10/20/2005 1
Customer Oriented CDISC Implementation
Edelbert Arnold and Ulrike Plank
Accovion, Germany
October 12, 2005
10/20/2005 2
Efficient Strategy in Clinical Development
Cornerstones:
CDISC recommendations
FDA expectations
Sponsor’s requirements
Accovion’s inventory
10/20/2005 3
Cornerstones (I)
CDISC recommendations:
FDA expectations:
• Study Data Tabulation Model (SDTM)
• Analysis Dataset Model (ADaM)
• SDTM and ADaM compliance recommended (less strict for ADaM)
• Documentation (metadata and links, preferably define.xml)
• Supplemental documents (annotated CRF, programs)
10/20/2005 4
Cornerstones (II)
Sponsor’s requirements
• In-house vs out-sourcing
• CDISC relevance
• Sponsor provided material:
CRFs only
CDMS based study data
Other source data (e.g. analysis data)
10/20/2005 5
Cornerstones (III)
Accovion’s inventory:
• Core functions
Database Programmers (DBP)
Statistical Programmers (SP)
• Software
CDMS: Clintrial (CT) and Oracle Clinical (OC)
Analysis software: SAS 8.2 and 9.1.3
10/20/2005 6
Accovion’s CDISC relevant tasks
Database set-up (DBP)
Generation of Full-SDTM and ADaM datasets (SP)
Integration of different data structures (SP)
Analysis output (SP)
10/20/2005 7
CT / OC“Pre-SDTM“
Annotated
CRF
Mappingprograms
SDTM
ADaM
Report, submissionAnalysis output
(tables, listings and graphs)
SAS macros, sample
programs
Submission to the FDA?
N
Standard tool
xpt
define. xml
Macro to create xpt files
YSponsor’s
dataData
Integration Tool
Our workflow
10/20/2005 8
Database set-up
Current situation:SDTM Implementation Guide (SDTM-IG) recommended by the FDA
Consequences:
Consider SDTM-IG
Fill gap, if CDMS is not 100% SDTM compliant
Implementation strategies differ between companies
10/20/2005 9
Strategy for Database Set-up
Use of metadata libraries based on SDTM
Efficient implementation of changes in data structure
Flexibility in case of complex derivations
Additional variables needed for data analysis
10/20/2005 10
Use global libraries including:
Clintrial /Oracle Clinical
data
Accovion’s strategy
(Pre-SDTM= SDTM pure+
SDTM comfort)Copy and adapt on study level
SDTM pure: SDTM variables and metadata
SDTM comfort: Additional variables foranalysis
CDMS generates a Pre-SDTM: Some variables will be derived in SAS
10/20/2005 11
SDTM comfort: Benefit in relation to SDTM pure
Numeric codes:
Date variables:
Used in analysis and listings
Used for convenient data selection and sorting
DSDECOD (SDTM pure) DSDECODN (SDTM comfort)
ADVERSE EVENT 1COMPLETED 2... ...NON-COMPLIANCE WITH STUDY DRUG 6... ...STUDY TERMINATED BY SPONSOR 13... ...OTHER 999
EGDTC (SDTM pure) EGDT+EGDC (SDTM comfort)
2005-10-12T09:15 12OCT2005 (16721)
12-OCT-2005
10/20/2005 12
SDTM and ADaM implementation
Topic SDTM ADaMTopic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Analysis friendly: “One PROC away”
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Analysis friendly: “One PROC away”
Characteristics Domain concept Analysis oriented
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Analysis friendly: “One PROC away”
Characteristics Domain concept
No redundancyAnalysis oriented
Common variables in each dataset
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Analysis friendly: “One PROC away”
Characteristics Domain concept
No redundancy
CRF data and trial design data
Analysis oriented
Common variables in each dataset
More derived data
Topic SDTM ADaM
Version 1.1 (SDS V3.1) General Considerations: 1.0, other models 0.x
Reviewer Medical Reviewer Statistical Reviewer
Requirements Standardization Standardization
Analysis friendly: “One PROC away”
Characteristics Domain concept
No redundancy
CRF data and trial design data
Textual results
Analysis oriented
Common variables in each dataset
More derived data
Numeric codes
10/20/2005 13
Possible strategies for SDTM and ADaM implementation
• Independent creation: CDMS ⇒ SDTM, CDMS ⇒ADaM
• Consecutive creation: CDMS ⇒ SDTM ⇒ ADaM
• Other concepts
CDMS (=SDTM) ⇒ ADaM
If CDMS not 100% SDTM compliant:
10/20/2005 14
Selecting a strategy
What is available in the CDMS?
