104
How to Develop Research Proposal for Original Study Alongside Economic Evaluation: HTA program 06/06/2018 Ammarin Thakkinstian, Ph.D. E-mail:[email protected] www.ceb-rama.org
How to Develop Research
Proposal for Original Study
Alongside Economic Evaluation
HTA program 06062018
Ammarin Thakkinstian PhD
E-mailammarinthamahidolacth
wwwceb-ramaorg
Outline of talk
Proposal for conducting original study
bull Research question
bull Literature review
bull Background amp rationalendash Magnitude of problem
ndash How is it important and burden
ndash Previous evidences
ndash Research objectives
bull Methodologyndash Study design
ndash Participants amp settingbull Inclusion criteria
bull Exclusion criteria
ndash Randomisation (if RCT)
ndash Data collectionbull Variables amp measurement
ndash Study factor (Treatment protocol for RCT)
ndash Outcome factor
ndash Confounderscovaraibles
bull DRF
ndash Data managementbull Databases
bull Data analysis
bull Dummy tables
ndash Ethic consideration bull Information sheet
bull Inform consent
ndash Budget
ndash Time table
What can make a good research question
bull Interesting
bull Feasible (answerable with a robust method)
bull Novel
bull Ethical
bull Relevant to biological mechanismrationale
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
How to focus on your question
bull Quick literature search for previous evidences
bull Assess feasibility
bull Discuss and convince your team
bull Perform
ndash individual systematic review (SR) withwithout meta-
analysis (MA)
ndash Umbrella review of previous SRsMAsbull Locate relevant studies as many as possible
bull Medline Embase Scopus EndNote
bull Create search strategies
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Outline of talk
Proposal for conducting original study
bull Research question
bull Literature review
bull Background amp rationalendash Magnitude of problem
ndash How is it important and burden
ndash Previous evidences
ndash Research objectives
bull Methodologyndash Study design
ndash Participants amp settingbull Inclusion criteria
bull Exclusion criteria
ndash Randomisation (if RCT)
ndash Data collectionbull Variables amp measurement
ndash Study factor (Treatment protocol for RCT)
ndash Outcome factor
ndash Confounderscovaraibles
bull DRF
ndash Data managementbull Databases
bull Data analysis
bull Dummy tables
ndash Ethic consideration bull Information sheet
bull Inform consent
ndash Budget
ndash Time table
What can make a good research question
bull Interesting
bull Feasible (answerable with a robust method)
bull Novel
bull Ethical
bull Relevant to biological mechanismrationale
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
How to focus on your question
bull Quick literature search for previous evidences
bull Assess feasibility
bull Discuss and convince your team
bull Perform
ndash individual systematic review (SR) withwithout meta-
analysis (MA)
ndash Umbrella review of previous SRsMAsbull Locate relevant studies as many as possible
bull Medline Embase Scopus EndNote
bull Create search strategies
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Methodologyndash Study design
ndash Participants amp settingbull Inclusion criteria
bull Exclusion criteria
ndash Randomisation (if RCT)
ndash Data collectionbull Variables amp measurement
ndash Study factor (Treatment protocol for RCT)
ndash Outcome factor
ndash Confounderscovaraibles
bull DRF
ndash Data managementbull Databases
bull Data analysis
bull Dummy tables
ndash Ethic consideration bull Information sheet
bull Inform consent
ndash Budget
ndash Time table
What can make a good research question
bull Interesting
bull Feasible (answerable with a robust method)
bull Novel
bull Ethical
bull Relevant to biological mechanismrationale
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
How to focus on your question
bull Quick literature search for previous evidences
bull Assess feasibility
bull Discuss and convince your team
bull Perform
ndash individual systematic review (SR) withwithout meta-
analysis (MA)
ndash Umbrella review of previous SRsMAsbull Locate relevant studies as many as possible
bull Medline Embase Scopus EndNote
bull Create search strategies
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
What can make a good research question
bull Interesting
bull Feasible (answerable with a robust method)
bull Novel
bull Ethical
bull Relevant to biological mechanismrationale
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
How to focus on your question
bull Quick literature search for previous evidences
bull Assess feasibility
bull Discuss and convince your team
bull Perform
ndash individual systematic review (SR) withwithout meta-
analysis (MA)
ndash Umbrella review of previous SRsMAsbull Locate relevant studies as many as possible
bull Medline Embase Scopus EndNote
bull Create search strategies
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
How to focus on your question
bull Quick literature search for previous evidences
