A robust meta-analytic approach for evaluating regional treatment effect in a multi-regional clinical trial.

Abstract

A multi-regional clinical trial under one protocol is the gold standard in global drug development. In addition to the positive benefit-risk ratio in the overall population, evaluating local treatment effects and assessing regional consistency are significant components of labeling discussions with the local health authorities. However, relatively small sample sizes in some regions lead to lower precision in the local treatment effect estimates. Such hurdles can be overcome by using results from other regions while estimating the region-specific treatment effect. Commonly used methods are complete pooling of data from specific regions assuming constant effect, and a Bayesian hierarchical model assuming exchangeability or similarity. Pooling patient data from multiple regions increases the precision of the treatment effect estimated in one region provided the medical practice is similar across regions. However, such an approach ignores heterogeneity between different regions and can cause an unrealistic estimate. On the other hand, the Bayesian hierarchical model considers the between-region heterogeneity and provides shrinkage estimates for region-specific treatment effects. However, shrinkage estimators are often criticized for providing too optimistic information borrowing in the presence of outliers. This article presents a robust meta-analytic approach for estimating region-specific treatment effects. While improving the precision, the proposed assessment is less influenced by other regions and provides a more robust estimation of the treatment effect. Simulation studies and a real-life example illustrate the proposed methodology.