cdisc adam 2.1 implementation: a challenging next step …€¦ · 1 cdisc adam 2.1 implementation:...
TRANSCRIPT
1
CDISC ADaM 2.1 Implementation:A Challenging Next Step in the Process
Presented by Tineke Callant
2014-03-14
2
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
2
3
Clinical Data Interchange Standards Consortium - Introduction
n 1997 - Inception
n 2000 - 32 global companies
CDISC is a global, open, multidisciplinary, non-profit organization that has established standards to support the acquisition, exchange, submission and archive of clinical research data and metadata.
n 2014 - ± 200 organizations• biotechnology and pharmaceutical development companies• device and diagnostic companies• CROs and technology providers• government institutions, academic research centers and other non-profit
organizations
4
Clinical Data Interchange Standards Consortium - Introduction
3
5
Clinical Data Interchange Standards Consortium - Introduction
n Mission statement
The CDISC mission is to develop and support global, platform-independent data standards that enable information system interoperability to improve medical research and related areas of healthcare.
Data standards to improve clinical research
6
Clinical Data Interchange Standards Consortium - Introduction
- 2001: Biomedical Research Integrated Domain Group (BRIDG) Model
4
7
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
8
CDISC - Foundational standards
6
11
CDISC - Foundational standards
n Study Data Tabulation Model (SDTM)
The content standard for regulatory submission of case report form data tabulations from clinical research studies.
Datasets containing data collected during the study and organized by clinical domain.
n Analysis Data Model (ADaM)
The content standard for regulatory submission of analysis datasets and associated files.
Datasets used for statistical analysis and reporting by the sponsor, submitted in addition to the SDTM domains.
12
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
7
13
CDISC ADaM V2.1 - Analysis data flow
ADaM
14
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
8
15
CDISC ADaM V2.1 - ADaM data structures
n The Subject-Level Analysis Dataset (ADSL) structure
n The Basic Data Structure (BDS)
n Other
16
CDISC ADaM V2.1 - ADaM data structuresThe Subject-Level Analysis Dataset (ADSL) structure
n One record per subject
n Variables (required + other)• Study identifiers (e.g. DM.STUDYID)• Subject demographics (e.g. DM.AGE)• Population indicator(s) (e.g. RANDFL)• Treatment variables (e.g. DM.ARM)• Trial dates (e.g. RANDDT)
n Required in a CDISC-based submission
9
17
CDISC ADaM V2.1 - ADaM data structures
n The Subject-Level Analysis Dataset (ADSL) structure
n The Basic Data Structure (BDS)
n Other
18
CDISC ADaM V2.1 - ADaM data structuresThe Basic Data Structure (BDS)
n One or more records per subject, per analysis parameter, per analysis time point (conditionally required)
n Variables• e.g. PARAM and related variables• e.g. AVAL and AVALC and related variables• e.g. the subject identification• e.g. DTYPE• e.g. treatment variables, covariates
n Supports the majority of statistical analyses
10
19
CDISC ADaM V2.1 - ADaM data structures
n The Subject-Level Analysis Dataset (ADSL) structure
n The Basic Data Structure (BDS)
n Other
20
CDISC ADaM V2.1 - ADaM data structuresOther
n CDISC ADaM Basic Data Structure for Time-to-Event Analysis Version 1.0 - May 8, 2012
n CDISC ADaM Data Structure for Adverse Event Analysis Version 1.0 - May 10, 2012
11
21
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
22
CDISC ADaM V2.1 - Analysis data flow
ADaM
12
23
n Understanding the relationship of element vs. predecessor
n Enabling transparancy
n Analysis results → Analysis datasets → SDTM
CDISC ADaM V2.1 - Traceability
24
CDISC ADaM V2.1 - TraceabilityStrategies for implementing SDTM and ADaM standardsSusan Kenny – Michael Litzsinger
n Parallel method
SDTM DomainsDBMS Extract
Analysis Datasets
n Retrospective method
DBMS Extract → Analysis Datasets → SDTM Domains
n Linear method
DBMS Extract → SDTM Domains → Analysis Datasets
n Hybrid method
DBMS Extract → SDTM Draft Domains → Analysis Datasets → SDTM Final Domains
13