bull Assess feasibility
bull Discuss and convince your team
bull Perform
ndash individual systematic review (SR) withwithout meta-
analysis (MA)
ndash Umbrella review of previous SRsMAsbull Locate relevant studies as many as possible
bull Medline Embase Scopus EndNote
bull Create search strategies
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
ndash Data extractions
ndash Pooling data if possible
ndash Critical appraisal of previous studiesbull Study designs
bull Variables and measurements
bull Interventions comparators
bull Outcomes
bull Results
bull Can answer your interested research question
bull Limitation of those studies
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Summarizing review results
bull Identify a gab of knowledge
bull Rationale for conducting your research
bull Plan to do better
bull Conceptual framework
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Turning a research question into a proposal
Things to consider bull Whom will be your studied patients
bull What kinds of data do you need
bull How large data will you require
bull Where will you get data
bull How will you collect the data ethically
bull How will you minimise chancebiasconfounding
bull How will you analyse data to answer research question
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Turning down your research question to be
what you are exactly planning to do
PICO
bull P - who are the studied patientsndash Disease event condition
ndash Stage severity
bull I - what isare the interventions or study
factorsndash RCTs
bull Type of interventions
bull Dosage frequencyday course route
ndash Observational study
bull Exposure versus non-exposure
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull C ndash what is the comparator
ndash Placebo
ndash Standard treatment
ndash Other active treatments
bull O - what are the interested outcomes
clinical endpoints
ndash Primary outcomes
bull Efficacy
bull Safety
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Secondary outcomes
ndash Surrogate or intermediate outcome
bull Type of outcome data
ndash Dichotomous outcome
ndash Continuous outcome
ndash Time to event
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Introduction and rationale bull What is the problem
ndash Magnitude of problem prevalence incidence
bull How is it burden ndash Morbidity complications disability QoL cost of
treatment for individual andor country levels
bull What are evidences available ndash Previous evidences
bull Individual studies
bull SRsMAs
bull Umbrella reviews of previous SRsMas
ndash Conduct your own review
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Why do we need further evidences ndash To do not replicate with previous studies
ndash A gab of knowledge
ndash Previous studies could not answer the question
ndash Limitation of previous studies bull Poor methodology
bull Did not consider important interventionsexposures
bull Did not consider clinical end outcomes
bull How can we do better
bull What new evidence will be added
bull TRY TO CONVINCE FUNDING AGENCYIRB AS MUCH AS POSSIBLE
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Methodologybull Study design
ndash RCT
ndash Cohort study
ndash Case-control study
ndash Cross-sectional study
ndash Descriptive study
bull Setting
bull Period of study
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Will you observe or experiment
bull Observe ndash Cross sectional case-control studies cohort
studiesbull Patients
bull Observe and collect characteristics
bull Associations
bull Experimentbull Parallel Randomised controlled trials (RCT)
bull Cross-over RCT
bull
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Cross-sectional studies
bull Frequency or prevalence survey
ndash Health National-Survey
ndash Thai-Seek project
bull Aim to
ndash Estimate prevalence of disease
bull Hypertension T2D Dyslipidemia CKD
bull Assess association between study factors and
diseases
bull Diagnostic test accuracy
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull All variables (exposures and outcomes)
are measured only once at a particular
pointrange of time
bull May require a sampling technique for
survey study
ndash Simple random sampling
ndash Systematic random sampling
ndash Stratified random sampling
ndash Cluster sampling
ndash Stratified-cluster random sampling
bull
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Pros and cons of cross-sectional study
Advantages
bull Cheap and simple
bull Ethically safe
Disadvantages
bull Establishes association at most not causality
bull Recall bias
bull Group sizes may be unequal
bull Confounders may be unequally distributed
Trish Groves Deputy editor Different types of clinical evidence and study design BMJ group
Accessed 24-92017
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Case-control bull Outcome already occurred
bull Unmatched versus matched case-control
bull Start with selection of cases and controls ndash How to identify cases and control
