Cohort StudyPresenter : Dr Himani
Moderator: Dr Subodh S Gupta
Types of Epidemiological study
1. Definition2. Study design3. Steps of study Identification of study population Measurement of exposure selection of study and comparison group Follow-up Analysis4. Advantages and disadvantages 5. Biases6. Variants of cohort study7. Cohort study and other study designs 8. Role of Cohort study in Epidemiology
Framework
The cohort study is an observational analytical epidemiological study which, after the manner of an experiment, attempts to study the relationship between a purported cause (exposure)and the subsequent risk of developing disease.
Synonyms: Incidence studies, prospective studies, follow-up studies, longitudinal studies
Definition
Study design
When there is good evidence of exposure and disease.
When exposure is rare and incidence of disease amongst exposed group is more
When follow-up is easy
When sufficient funds are available
Indication of Cohort study
Open Cohort: Subjects can leave and enter the cohort at any time
Closed Cohort : Fixed number of persons are followed over a specified time till the end point.
Types of cohort
Identification of study population
Measurement of exposure
Selection of study and comparison group
Follow-up (for outcome measurement)
Analysis
Steps of Cohort study
Free of disease of interest
Equally susceptible to development of the disease at the beginning of the study
Equivalent information (quantity and quality) should be available on exposure and disease
Both the groups should be accessible and available for follow- up.
Criteria for the selection of subjects in cohort
Internal comparison◦ Single cohort enters the study◦ Later on, classified into study and comparison
cohort based on exposure
External comparison◦ More than one cohort identified ◦ e.g. Cohort of radiologist compared with
ophthalmologists or smokers with non smokers
Comparison with general population rates◦ If no comparison group is available we can
compare the rates of study cohort with general population
◦ Frequency of cancer amongst asbestos workers compared with cancer in general population
Selection of Comparison group
Existing Records
Interviews, self administered questionnaire, mailed questionnaire
Examinations: medical and other special examinations
Measure of environment: measure of air pollution, exposure to radiation or other toxicological substances
Gathering data on exposure
Objectives of Follow up uniform and complete follow-up of all cohort groups complete ascertainment of outcome events standardized diagnosis of outcome events.
Methods: correspondence with the subject and other
informants, periodic re-examination of the subjects, and indirect surveillance of hospital records and death
certificates.
Follow up
The estimated sample size n for cohort studies is calculated as
N= [Zα √(1+1/m)p(1-p) +Zβ √ {p0(1-p0) / m} +p1(1-p1)]2
(po-p1)2
Where
m = number of control subjects per experimental subject
po = probability of events in control
p1 =probability of events in experimental subjects
p= p1+ m po
m+1
Sample size calculation
Sample Size calculation
n=sample size µ1= population mean in group1 µ2= population mean in group2µ1- µ2 = difference the investigator wishes to detectσ = population variance (SD)a= conventional multiplier for alpha =0.05b= conventional multiplier for power = 0.80
When mean is available
Relative risk or risk ratio
Risk difference
Attributable risk
Population Attributable Risk
Rate Ratio
Analysis of Cohort study
Table: Summary of Risk Data from a cohort study
Disease No Disease Total
Exposed a b a+b
Not exposed c d c+d
Total a+c b+d a+b+c+d
Incidence (exposed) = Exposed persons who developed ds = a/(a+b) All exposed person Incidence (Unexposed) = Unexposed persons who develop ds = c/(c+d) All unexposed persons
Relative Risk = _Incidence (exposed)__ = _a / (a+b)_ Incidence (unexposed) c / (c+d)
Risk difference = Incidence (exposed) – Incidence(nonexposed) = [a/(a+b)] – c/(c+d)]
Attributable Risk = Risk(exposed) – Risk(nonexposed) * 100 Risk (exposed) = (RR-1) * 100 RR
Population Attributable Risk= Risk(Total) – Risk(unexposed) *100 Risk (total)
Table: Summary of Risk Data from a cohort studyDeveloped
diseaseDo not develop
diseaseTotal
Exposed 200 9800 10,000Not exposed 100 9900 10,000
Total 300 19700 20,000
Incidence (exposed) = Exposed persons who develop ds = 200/10,000 = 0.02 All exposed person Incidence (Unexposed) = Unexposed persons who develop ds = 100/10,000 =0.01 All unexposed persons
Relative Risk = _Incidence (exposed)__ = _0.02 = 2 Incidence (unexposed) 0.01
Risk difference = Incidence (exposed) – Incidence(nonexposed) = 0.02 – 0.01 = 0.01
Attributable risk: {(2-1)/2}*100 = 50%
Population attributable risk= {(0.015- 0.01)/0.02}*100 = 25%
Cumulative incidence Number of new cases of disease occurring over a
specified period of time in a population at risk
Incidence density Number of new cases of disease occurring over a
specified period of time in a population at risk throughout an interval
Table: Summary format of Rate data from a cohort study
Exposed Unexposed Total
Number of outcomes
a b a+b
Person time(PT)
PT(exposed) PT(unexposed) PT(total)
Rate ratio=rate of outcome amongst exposed/ rate of outcome among unexposed
Rate ratio= [a/ PT(exposed)] / [b/PT(unexposed)]
Statistical method for analyzing longitudinal data on occurrence of events.
