PCN12 USING CLINICAL DATA STANDARDS TO FACILITATE COMPARISONS OF ACUTE MYELOID LEUKEMIA TREATMENT PATTERNS AND OUTCOMES IN REAL WORLD CLINICAL CARE AND POOLED CLINICAL TRIAL POPULATIONS

Abstract

While randomized controlled clinical trials are the gold standard for demonstrating efficacy, there is a need to facilitate comparison of trial findings with real world populations. We leverage common data modeling and vocabularies to compare Acute Myeloid Leukemia in real-world clinical practice to a pooled synthetic cohort of clinical trial subjects. Clinical trial data was derived from a pooled dataset of 7 clinical trials (n=719) for relapsed/refractory AML from 2012-2017Medidata archive of >3000 trials, created using CDISC SDTM. Real-world data was obtained from a US-based geographically representative oncology-focused electronic medical record (EMR) and a US medical insurance claims dataset. All data sets were converted into the OMOP Common Data Model, v5, and standard vocabularies, with conversion completeness >99%. Analyses were conducted in SHYFT Quantum v6.7.0. Descriptive comparisons were conducted for baseline patient demographics, comorbidities, treatment patterns and outcomes. Systematic comparison of population heterogeneity across clinical trial, EMR and claims data was also conducted, with respect to patient, treatment, and outcome variation. The approach employed patient clustering, treatment propensity score distribution, and Extreme Quartile Risk Ratio (EQRR) and Median-to-Mean Risk Ratio (MMSR) calculations. Real world populations showed similar age and gender distributions, with lower use of cytotoxic agents than clinical trials. Heterogeneity measures were directionally greater in real world data but remarkably consistent with trial findings. Use of common data standards can enable systematic and consistent comparisons, and drive potential applications ranging from in silico modeling and synthetic control arm creation to extrapolation of clinical trial findings to real world practice.