|
Abstract:
|
Post-marketing safety studies are increasingly used to monitor the potential risk of specific drug-induced adverse events (AEs). Among the various types of study designs, the observational cohort study is an appropriate study design to identify the possible causal relationship between drugs and rare/severe AEs. The current single-group cohort study, based on the frequentist approach, pre-specifies a single reference level for comparison, suggesting that the background rate is a fixed constant. In many cases, however, there is uncertainty regarding the background rate. In this talk, a Bayesian monitoring procedure is proposed which can account for the inherent variability in the background rate and also allow for monitoring during the course of the study. The Bayesian procedure can improve the study efficiency by possibly terminating the study early while keeping a similar level of sample size.
|
