Combined Estimation of Treatment Effects Under a Discrete Random Effects Model

Combining information from different groups of data, such as regions in a multi-regional study or trials in a meta-analysis, is an increasingly important problem in clinical drug development. This paper focuses on the combination of treatment effect estimates from independent sources of data (e.g., regions, trials) under a discrete, patient-level random effects model. The approach is motivated by multi-regional clinical studies, being also illustrated in the context of meta-analysis. Comparisons to traditional combination of information methods based on fixed effects (multi-regional trials) and study-level random effects (meta-analysis) are also discussed.