ndash Definition of case and ascertainment
ndash Definition of control and ascertainment
bull Obtain information whether subjects had
been exposed to interested factors
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Advantages
bull Only feasible design for very rare disease or
long term disease onset
bull Quick and cheap may required lower sample
size than cross-sectional study
Disadvantages
bull Have only one outcome of interest
bull Confounders
bull Selection of control groups is difficult
bull Potential bias recall selection
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Cohort studies
bull Proceed in a logical sequence
ndash Exposureoutcome
bull Subjects with or without exposure of
interested factors are identified at the
beginning
bull Interested factors can be gt 1 factor
bull Subjects are then followed-up until disease or event occurs
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Type of cohorts
bull Retrospective cohort
bull Prospective cohort
bull Bidirectional cohort
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
PresentTime
PastFuture
Retrospective Prospective Historical
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Strength
bull Causal relationship can be claimed
bull Can have more than one outcome
bull New cases incidence can be estimated
bull Recall bias is less likely
Weakness
bull Rare event is required very large sample size
bull Long term disease-onset requires long term
follow-up
bull Expensive
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Parallel RCT
bull Participants are randomly allocated or
randomised to receive
treatmentintervention or standard
treatmentplacebo
bull The best for studying the effect of
treatmentintervention
bull Can balance known amp unknown
riskprognostic factors between
treatmentintervention groups
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Pros and cons of the RCT
Advantages
bull Controladjust for confounders
bull Blinding may be applicable
Disadvantages
bull Expensive in both time and money
bull Volunteer bias
bull Ethically problematic at times
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Sequence generation method
bull Simple randomisation
bull Block randomisation
bull Stratification block randomisation
bull Good randomisation technique
bull Unpredictable
bull Repeatable
bull Auditable
bull
Randomisation technique
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Simple randomization
Random number list92612 19036 59921 86723 50318 29715 26308 94840 27961 8397531650 10804 49276 74103 80429 47242 03448 51271 29745 7087405211 69788 29346 96431 33172 22474 17227 58400 57297 2182644236 27790 91491 06683 26116 66103 32217 79052 46429 3370969081 00820 58799 94787 78338 04984 61901 86054 35234 0115545650 41433 06142 79961 32014 89878 06416 46079 58892 6852934907 34555 80506 18379 21074 90125 82719 54658 45635 1420636743 29107 20687 09246 39661 26334 40234 75131 15200 6143637071 37332 73198 71088 06945 11467 51206 15266 85939 3398986637 31805 92023 13669 92942 10984 66697 83562 38217 82306
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Simple randomization
bull Coin tossing
bull Shuffling cards or envelopes
bull Throwing dice
bull Drawing of lots
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Sequence generation (cont)
bull Block randomisation OR
bull Restricted randomisation OR
bull Balance randomisation
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Block randomisation (cont)ndash Number of subjects
ndash Number of Rx
ndash Ratio Rx vs Placebo
ndash Block sizebull The smallest block size = No of Rx andor ratio
bull Numbers of block
ndash Should be varied
ndash Small block size next Rx allocation can be guessed breaking the blind
ndash Too many large blocks may cause imbalance of treatments in case of premature termination
ndash Seed number if using computer software bull Reproducibility
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Example of designing block size
bull Rx=2 ratio 11 block size = 2 46
bull Rx = 3 ratio 111 block size = 36 9
bull Rx=2 ratio 12 block size = ratio= 369
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Participant and setting
ndash Eligibility criteria
bull Inclusion
bull Exclusion
ndash Setting amp period of study
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Method (cont)
bull Treatmentintervention protocol (For RCT)
ndash Describe detail of treatmentsinterventions
bull Dosage frequency course of treatment
ndash How it will be administered
ndash By whom
ndash When
bull
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Study variables and measurements
For observational studies
bull Describe all interested variables
ndash Study factors
ndash Co-variables
bull Howwhen to measure
bull Are they varied or fixed
bull How many times will they be measured
bull Continuous or categorical data
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Outcome of interests
bull Primary outcome
bull Secondary outcomes
ndash Howwhen to measure
