CDISC Standards Problems and Solutions: Some Examples Paul Terrill and Sarah Brittain
2 Aim To discuss some problems met when creating and processing datasets that follow SDTM (and ADaM) standards.
3 Introduction MDSL International Specialist CRO Supporting Pharma/Biotech companies with limited in-house statistical and data management experience
4 Introduction Overall Problem Clients have limited knowledge about CDISC Not involved from beginning Consequences Difficult to retrospectively follow CDISC Solutions Take on decisions for clients Increased consultancy work Be involved from beginning!
5 Protocol/CRF Design not CDISC Tables and listings to follow protocol/crf SDTM datasets required Problems: 1. CRF not CDASH 2. Trial Dataset Creation
Problem 1: CRF not CDASH 6 Original data (CRF) not collected according to CDASH / controlled terminology Solution Map CRF data to SDTM controlled terminology Back code in ADaM for tables and listings
Problem 1: CRF not CDASH Example 1: Study Termination 7 Study completed according to protocol? If no, indicate one primary reason SDTM (DS domain) - Lack of efficacy - Adverse event - Subject Request - Protocol Deviation - Lost to Follow-up - Death - Other LACK OF EFFICACY ADVERSE EVENT WITHDRAWAL BY SUBJECT PROTOCOL VIOLATION LOST TO FOLLOW-UP DEATH OTHER ADaM (ADDS) LACK OF EFFICACY ADVERSE EVENT SUBJECT REQUEST PROTOCOL DEVIATION LOST TO FOLLOW-UP DEATH OTHER
Problem 1: CRF not CDASH Example 2: Adverse Events 8 Action taken (tick all that apply): - None (1) - Study Drug Discontinued (2) - Study Drug Stop and Restart (3) - Treatment (4) SDTM (1) SUPPAE: QLABEL= Action Taken None QVAL= NONE ADaM (ADAE.AEACT) NONE (2) AE.AEACN= DRUG WITHDRAWN STUDY DRUG DISCONTINUED (3) AE.AEACN= DRUG INTERRUPTED STUDY DRUG STOP AND RESTART (Not 2 or 3) AE.AEACN= DOSE NOT CHANGED (4) AE.AECONTRT= Y TREATMENT
9 Problem 2: Trial Datasets Trial datasets not thought about up front Solutions: Retrospectively produce datasets (time consuming and tricky) Recommend create trial design datasets whilst developing protocols for future projects
Problem 2: Trial Datasets 10 Example 1: Trial Arms Study with open-label period followed by doubleblind period (two treatments) followed by an optional extension study
Problem 2: Trial Datasets 11 Example 1: Trial Arms Create 4 trial arms, although there are only 2 blinded treatments: 1. Open-label Treatment 1 2. Open-label Treatment 1 Extension study 3. Open-label Treatment 2 4. Open-label Treatment 2 Extension study
Problem 2: Trial Datasets 12 ARM CD Example 1: Trial Arms ARM TAET ORD ETCD ELEMENT TABRANCH EPOCH AB TRT1 1 SCREEN SCREENING SCREENING AB TRT1 2 OLTRT OPEN LABEL TRT RANDOMISATION TO TRT1 OPEN-LABEL AB TRT1 3 TRT1 TRT1 DOUBLE-BLIND AB TRT1 4 PT POST-TREATMENT NOT ENTER EXTENSION POST-TREATMENT AB TRT1 5 FU FOLLOW UP FOLLOW UP ABX TRT1-EXT 1 SCREEN SCREENING SCREENING ABX TRT1-EXT 2 OLTRT OPEN LABEL TRT RANDOMISATION TO TRT1 OPEN-LABEL ABX TRT1-EXT 3 TRT1 TRT1 DOUBLE-BLIND ABX TRT1-EXT 4 PT POST-TREATMENT ENTER EXTENSION POST-TREATMENT ABX TRT1-EXT 5 FUX FOLLOW UP EXTENSION AC TRT2 Etc Etc Etc Etc Etc FOLLOW UP
13 Problem 2: Trial Datasets Example 2: Protocol with several Parts Study split into different consecutive parts Part 1: Three period crossover Parts 2 and 3: Placebo controlled, single dose based on dose selection from Part 1 Part 4: Three period crossover
