Keywords: multi-regional clinical trials, shrinkage estimator, consistency of treatment effects
Multi-regional clinical trials (MRCTs), defined as a clinical trial conducted in more than one region under a single protocol, inherently anticipates a consistency of treatment effects across regions. We, therefore, calculate overall sample size so that the trial demonstrates statistical significance of the treatment effect based on the data observed in all the regions included in the trial, partitioning it into each region based primarily on the feasibility of patient enrollment along with regulatory requirement. The assessment of consistency in treatment effect across regions under the heterogeneity of sample size is challenging in the MRCT. Some authors proposed to use a shrinkage estimator to assess the consistency of treatment effects. It is natural to use the shrinkage estimator borrowing information from the other regions, but this possibly bears the risk of too much shrinkage of treatment effect. In this study, we propose a simple alternative method that uses a robust estimator to estimate the treatment effect in each region. In the proposed method, we also adjust the robust estimate in each region by using an ordinal shrinkage approach. We evaluate the probabilities of consistency between the treatment effect in a specific region and the overall treatment effect and the probabilities of consistency between the treatment effect in all the regions and the overall treatment effect in the proposed and rival methods through the simulation studies under the several scenarios with respect to the number of regions, sample size for each region, and heterogeneity of treatment effects across regions. We discuss the operating characteristics of the proposed method based on the results of simulation studies.