25
CDISC ADaM V2.1 - Traceability
n Traceability
26
CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 • Fundamental principles
– Provide traceability between the analysis data and its source data
• Practical considerations– Maintain the values and attributes of SDTM variables
n CDISC ADaM implementation guide (IG) V1.0• General variable naming conventions
14
27
CDISC ADaM V2.1 - TraceabilityGeneral variable naming conventions
Any ADaM variable whose name is the
same as an SDTM variable must be a
copy of the SDTM variable, and its label,
meaning, and values must not be
modified
28
n Parallel method
SDTM DomainsDBMS Extract
Analysis Datasets
n Retrospective method
DBMS Extract → Analysis Datasets → SDTM Domains
n Linear method
DBMS Extract → SDTM Domains → Analysis Datasets
n Hybrid method
DBMS Extract → SDTM Draft Domains → Analysis Datasets → SDTM Final Domains
CDISC ADaM V2.1 - TraceabilityStrategies for implementing SDTM and ADaM standardsSusan Kenny – Michael Litzsinger
15
29
n Linear method
DBMS Extract → SDTM Domains → Analysis Datasets
• Traceability• CDISC SDTM/ADaM Pilot Project• Recommended
CDISC ADaM V2.1 - TraceabilityStrategies for implementing SDTM and ADaM standardsSusan Kenny – Michael Litzsinger
30
n Hybrid method
DBMS Extract → SDTM Draft Domains → Analysis Datasets → SDTM Final Domains
• Traceability• Amendment 1 SDTM V1.2 and SDTM IG V3.1.2• Future?!?
CDISC ADaM V2.1 - TraceabilityStrategies for implementing SDTM and ADaM standardsSusan Kenny – Michael Litzsinger
16
31
n Traceability → Recommended: Linear method
n Flexible
n Delivery of consistent analysis datasets
n Easy to use (Excel file)
n Easy to maintain (Excel file)
CDISC ADaM V2.1 - Traceability
32
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
17
33
CDISC ADaM V2.1 - ADaM metadata
n Microsoft Office Excel spreadsheet as framework
n Metadata
34
CDISC ADaM V2.1 - ADaM metadata
n Microsoft Office Excel spreadsheet as framework
n analysis dataset
n %CHKSTRUCT(ds_ = )• Automatization • Compliance
n define.xml
18
35
CDISC ADaM V2.1 - ADaM metadata
n Analysis dataset metadata
n Analysis variable metadata
n Analysis parameter value-level metadata
n Analysis results metadata
36
CDISC ADaM V2.1 - ADaM metadataAnalysis dataset metadata
n Illustration from CDISC ADaM V2.1
n Practical consideration: ADxxxxxx
! ≠ SDTM !The key variables should define uniqueness
19
37
Analysis dataset naming convention
n ADxxxxxx
n The subject-level analysis dataset is named ADSL
n max. 8 characters
CDISC ADaM V2.1 - ADaM metadataAnalysis dataset metadata
38
CDISC ADaM V2.1 - ADaM metadata
n Analysis dataset metadata
n Analysis variable metadata
n Analysis parameter value-level metadata
n Analysis results metadata
20
39
n Illustration from CDISC ADaM V2.1
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata
40
CDISC ADaM V2.1 - ADaM metadata
n Analysis dataset metadata
n Analysis variable metadata
n Analysis parameter value-level metadata
n Analysis results metadata
21
41
n Illustration from CDISC ADaM V2.1
CDISC ADaM V2.1 - ADaM metadataAnalysis parameter value-level metadata
42
CDISC ADaM V2.1 - ADaM metadata
n Analysis dataset metadata
n Analysis variable metadata
n Analysis parameter value-level metadata
n Analysis results metadata (not required)
22
43
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice
n Analysis dataset metadatan Analysis variable metadata
Dataset name Display formatVariable name Codelist / Controlled termsVariable label Source / DerivationVariable type
Parameter identifier (Basic Data Structure (BDS))
n Analysis results metadata (not required)
44
CDISC ADaM V2.1 - ADaM metadata
n Microsoft Office Excel spreadsheet as framework
n Metadata
23
45
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice
n SAS variable attributes
n To work in a SAS environment– NAME– TYPE– LENGTH– FORMAT– INFORMAT– LABEL– POSITION IN
OBSERVATION– INDEX TYPE
n Analysis variable metadata fields
– DATASET NAME– VARIABLE NAME– VARIABLE LABEL– VARIABLE TYPE– DISPLAY FORMAT– CODELIST /
CONTROLLED TERMS– SOURCE / DERIVATION– BASIC DATA STRUCTURE:
PARAMETER IDENTIFIER
46
n Example
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice
...
