An Application of the Beta-Binomial Model to Combine and Monitor Medical Event Rates in Clinical Trials

Abstract

Safety assessment is an important part of all clinical trials. Since safety data, especially those related to the occurrence of adverse medical events, are typically pooled from different studies to help us understand the safety profile of a new drug, we need to apply statistical techniques that allow us to combine data from different studies in a reasonable manner. In this paper, we propose to apply the beta-binomial model to summarize the medical event rates observed in different trials. We use results obtained by Kleinman (1) to estimate the parameters in the proposed beta-binomial model. In addition to providing an estimate for the average medical event occurrence rate, the estimated model can be used to look for unusual medical event patterns in future trials. We will illustrate the application by an example.