Dual derivations
Inconsistencies
Reviewer’s needs
Potential of standardization
10/20/2005 15
CDMS DraftSDTM/ADaMData
Mapping
Data Mapping
Data Mapping
Note: SAS is used for Data Mapping
Accovion’s strategy
SDTM
ADaM
10/20/2005 16
SAS Post-Processing for Full-SDTM
Draft SDTM/ADaM ⇒ SDTM
• Delete comfort variables
• Add datasets, e.g. supplemental qualifier datasets SUPP~ (only if necessary)
CDMS ⇒ Draft SDTM/ADaM• Derivation of variables, e.g. treatment start
• Add records, where needed, e.g. baseline records
10/20/2005 17
ADaM Implementation
CDISC guidance for ADaM:
General Considerations V1.0
Subject-Level Analysis (ADSL dataset, draft)
Other guidance (draft)
ADaM guidelines are not as restrictive as those for SDTM!
10/20/2005 18
ADaM implementation at Accovion
• Add variables, e.g change from baseline
• Add ADSL information → supports “One PROC away”
• Derive variables only once
Use Draft SDTM/ADaM as input
Create ADSL dataset
Create other ADaM datasets
Define standards across projects
Few study/project specific changes
10/20/2005 19
Tools for SDTM and ADaM Generation
SAS program generator
Standardized data mapping using SAS
• Dataset mapping table (one per study)
• Variable mapping tables (one per domain)
• Macro creates programs and datasets
At Accovion:
10/20/2005 20
Dataset mapping table: Example study 1234
Studyid Domain Description Structure Sort order1234 ADSL Subject Level Key Information Special Purpose USUBJID1234 ADAE Adverse Events Events USUBJID, AESEQ1234 ADEF Efficacy Findings USUBJID, EFTESTCD, EFANLTMN
ADaM domain metadata information
Studyid Domain Variable Label Type Length Format SAS library
Dataset name
Variable name
Task Transform (SAS code)
Com-ments
Exec. order
1234 ADSL SEX Sex Char 1 SDTM DM SEX no change1234 ADSL TRTAN Actual Treatment
Group NumberNum SDTM EX EXTRT macro %trta(indata,
outdata)… 2
ADaM variable metadata information Source information Mapping information
Variable mapping tables: Example domain ADSL
Tables to feed program generator
10/20/2005 21
Automation of Documentation
For submission to the FDA:
Documentation at Accovion:
Datasets as xpt files
Documentation for SDTM and ADaM in define.xmlformat
Standard tool
Usage of mapping tables
Additional information, e.g. role
10/20/2005 22
Output 1: Domain level metadata, example study 1234
Dataset Description Structure Purpose Keys Location DocumentationADSL Subject Level
Key Information1 record per subject Analysis USUBJID ../1234/dds/
adsl.xptSAP and/or adsl.sas
ADAE Adverse Events 1 record per subject per event Analysis USUBJID, AESEQ ../1234/dds/adae.xpt
SAP and/or adae.sas
ADEF Efficacy 1 record per subject per parameter per analysis time window
Analysis USUBJID, EFTESTCD, EFANLTMN
../1234/dds/adef.xpt
SAP and/or adef.sas
ADaM Datasets for study 1234
Variable Label Type Controlled Terms or Format
Origin Role Comment
SEX Sex Char DM.SEX Result qualifier No changeTRTAN Actual Treatment
Group NumberNum 0 = ‘Placebo’
1 = 'Drug A'2 = 'Drug B'3 = 'Drugs A+B'
Derived from EX.EXTRT
Selection, result qualifier
For one subject either one or two treatments in EXTRT are possible
Variable Metadata for Dataset ADSL – 1 record per subject
Output 2: Variable level metadata, example dataset ADSL
Output of Accovion’s standard tool
10/20/2005 23
Integrating different data structures (I)
Deviating data structures:
• Legacy studies
• Sponsor’s structures
Accovion full service:
• Standardized “Pre-SDTM”
• SDTM
• ADaM
A migration to SDTM and ADaM may be necessary!