24
47
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice -Subposition in observation
n Example• ADSL – SITEGR* (Char) and SITEGR*N (Num)
* = a single digit [1-9]
• SITEID
• SITEID grouped together by city in the variable SITEGR1 (SITEGR1N)
• SITEID grouped together by province in the variable SITEGR2 (SITEGR2N)
48
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice -Subposition in observation
%CHKSTRUCT(ds_ = ADSL)
1 21 2ORDER
25
49
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice -Subposition in observation
ORDER 1 2 1 2
50
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice -Subposition in observation
n Example• ADSL – SITEGR* (Char) and SITEGR*N (Num)
* = a single digit [1-9]
POSITION IN OBSERVATION
SUBPOSITION IN OBSERVATION
VARIABLE NAME
1 STUDYID
2 USUBJID
3 SITEID
4 1 SITEGR*
4 2 SITEGR*N
26
51
n Example
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice
n Example
...
52
CDISC SDTM CDISC ADaMReq - Required
The variable must be included in the dataset and cannot be null for any record.
Req - Required
The variable must be included in the dataset.
Exp - Expected
... and may contain some null values.
Cond - Conditionally required
... in certain circumstances.Perm - Permissible
The variable should be used in a domainas appropriate when collected or derived.
Perm - Permissible
The variable may be included in the dataset, but is not required.
CDISC ADaM V2.1 - ADaM metadataAnalysis variable metadata in practice - Core
n Nulls are allowed
27
53
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
54
CHKSTRUCT macro
n Microsoft Office Excel spreadsheet as framework
n analysis dataset
n %CHKSTRUCT(ds_ = )• Automatization • Compliance
n define.xml
28
55
CHKSTRUCT macro - Automatization
%CHKSTRUCT(ds_ = ADSL)
Before
After
4 6 5 7 1 2 3
1 2 3 4 5 6 7
ORDER THE ANALYSIS VARIABLES
56
CHKSTRUCT macro - Automatization
%CHKSTRUCT(ds_ = ADSL)
Before
After
LABEL THE ANALYSIS VARIABLES
29
57
CHKSTRUCT macro - Automatization
%CHKSTRUCT(ds_ = ADSL)
Key variables
7
2134
5698
10
5
1234
6789
10
Key variables
Before
After
SORT THE ANALYSIS DATASET
58
CHKSTRUCT macro – Compliance
Analysis dataset Analysis variable metadata
30
59
CHKSTRUCT macro – Compliance
Analysis dataset Analysis variable metadata
60
CHKSTRUCT macro – Compliance
Analysis dataset Analysis variable metadata
31
61
CHKSTRUCT macro
n Excel spreadsheet as framework
n Purpose• Reference • Automatization• Compliance
62
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
32
63
Linear method - Challenges and solutions
Step 1
64
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation Guide
...
...
33
65
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation Guide
Any ADaM variable whose name is the
same as an SDTM variable must be a
copy of the SDTM variable, and its label,
meaning, and values must not be
modified
66
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation GuideChallenge: Flexible variable length
...
...
...
34
67
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation GuideChallenge: Flexible variable length
n CDISC SDTM IG• Variables of the same name in split datasets should have the same
SAS Length attribute• Version 5 SAS transport file format: max. 200 characters• -- TESTCD and QNAM: max. 8 characters• -- TEST and QLABEL: max. 40 characters
n Example: DM.RACE: $41, $50, and $200
n Amendment 1 to SDTM V1.2 and SDTM IG V3.1.2• Version 5 SAS transport file format: max. 200 characters
! only if necessary !
