A SYSTEMATIC APPROACH FOR SYNTHETIC REPLICATION OF CLINICAL TRIAL COHORTS USING RETROSPECTIVE REAL-WORLD AND CLINICAL TRIAL DATA
Author(s)
Galaznik A1, Berger M2, Lempernesse B3, Ransom J3, Shilnikova A4
1SHYFT Analytics, Belmont, MA, USA, 2Self Employed, New York, NY, USA, 3SHYFT Analytics, Waltham, MA, USA, 4Medidata Solutions, Boston, MA, USA
Presentation Documents
| OBJECTIVES: Replication of clinical trials through retrospective data has potential applications ranging from in silico modeling and synthetic control arm creation to extrapolation of clinical trial findings to real world practice. We outline here a systematic approach leveraging common data standards for data pooling and study replication. METHODS: 1. Clinical Trial Cohort Creation. Starting with the design of the target trial to be replicated, data from similar trials can be gathered and pooled leveraging the CDISC SDTM common data standard.1 Protocol inclusion/exclusion criteria are applied. 2. Cohort Adjustment. Sources of bias and imbalance are assessed, both according to established Cochrane criteria for clinical data pooing.2 Adjustments can include: matching, stratification, weighting, and/or multivariate regression modeling approaches.3 (Table 1: Example – Synthetic Control in AML) 3. Real World Cohort Creation. Adjusted trial cohort is converted from CDISC to a common data model for real world data comparison, such as the Observational Medicine Outcomes Partnership (OMOP) format.5 (Table 2: Example - CDISC SDTM to OMOP Conversion in Alzheimer’s Disease) 4. Cohort Adjustment. Assessments of bias and imbalance between the clinical trial and real-world cohort(s) can be conducted according to standard observational research practice.3,7 Common adjustments include high-dimensional propensity score matching or weighting.8 5. Cohort Replication. Benchmarking analysis can be conducted across both the clinical trial and real-world cohort(s). Common data modeling ensures that the same analysis can be executed in different cohorts, with consistent outcomes definitions. CONCLUSIONS: Common data standards can facilitate consistent, rapid and transparent replication of study cohorts across both clinical trial and real -world data sources. This has tremendous application for developing synthetic trial cohorts, in applications ranging from historical controls for pre-approval single arm trials to post-approval assessments of patient sub-populations. |
Conference/Value in Health Info
2019-05, ISPOR 2019, New Orleans, LA, USA
Value in Health, Volume 22, Issue S1 (2019 May)
Code
PMU8
Topic
Real World Data & Information Systems
Topic Subcategory
Reproducibility & Replicability
Disease
Neurological Disorders, Oncology