A decision rule for evaluating several independent clinical trials collectively

Usually, in applying for market approval of a new drug, more than one similarly designed clinical trial is conducted to support efficacy claims. How to evaluate these trials collectively and assess the overall type I error of a decision rule can be a challenging statistical issue. In this paper, we propose a decision rule to collectively evaluate p-values from several similarly designed and independently conducted trials. A concept of overall hypotheses, which is essentially union or intersection combinations of individual trials' hypotheses, is used so that the overall type I error can be controlled at desired levels. We also discuss some important properties of the approach, including the selection of the overall type I error rates and power. Examples are presented.