ndash Definition or criteria use for diagnose with
references
ndash Whom will be involved in measurements
ndash When to measure
ndash Blinding
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Sample size estimation
bull Describe how sample size is estimated
bull For hypothesis testing
ndash Type I amp II error
ndash Size of detectable
bull Minimum size that is clinically meaningful
ndash Basic information control group ndash Mean (SD) headache score in control
ndash Cure rate in standard treatment
ndash Median disease free time in the standard treatment
ndash Mortality rate in the standard treatment
ndash Strongly recommend to pool data using meta-analysis
ndash Size of studied population (feasibility of recruitment)
raquo Incideneprevalence
ndash
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Ho microSBP(A) = microSBP(B)
Ha microSBP(A) ne microSBP(B)
Statistical Decision In Population
Based on Sample Ho True Ho False
-----------------------------------------------------------
Reject (Positive) a (FP) 1-b (Sen)
Fail (Negative) 1-a (Spec) β(FN)-----------------------------------------------------------
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Methodology (cont)
bull Data collection forms
ndash Include only relevant variables into the form
ndash Collect only relevant variables
ndash Collect variables as their actual measurements
bull Time duration date start date end
bull Age birth date to current date
bull BMI height weight
bull DM HT Dislipidemia BS BP Chol
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
ndash Design all variables as domains
bull Demographic data
bull Riskpreventive behavior
bull Environmental variables
bull Clinical sign amp symptoms
bull Treatments
bull Physical examination
bull Lab
bull Radiography
bull Outcomes
ndash Short intermediate long terms
ndash Efficacy safety
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
For follow-up study
ndash What variables are constant and varied
over time
bull Enrollment form
ndash Contains all variables
ndash Domain amp sequence
bull Follow-up
ndash Time varying variables
ndash Outcome assessments
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Methodology (cont)
bull Data management
ndash Databases
bull Data quality control
bull Data cleaning
bull Software
bull Statistical analysisbull Describe all detail what statistics are going to apply for
bull Descriptive statistic
bull Analytic statistics
bull Software
bull Level of significance
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Statistical analysis plan
(SAP)bull Data management
ndash How to entry
ndash Data checking amp validating
bull Descriptive analysis
ndash How to describesummary data
bull Univariate analysis
bull Muti-variate analysis
bull Software
bull P value
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Methodology (cont)
Dummy tables bull Imagine how results will be reported
bull Depends on study design
bull Follow step of data analysis
bull Table 1
ndash Describe general characteristic of subjects for cohortcross-sectional studies
ndash Describe amp compare general characteristic of subjects between treatmentcase and control groups for RCTcase-control studies
bull Table 2
ndash Describe amp compare general characteristic of subjects between disease versus non-disease groups for cohortcross-sectional studies
ndash Assessment of treatment effects for RCT effect of riskprognostic factors adjusting for confounders for observational studies
bull Table 3
bull Table 4
bull Figure 1
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Budget
ndashPer-patient cost
ndashStudy-level cost
ndashIndirect cost
ndashTotal costs
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Ethic consideration
ndashInform consent
ndashInformation sheet
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Time frame bull Proposal phase
bull Study-phase
ndash Conducting study
ndash Data management
ndash Analysis
ndash Writing
bull Reports
bull Manuscripts
ndash Publication
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Proposal developing phase
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Activity 2009 2010 2011
Aug Sep Oct Nov Dec Jan Feb Mar Apr May June Jul Aug Sep Oct Nov Dec Jan Feb Mar
Enrollment
Data collection
Monitor and audit
CRF transferring
Data entry
Investigator meeting
Cleaningamp checking data
Data analysis Enrollment Enroll + FU
Report
Manuscript I
Manuscript II
Study-phase
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Guideline for reporting studiesrsquo results
From EQUATOR network httpwwwequator-
networkorg
bull CONSORT for randomised controlled trials
bull STARD for diagnostic accuracy studies
bull STROBE for observational studies
bull MOOSE for meta-analyses of observational
studies
bull CHEERS for health economic evaluation
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Economic evaluation alongside
originalindividual study
bull Cost effectiveness study is incorporated
with
ndash RCT
ndash Cohort
