PND90 REAL WORLD REPLICATION OF CLINICAL TRIAL FINDINGS IN ALZHEIMER'S DISEASE

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. In this study we replicate the study cohort from an Alzheimer’s trial in a real-world data source, leveraging common data modeling and vocabularies. This study was conducted using a publicly available Alzheimer’s clinical trial dataset from CDISC, and a nationally representative commercial insurance claims dataset in OMOP CDM v5. The clinical trial dataset was converted from the CDISC SDTM into the OMOP Common Data Model, v5, and standard vocabularies, with conversion completeness >99%.1 Clinical converted data was loaded into the SHYFT Quantum V6.7.0 solution, and inclusion/exclusion criteria for the trial were applied to the claims data to create a real world benchmark cohort for comparison. Common OMOP vocabularies were used to derive equivalent clinical (concomitant medications, comorbidities), demographic (age, gender), and outcomes variables across both cohorts, and to replicate generation of descriptive statistics and Kaplan-Meier time-to-event analyses for key outcomes measurable in both datasets. Due to limited clinical outcomes in the insurance claims dataset, safety events were assessed (application site disorder, erythema, rash, site irritation, application site pain, edema). Demographic and clinical characteristics of both cohorts were statistically similar between clinical trial and real-world cohorts (Figure 1). Event rates and median time-to-event for safety outcomes were either statistically or directionally consistent (Figure 2) By leveraging standard OMOP vocabularies and common data modeling, cohort replication and analysis can be executed rapidly and consistently. Next steps include application of propensity score matching or weighting to assess impact on comparative outcomes event-rate and time-to-event measures between cohorts.