Abstract:
|
Different clinical trials conducted by different pharmaceutical companies often study the same ingredients and a comprehensive estimate pooling information of individual trials is desirable. At FDA, Individual participant data (IPD) are available and provide a unique opportunity for individual level meta analysis, which adjusts for baseline characteristics and potential confounding variables, leading to increased power. In this study, we propose a novel method for IPD M-A for the incidence rate of rare adverse events. This exact likelihood method, built on a Poisson-Gamma hierarchical model, exhibits superior performance in terms of bias and coverage probability over the conventional approximate approaches including zero inflated models in extensive simulation studies. The methods are illustrated with a real set of trials at FDA.
|