68
n Traceability
n Flexible
n Delivery of consistent analysis datasets
n Easy to use
n Easy to maintain
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation GuideChallenge: Flexible variable length
35
69
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation GuideSolution: [sdtm] ↔ %CHKSTRUCT(ds_ = )
70
Example: LB.LBSCAT
Linear method - Challenges and solutionsStep 1 - CDISC SDTM Implementation GuideChallenge: Permissible variables
Solution: [sdtm] ↔ %CHKSTRUCT(ds_ = )
36
71
Linear method - Challenges and solutions
Step 2
72
Linear method - Challenges and solutionsStep 2 - SUPP--
n QNAM → variable name
n QLABEL → variable label
n QVAL → variable type
→ variable length
e.g. SUPPDM SDTM dataset e.g. ADSL ADaM dataset
37
73
Linear method - Challenges and solutionsStep 2 - SUPP--Challenge: Flexible code list
n QLABEL is different for the same QNAM– Example
ELIGCONF Subject Still EligibleELIGCONF Still Fulfill Eligibility Criteria
n QLABEL format– Example
RANDNO RANDOMIZATION NUMBERRANDNO Randomization Number
n QLABEL changes during the course of a study– Example
ELIGIBLE Suject Eligible For Dosing ELIGIBLE Subject Eligible For Dosing
74
Linear method - Challenges and solutionsStep 2 - SUPP--Solution: [supp] ↔ %CHKSTRUCT(ds_ = )
38
75
Linear method - Challenges and solutions
Step 3
76
Linear method - Challenges and solutions - Step 3
ADaM
39
77
Linear method - Challenges and solutions - Step 3Challenge: 12 SDTM → 12 ADaM?!?
1
3
2
4
5
6
8
7
910
SDTM
12
11
ADaM
?
?
??
??
??
??
??
78
Linear method - Challenges and solutions - Step 3Solution: 1 central model + sponsor specific add-ons
sponsorspecificadd-on
centralADaMmodel
domlist.sas7bdat
varlist.sas7bdat
codelist.sas7bdat
domlist.sas7bdat
varlist.sas7bdat
codelist.sas7bdat
domlist.sas7bdat
varlist.sas7bdat
codelist.sas7bdat
1
1 Convert Excel file to SAS datasets (by ADaM administrator)
2
2 Combine central model and sponsor specific add-on (by study programmer)
1
40
79
n Traceability
n Flexible
n Delivery of consistent analysis datasets
n Easy to use
n Easy to maintain
Linear method - Challenges and solutions - Step 3Solution: 1 central model + sponsor specific add-ons
80
Linear method - Challenges and solutions
Step 4
41
81
Linear method - Challenges and solutions - Step 4Challenge: SDTM model no. 1, 2, 3 ... ?
1
3
2
4
5
6
8
7
910
SDTM
12
11
ADaM
?
?
??
??
??
??
??
82
Linear method - Challenges and solutions - Step 4 Solution: Central metadata repository
n CDISC metadata• SDTM version• SDTM metadata• ...
n Study characteristics • Therapeutic area• Clinical phase • Trial design characteristics• ...
n Project metadata• Study timelines• Key Performance Indicators• ...
42
83
Linear method - Challenges and solutions
Step 5
84
Linear method - Challenges and solutions – Step 5Challenge: Future
43
85
Linear method - Challenges and solutions – Step 5Challenge: Future
86
Agenda
n CDISC - Introduction
n CDISC - Foundational standards
n CDISC ADaM V2.1 - Analysis data flow
n CDISC ADaM V2.1 - ADaM data structures
n CDISC ADaM V2.1 - Traceability
n CDISC ADaM V2.1 - ADaM metadata
n CHKSTRUCT macro
n Linear method - Challenges and solutions
n Take home messages
44
87
Take home messagesMessage no. 1
ADaM SDTM
n SDTM and ADaM go hand in hand
n Thus, without a CDISC compliant SDTM database to start from, ADaM cannot exist
n But do realize a strong analysis data model needs more than a CDISC compliant SDTM database alone
88
n Linear method:• Recommended• Challenging
n Solution:• SDTM: Central metadata repository• ADaM: Automatization, e.g. [sdtm], [supp] …
Study medata differences are handled efficiently
Take home messagesMessage no. 2