ndash Cross-sectional study
bull Aims to compare ndash Delta cost (new versus old treatments)
ndash Delta benefit (efficacy of new versus old treatments)
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Incremental cost effectiveness
ratio (ICER)
E
C
1E1E
2C1C
ˆ
ˆ
)ˆˆ(
)ˆˆ(ICER
Costs and effectiveness will be collected from all
individual patients then average them and plug into the
ICER equation
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Design of RCT based economic
evaluation
bull Research team
ndash Methodologists specialised in RCT
ndash Biostatistician
ndash Clinical expert
ndash Health Economist
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Standard research methods for RCTs
must be applied
bull Good research question
bull Rationale for conducting RCT
ndash May need to do systematic review
withwithout meta-analysis
ndash Identify a gab of knowledge
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Methodology
ndash Study design
ndash Participant and setting
bull Inclusionexclusion criteria
ndash Randomisation
ndash Treatment protocol
ndash Outcome of interests
bull Primary and secondary outcomes
ndash Data collection
bull Study recruitment flow
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
ndash How
ndash When
ndash Case record forms
ndash Data monitoring
ndash Data safety and monitoring broad (DSMB)
bull Data management
ndash Databases
ndash Data analysis
ndash Dummy table
bull Ethic consideration
bull Budgets amp time frame
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Comparison of Superficial Surgical Site
Infection between Delayed Primary
versus Primary Wound Closure in
Complicated Appendicitis A
Randomized Controlled Trial
bull Ann Surg 2017 Aug 4 doi 101097SLA0000000000002464
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Introduction and background
bull Appendicitis is a common surgical emergency
the rate of appendectomy in a Korean study was
14 per 10000 population per year of which 21
was for complicated appendicitis (ie
gangrenous and ruptured)
bull Superficial surgical site infection (SSI) is a common complication (ie 9 to 53 2) in this condition as compared to simple appendicitis3 and adversely affects both patients and the health care system4
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Delayed primary wound closure (DPC)
ndash Firstly introduced in World War I5
ndash The wound is left open and sutured on
postoperative day 3 to 5
ndash Increasing blood supply6 and oxygen7 at the
surgical site
ndash Decrease SSI in other contaminated wounds9 10
ndash It is invasive with dressing changes and re-suturing and increases length of stay and cost of treatment11
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Primary wound closure (PC)
ndash The wound is sutured immediately after
operation
ndash Given advances in antibiotics and
perioperative care over recent years12 we
questioned whether DPC is still required as a
routine practice
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
A systematic review and meta-analysis of
randomised controlled trials of delayed
primary wound closure in contaminated
abdominal wounds
World J Emerg Surg 2014 9(1)49
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Six RCTs
bull n = 234 vs 182 for PC vs DPC
bull Pooled SSI rates 23 vs 26
bull However ndash These included RCTs were of low quality
ndash The pooled effect was imprecise small
numbers of RCTs and subjects
bull Although DPC is still debated it is still the
current standard of practice
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Drawbacks of DPC
ndash Pain from dressing
ndash Required suturing afterward thus 1 to 2 days
longer hospital stay5 22
ndash Higher cost of treatment
bull As high as $400 per one DPC in Canada22
bull About 2500 (3 dressing changes 1 suturing and
1-2 days hospital stay) baht in Thailand
bull We conducted a RCT
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Objectives
Primary Objective
bull To compare the rate of postoperative SSI
in complicated appendicitis between PC
and DPC
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Secondary Objectives
bull To compare postoperative pain scores at days 1
3 and 5 after appendectomy between primary
and delayed primary wound closure
bull To compare quality of life at postoperative days
3 5 and 30 after appendectomy between
primary and delayed primary wound closure
bull To do cost-utility analysis between primary
and delayed primary closure in complicated
appendicitis
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
MethodsStudy design and setting
bull A multicenter-parallel-RCT was conducted
from November 2012 to February 2016
bull Six hospitals in Thailand ndash Two university hospitals (Thammasat and Ramathibodi hospitals)
ndash Pathum Thani (the Central region)
ndash Chonburi (the Eastern region)
ndash Surin (the North-Eastern region) and
ndash Lampang (the Northern region)
ndash This study was approved by the Ethics committee of Ramathibodi