10/20/2005 24
Integrating different data structures (II)
Study specific source data:
• CDMS structure
• SDTM
• Analysis datasets
Migration to fixed target data:
• Meta-analysis: ADaM only
• Data for submissions: SDTM + ADaM
10/20/2005 25
Issues for consideration
Programming effort
Validation effort
Reusability of programs/potential of standardization
Risk of errors
Risk of inconsistencies
10/20/2005 26
Usage of standard mapping tool for data integration!
CDMS
SDTM
ADS
SDTM
ADaM
1. Direct Migration
CDMS
2. Migration to “Pre-SDTM”
Pre-SDTM(= SDTM pure+SDTM comfort)
Data Integration Tool
Accovion standard mapping
Accovion’s proposals
Data Integration Tool
10/20/2005 27
Analysis output
• Review of SAP and TLG shells
• “Analysis-ready” datasets
Programming based on analysis datasets (ADaM)
ADaM standards support usage of standard analysis tools
At Accovion a set of 20 standard tables was developed
Consider analysis output for ADaM set-up:
10/20/2005 28
Example “Change from Baseline in Primary Efficacy Variable – ITT population”:
Use efficacy dataset ADEF with selection criteria “WHERE (EFTESTCD=‘PEF’AND ITT=‘Y’);” T a b l e 1 . 1 C h a n g e f r o m B a s e l i n e i n P r i m a r y E f f i c a c y V a r i a b l e – I T T p o p u l a t i o n
N o t e : 1 p - v a l u e s f r o m p a i r w i s e c o m p a r i s o n A B v e r s u s A r e s p . B f r o m t - t e s t ; I T T = i n t e n t i o n - t o - t r e a t ; t r e a t m e n t g r o u p s a s t r e a t e d .
A n a l y s i s S t a t i s t i c R e s u l t C h a n g e f r o m B a s e l i n eW e e k _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ A B A B A B A B ´ ( N = x x ) ( N = x x ) ( N = x x ) ( N = x x ) ( N = x x ) ( N = x x ) _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ B a s e l i n e N w i t h d a t a x x x x x x x x x N m i s s i n g x x x x x x x x x M e a n x x x . x x x x . x x x x . x S D x x x . x x x x x . x x x x x . x x M i n x x x x x x x x x M e d i a n x x x . x x x x . x x x x . x M a x x x x x x x x x x 2 N w i t h d a t a x x x x x x x x x x x x x x x x x x N m i s s i n g x x x x x x x x x x x x x x x x x x M e a n x x x . x x x x . x x x x . x x x x . x x x x . x x x x . x S D x x x . x x x x x . x x x x x . x x x x x . x x x x x . x x x x x . x x M i n x x x x x x x x x x x x x x x x x x M e d i a n x x x . x x x x . x x x x . x x x x . x x x x . x x x x . x M a x x x x x x x x x x x x x x x x x x x p - v a l u e 1 x . x x x x x . x x x x 9 5 % C I [ x x . x ; x x . x ] [ x x . x ; x x . x ] e t c .
EFANLTMN
TRTAN
EFSTRESN
EFSTRESN=EFBLRESN
EFCHGBL
Relationship ADaM ⇔ analysis output
10/20/2005 29
FDA: Recommendation for CDISC analysis metadata structure not yet available
Analysis Name Description Reason Dataset DocumentationTable 1.1: ef00001t.lst Change from Baseline in Primary
Efficacy Variable – ITT populationPrespecified in Protocol
../1234/dds/adef.xpt SAP, Section X.Y and/or ef00001t.sas
Table 1.2: ef00002t.lst Proportion of Responders at Endpoint – Per-Protocol population
Prespecified in Protocol
../1234/dds/adef.xpt SAP, Section X.Y and/or ef00002t.sas
Analysis-Level Metadata for study 1234
Documentation of analysis
Follow ADaM General Considerations:
10/20/2005 30
Summary
CDISC impacts the clinical development process
CDISC standardization:• Efficiency ↑
• Risk of errors ↓
Accovion’s modular approach:• Benefit from standardization
Study specific needs
Changes in SDTM and ADaM models
Customer specific needs
• Flexibility:
10/20/2005 31
Thank you!
Thank you for your attention!
Thanks to our colleagues for all the efforts todevelop this strategy!
Edelbert Arnold and Ulrike PlankAccovion GmbH, Germany