14 Problem 2: Trial Datasets Example 2: Protocol with several Parts Resulting trial design datasets large Use of ARMCD and EPOCH to help distinguish between parts
Problem 2: Trial Datasets Example 2: Protocol with several Parts 15 ARMCD ARM TAET ORD ETCD ELEMENT TABRANCH EPOCH ABC ABC 1 SCREEN SCREENING RANDOMISATION TO TRT ABC SCREENING ABC ABC 2 A TRT A PART 1 FIRST TREATMENT EPOCH ABC ABC 3 B TRT B PART 1 SECOND TREATMENT EPOCH ABC ABC 4 C TRT C PART 1 THIRD TREATMENT EPOCH ABC ABC 5 FU FOLLOW UP FOLLOW UP BAC BAC 1 Etc Etc Etc Etc B B 1 SCREEN SCREENING RANDOMISATION TO TRT B SCREENING B B 2 B TRT B PART 2 AND 3 TREATMENT EPOCH B B 3 FU FOLLOW UP FOLLOW UP A A 1 Etc Etc Etc Etc
16 Processing of Data SDTM datasets repeatedly processed Creation of ADaM Some Listings Efficient methods required Problem: 3. SUPPxx Domains
17 Problem 3: SUPPxx Domains Processing SUPPxx domains Solutions: Use macros that combine SUPP datasets to original domain Create additional database where SUPP dataset variables are included in parent domain (QNAM)
18 Problem 3: SUPPxx Domains Example: Reason for unscheduled visit SUPPSV: RDOMAIN USUBJID IDVAR IDVARVAL QNAM QLABEL QVAL SV 001 VISITNUM 3.1 SVUPREAS Primary Reason for Visit REPEAT LAB TESTS ADSV: Unique Subject Identifier Visit Number Primary Reason for Visit USUBJID VISITNUM SVUPREAS 001 3.1 REPEAT LAB TESTS
19 Non-Standard Data Data collected / CRF does not fit into standard domains SDTM still required Create custom domain or... Put into QS (Questionnaire) type domain
20 Problem 4: Non-Standard Data Example: Subject Diary Drug Accountability Primary source for drug accountability should be CRF but daily Dose taken? Yes/No also collected on a subject diary Solution: Put diary data into QS type domain Use this to derive compliance if necessary
21 Problem 4: Non-Standard Data Example: Subject Diary Drug Accountability QSSEQ QSTESTCD QSTEST QSCAT QSORRES VISITNUM QSDTC QSTPT 1 TRT TRT TAKEN? DRUG YES 2 2010-08-01 DAY 1 2 TRT TRT TAKEN? DRUG YES 2 2010-08-02 DAY 2 3 TRT TRT TAKEN? DRUG NO 2 2010-08-03 DAY 3 4 TRT TRT TAKEN? DRUG YES 2 2010-08-04 DAY 4 5 TRT TRT TAKEN? DRUG YES 2 2010-08-05 DAY 5 Etc Etc Etc Etc Etc Etc Etc Etc
Development Program 22 Consistency Many studies form part of development program Consistency between studies required Problems: 5. TESTCD/PARAMCD 6. Changing Standards
23 Problem 5: TESTCD/PARAMCD Consistency of endpoints and associated xxtestcd / PARAMCD across studies Solutions: Create ongoing master test code list for each program. Use csv format so it can be easily read in to create a format. Try and use the same team within indications / development programs
24 Problem 5: TESTCD/PARAMCD Example: Test codes in a uterine myoma study TESTCD M1VOL M1LOC M1TYP Etc TMVOL ULEN UHGT UDEP UVOL TEST MYOMA 1 VOLUME MYOMA 1 LOCATION MYOMA 1 TYPE Etc TOTAL MYOMA VOLUME UTERUS VOLUME UTERUS HEIGHT UTERUS DEPTH UTERUS VOLUME
25 Problem 6: Changing Standards Changing standards over long-running development programs Solutions: Generally try to use most up to date standard, but... Continually assess backwards compatibility Try to keep the same team working on development programs
Questions 26