Hospital and the collaborating hospitals
ndash Registered at ClinicalTrialsgov (ID NCT01659983)
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Participants Preoperative eligibility criteria for step I
bull Age ge18 years
bull Appendectomy with a right lower quadrant abdominal incision
bull Not
ndash Pregnant
ndash Very obesity (BMI ge40 kgm2)
ndash Auto-immune diseases end-stage renalliver disease or HIV
bull Providing informed consent
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Intraoperative eligibility criteria step II
bull Gangrenous appendicitis
ndash Erythematous or swollen appendix and
ndash Necrotic wall (dark grayish color)
bull Ruptured appendicitis
ndash Appearance of hole in an appendix
ndash Rupture during the procedure or
ndash Presence of frank pus3
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Randomization
ndash Balance known amp unknown prognostic factors
on average evenly across intervention
groups
ndash Good randomisation
bull Unpredictable
bull Repeatable
bull Auditable
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Randomizationbull Permuted blocks of 4 to 6
bull Stratified by hospital
bull Assignment provided by an independent
statistician using STATA version 120
bull The random sequences will be kept
confidential at the central data
management unit (DMU) at the section of
Clinical Epidemiology and Biostatistics
Faculty of Medicine Ramathibodi Hospital
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
NT=2 ratio 11 No blocks = 2 initial block size of 4
ralloc bn bs Rx nsubj(600) nt(2) osize(2) ra(1) init(4) seed(23456) saving(PC-DPC) trtlab(PC DPC)
Rx | Freq Percent Cum
------------+-----------------------------------
PC | 301 5000 5000
DPC | 301 5000 10000
------------+-----------------------------------
Total | 602 10000
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Allocation concealment bull Concealment is aimed to prevent selection
bias before randomization
bull Sequentially numbered opaque sealed
envelopes will be used to conceal the
randomization lists
bull The envelopes will be created by an
independent statistician who will not be
involved with the study
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Allocation concealment
bull The random number will be sequentially
specified within each envelope and will be
delivered to the local research sites
bull The sealed opaque envelopes will be
opened by scrub nurses intra-operatively
prior to closure of the appendectomy
incision
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Blinding bull Patients Investigators outcome assessors
bull Neither patients nor attending physicians were
blinded to intervention due to obvious
constraints
bull Superficial SSIs were assessed using standard
criteria and monitoring procedures to minimise
ascertainment bias
bull However research assistants involved in
subjective outcome assessments (ie recovery
period pain and QoL) were blinded
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Interventionsbull Appendectomy and wound closure were done by
surgical staff or surgical residents under supervision
bull For PC the wound was closed immediately after the
operation using a non-absorbable monofilament
suture or stapler
bull Dry dressings were applied daily until stitches were
removed
bull For DPC the wound was left open with twice daily
saline-soaked gauze and closed on postoperative
day 3 to 7 using the same suture as PC
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Co-interventionsbull Co-interventions were standardised
ndash Wound dressings and closed suction drains
ndash Antibioticsbull Cover enteric gram-negative and facultativeanaerobic bacilli14
bull Prepost-operative intravenous antibiotics until body temperature was lt 378degc for 24-48 hours
bull Oral antibiotics for 7-10 days
ndash Pain controlbull Intravenous opioids
ndash Morphine 3-5 mg or pethidine 25-50 mg as requested every 4 hours
ndash Paracetamol or non-steroidal anti-inflammatory drugs after resuming an oral diet
ndash
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Outcome of interests The primary outcome was superficial SSI
ndash Defined using the Center for Disease Control (CDC)
criteria15
ndash Occurring within 30 days
ndash Only skinsubcutaneous tissue involvement
ndash Had one of the following
bull Purulent drainage organism isolated from fluid or tissue
bull Or at least one of the following signssymptoms
ndash Pain or tenderness
ndash Localized swelling
ndash Redness or heat
ndash Superficial incision was deliberately opened by surgeon without a
positive culture
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Assessed by the responsible physician before
discharge at 1 week and 1 month follow up
bull If patients did not visit the outpatient clinic a
standardized telephone interview was used to
ascertain wound swelling pain discharge and
any patient visits to other hospitalphysicians
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Secondary outcomes
