A Bayesian Framework for Identifying Safety-Signal from Multiple Studies

• Activity Number: 518 - Estimand, Causal Inference, and Other Statistical Considerations in Clinical Trials
• Type: Contributed
• Date/Time: Thursday, August 11, 2022 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Adrijo Chakraborty* and Ram Tiwari
• Companies: U.S.Food and Drug Administration and Bristol Myers Squibb
• Keywords: Safety surveillance; adverse events; Bayesian

Abstract

In recent years, a wide array of statistical methods have been developed to facilitate the safety surveillance of a medical product. These methods aim to identify safety signals by analyzing adverse events (AE) from clinical datasets. Berry and Berry (Biometrics, 2004), Xia et al. (J Biopharm Stat, 2011) discuss challenges with analyzing multiple adverse events and propose Bayesian models to address these challenges. Jung et al. (J Biopharm Stat., 2020), and the references therein, discuss likelihood ratio (LRT) based methods for detecting safety-signal from one or more datasets. Our article describes a Bayesian framework to identify potential safety signals by borrowing information from multiple clinical datasets. Our proposed framework can handle zero events for one or more AEs. In addition, our approach can accommodate more than one AE and account for interrelationship among these AEs. We discuss the motivation, advantages of our framework and demonstrate our approach using a real dataset.