Events can be Time to death time to onset of disease length of stay in hospital viral load measurement
Survival Analysis
Life table approach
Kaplan Mier
Cox Regression
Survival analysis
Clinically suspected casesEven
t fr
ee s
urv
ival
Direct calculation of risk ratio (relative risk)
Yield information on incidence of disease
Clear temporal relationship between exposure and disease
Particularly efficient for study of rare exposures
Can yield information on multiple outcomes of a particular exposure
Minimize bias
Strongest observational design for establishing cause and effect relationship
Advantages of cohort study
Time consuming
Often requires a large sample size
Expensive
Not efficient for study of rare diseases
Losses to follow-up may diminish validity
Changes over time in diagnostic methods may lead to biased results
Disadvantages of cohort study
Selection bias
Information bias/ Misclassification bias
Bias due to confounding
Post hoc bias.
Biases
the group actually studied does not reflect the same distribution of factors (such as age, smoking, race, etc.) as occurs in the general population.
This is due to members refuse to participate records are not available
Particular subgroup may not be representative of general population
Resolution: by careful selection of individuals for inclusion in the study and by making every attempt to characterize differences that may exist between respondents and non-respondents.
Selection Bias
Disease No disease
Exposure +
100 900 1000
Exposure - 50 950 1000
150 1850 2000
Selection Bias in historical cohort
True RR =2.0
20% of employee heath records were lost or discarded, except in “solvent” workers who reported illness (1%loss)
Disease No disease
Exposure +
99 720 819
Exposure - 40 760 800
139 1480 1619
RR=2.42
Some members of the original cohort drop-out of the study
If the loss to follow-up occurs equally in the exposed and unexposed groups, and their characteristics are nearly similar, the internal validity should not be affected.
Resolution: by intensive follow up comparison of baseline characteristics of those who
are loss to follow up and those who are not
Follow- up Bias
Differential loss to follow up in a prospective cohort study on oral contraceptives and thromboembolism(TE)
Without losses
TE Normal Total
OC+ 20 9,980 10,000
OC- 10 9,990 10,000
RR= 2 (truth)
After 40% loss to follow up
Final sample
TE Normal Total
OC+ 8 5,980 5,988
OC- 8 5,990 5,998
RR=1
Also known as ‘misclassification bias’
result from measurement errors, imprecise measurement, and misdiagnosis for whatever reason.
Resolution: by using well-defined precise measurements and classification criteria for which the sensitivity and specificity have been determined.
Information Bias
use of data from a cohort study to make observations that were not part of the original study intent.
These findings should be treated as hypotheses that are an appropriate subject for additional studies.
Post hoc bias
When other factors that are associated with both the outcome and exposure variables do not have the same distribution in the exposed and unexposed groups.
Confounder should have 3 criterias: it should be cause of disease it is associated with risk of disease it should not be an intermediate step in causal
pathway between exposure and disease
Bias due to Confounding
During design of study randomisation restriction matching
During analysis stratification statistical modelling
Nested Case-Control study
Variant of Cohort study
Nested Case-control study
Cohort study Case control study
Study group Exposed persons Persons with disease
Comparison group Unexposed persons Persons without disease
Outcome measurements Incidence in exposed group and incidence in unexposed
Proportion of cases exposed and proportion of controls exposed
Measures of risk Relative riskAttributable risk
Odds ratio
Best when Exposure is rare Disease is rare
Multiple associations Possible to study association of an exposure with several diseases
Possible to study associations of a disease with several exposures or factors
Cohort study and other studies in epidemiology
Cohort study Randomized control trial
Exposure is natural Exposure is not natural and decided by investigator
Randomization is not there in cohort studies
Randomization if heart of RCT
Natural history of the disease: For example, cohort studies of individuals who were
infected with HIV revealed that a drop in the level of T lymphocytes having the CD4 marker was associated with being infected with HIV, and that a further decline in CD4 cells was associated with developing clinical symptoms and AIDS
Causation of the disease: eg. role of high blood pressure as a major cause of
stroke, myocardial infarction, and chronic kidney disease. Evaluating public health programmes: If relevant statistics are not routinely available, cohort
studies can be used to assess whether the programme has an impact by calculating the incidence of a disease.
Role of cohort study in epidemiology
1. Gordis L, Epidemiology, Third Edition(2004), Elsevier Saunders.2. Detels R, McEwen J, Beaglehole R, Tanaka H, Oxford Textbook
of Public Health, Fourth edition(2002), Oxford University Press3. Beaglehole R, Bonita R, Kjellstrom T, Basic Epidemiology, 2nd
Edition(2006)WHO.4. Hill, Mc Graw, Medical Epidemiology, 4th Edition(2007), Lange.5. Bhalwar R, Vaidya R, Tilak R, Gupta R, Kunte R, Textbook of
Public Health and Community Medicine. 1st Edition (2009).6. Kasiulevičius V, Šapoka V, Filipavičiūtė R, Sample size
calculation in epidemiological studies, Gerontologija 2006; 7(4): 225–31
7. Grimes DA, Schulz KF, Cohort studies: marching towards outcome, Lancet2002;359:341-45.
References