bull Postoperative pain
ndash 0-100 visual analog scale (VAS) at day 13
ndash Cumulative dosage of opioids was also
recorded
bull Length of stay
ndash Admission date
ndash Discharge date
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Treatment costs ndash Direct medical costs from PC and DPC (LOS
dressing and re-suture) were estimated from
Thailandrsquos standard cost17
ndash Direct non-medical costs
bull Informal care included work time lost by caregivers
bull Transportation cost that patients caregivers spent
travelling to and from hospital
bull Thailandrsquos minimum wage of 300 Bahtday was
used for calculation
(httpwwwmolgothanonymousehome)
ndash Indirect treatment costs (income lost) were obtained
by interview using standard forms
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Sample size estimation
bull The pooled superficial SSI rate in DPC was 295 (95 confidence interval (CI) 148 442) based on 3 previous RCTs18-20
bull The type-I error power and ratio were set at 005 (two-sided) 080 and 11 respectively
bull A sample size was estimated based on a difference of plusmn10
bull A total of 570 patients (285 per group) was needed Taking into account loss to follow up of 5 600 patients were set as the target
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Data collectionbull Case record forms (CRF)
ndash Enrollment
bull Demographic data
bull Sings and symptoms
bull Physical examinations
bull Lab
bull Imaging
ndash Operation
ndash Follow up and outcome
ndash
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Data collectionbull The central DMU at CEB will prepare
CRFs and distribute CRFs to all
collaborative sites
bull Training
ndash Operative nurses will be trained for data
collections
bull Research project informed consent and
randomization process
bull
bull
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Flow of data collection
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Data management
bull A centralized data management system
bull All CRFs will be delivered to the central
DMU
bull Problem resolution and data queries will
be resolved by research assistants
principal investigator and local
investigators at each research site
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull The databases will be constructed based on the
CRFs using EPIDATA version 31
ndash Enrollment operation follow up and outcome
databases
bull PI will check for the completeness of CRFs before
performing data entry
bull The data will be double entered by two independent
staffs
bull The two data sets will then be validated to detect
typing errors Cleaning and checking of data will be
performed every month
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Any unclear answer missing or non-sensible
data will be inquired to the local investigators
bull Data will be kept in a safe area
bull Only principal investigator (BS) supervisor (AT)
and the data manager will be able to access to
the data
bull Data cleaning and checking will be performed
every week DMU meeting will be organised
once a month
bull The data will be automatically real-time backed
up at the DMU server to prevent loss of data
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Data safety and monitoring board (DSMB)
bull Responsibility
ndash Review and evaluate the accumulated study data concerning
participantrsquos safety study conduct and progress
ndash Make recommendations to the research team concerning the
continuation modification or termination of the trial
ndash The DSMB will be constructed
bull Surgeon Epidemiologist and
bull Biostatistician
bull They are independent and not involved in the study
bull Validation of the data including rate of recruitment process of the data
collection quality of the data (ie informed patient obtained consent form
randomization postoperative care and outcome assessment) will be
audited
bull Any adverse reaction that might be caused by the studied interventions will
be checked
bull Interim analysis will be explored and suggested by the DSMB
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Statistical analysisbull The primary analysis will be performed by
intention to treat (ITT)
bull All patients will be analysed according to
the groups to which they are originally
randomised accounting for loss to follow-
up (modified ITT)22
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Three additional post-hoc analyses will be
also performed22 23
ndash A per-protocol (PP) analysis will include only patients
who actually received and completed the randomly
assigned interventions
ndash As-treated (AT) analysis will include patients
according to the intervention actually received
bull A counterfactual approach using instrumental variable
(IV) analysis will be applied
ndash Assessing what outcome will have been seen for those
patients who dodo not comply with the assigned
intervention
ndash This is known as potential outcome mean
