(A-202) Active Safety Surveillance Using Real-world Evidence (ASSURE) Implementation: Transparent and Reproducible Real-World Evidence (RWE) Standardized Framework to Support Safety Signal Evaluation

Background: Real-world data (RWD) are increasingly utilized in safety signal evaluations, which are performed rapidly to meet tight regulatory timelines.

Objectives: To summarize experience with Johnson & Johnson’s ASSURE program for producing efficient, transparent, reproducible, and impactful RWE to support routine pharmacovigilance processes.

Methods: ASSURE follows a stepwise framework that encompasses database diagnostics to select fit-for-purpose real-world data sources, phenotype development, query specification and analytic implementation, objective study validity diagnostics, and reporting. Licensed RWD sources are transformed into the Observational Medical Outcomes Partnership Common Data Model, from which evidence is generated for specific questions using a suite of open-source large-scale analytic tools. For each signal request, we formed drug exposure outcome evaluations (some requests had multiple outcomes evaluated) and generated population characterization, population-level causal effect estimation, and patient-level prediction evidence. We provide an evaluation of the implementation of this framework, including results from database diagnostics, phenotype development, and objective study validity diagnostics prior to unblinding evidence. We defined efficiency as the time from question to results and integration of results into safety decisions.

Results: The initial 19 months of the ASSURE program supported 110 safety signal evaluations across 24 unique products that arose from 47 requests across multiple therapeutic areas. Many analyses failed at least some diagnostics: database (7 evaluations from 7 requests), phenotype specification (3 evaluations from 3 requests), or the requisite objective study validity diagnostics (26 evaluations). These 26 evaluations (24%) did provide population characterization of outcome incident rates, but no causal effect estimates. There were 74 evaluations (67%) that yielded effect estimation results with 62 (54%) providing propensity-score adjusted cohort method results and 40 (36%) providing self-controlled case series results. Overall, the median time from signal identification to RWE results was 34 (IQR 29-35) days, which afforded time to integrate RWE into regulatory safety decisions.

Conclusions: The ASSURE framework and implementation process collaborating with safety management teams enables rapid response to medical product safety signals under evaluation and produces actionable RWE that transparently supports pharmacovigilance decision making within regulatory timeframes.