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Baseline characteristics of the patients will be
described by intervention groups
bull SSI
ndash Binary regression will be used for comparing superficial SSI
between groups
ndash SSI incidence risk difference and risk ratio (RR) along with
95CI will be estimated
bull Continuous outcomes
ndash Linear or quartile regression analysis will be applied to
compare LOS and return to normal activitieswork between
PC and DPC groups where appropriate
ndash For pain and QoL scores
raquo A mixed-effect regression model will be used to compare
mean difference (MD) between groups
raquo The analysis of pain was adjusted for the total opioid
dose
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull For the counterfactual approach
ndash IV regression will be applied considering
bull The assigned intervention as the IV
bull The actually received intervention as endogenous
variable
bull Type of models
ndash Bivariate probit for SSI
ndash A two-stage least square regression for pain score QoL
ndash Adjusting for covariates (eg age sex BMI smoking
ASA classification diabetes hypertension onset time
type of appendicitis WBC count body temperature and
use of drain)
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Imputation
ndash Missing data will be imputed using a multi-chain
simulation-based approach25 26
ndash Assuming data were missing at random (MAR)
ndash Multi-chain equations bull A linear truncatedinterval regression for continuous data
bull Ordinal logit and logit equations for ordinal and categorical
data
bull Twenty imputations
ndash Performances
bull Fraction of missing information (FMI)
bull Relative variance increase (RVI)
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Economic evaluation
Societal perspective costs
bull Direct medical costs
ndash Drug Medical supplies laboratory test
medical fees operative fee hospital charges
nursing care
ndash Fees will be based on the price lists of each
collaborative hospitals
bull Direct non-medical (carersquos giver food etc) and
indirect costs from sick leave will be based on
Health Intervention and Technology Assessment
Program (HITAP) study
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Provider perspectives
ndash Direct medical cost hospital charges drugs
home nursing
bull Utility score will be estimated from the
EQ5D according to the guideline for
Economic Evaluation in Thailand 43
bull Incremental cost effectiveness ratio
(ICER) will be calculated by dividing the
difference of cost by the difference of utility
between delayed PC and DPC
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Decision tree model will be used for cost
analysis
bull Discounting and inflation rate will be not
used because of the short time horizon of
the analytical model
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull Uncertainty analysis will be performed by
one-way sensitivity analyses
bull Tornado diagram will be constructed by varying each parameter at a time with 95 confidence interval
bull Probabilistic sensitivity analysis will be simulated using the Monte-Carlo method
with 1000 replications
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Figure 6 Decision tree for cost-utility analysis comparing wound
infection between PC and DPC in complicated appendicitis
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull All analyses for efficacy will be performed
using STATA version 140
bull Cost analysis will be performed using in
Microsoft Excel 2007
bull P value of less than 005 will be
considered as statistically significant
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
Ethic consideration
bull Patients will be informed by a surgeon
whorsquos going to operate ndash Study objectives and procedure
ndash Benefit and
ndash Risk of the study following information sheets
ndash Patients will be free to ask any unclear information related
to the study during the information process
ndash The information process will be take place in a comfortable
room (in a ward or preoperative waiting room) before an
operation
ndash Patients will be given about 10 ndash 15 minutes to make
decision whether they will participate in the study
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
bull If patients are not willing to participate in the study they
will be treated according to the standard treatments of
appendicitis of that hospital
bull The information of the patient will be recorded as in
observational studies with the permission of the patients
bull If they agree to participate in the study they have the
right to withdraw from the study at any time without any
influence to the treatment of the patients
![Case Report Stump Appendicitis: An Uncompleted Surgery, a ... · a er appendectomy, but our case presented only four and half ((/)) months a er laparoscopic appendectomy [, ]. With](https://static.documents.pub/doc/80x56/60df3a7ef0e58c30304e41fe/case-report-stump-appendicitis-an-uncompleted-surgery-a-a-er-appendectomy.jpg)
