A Group Sequential Procedure for Interim Treatment Selection

We consider a group sequential design for interim treatment selection where several treatments are compared with a control. Depending on interim treatment comparisons, the number of winners after an interim look is random. In this article, we discuss the case when there are up to two winners and propose a multiple comparison approach. We derive the distribution of the test statistic and the sequential boundaries. We show that the proposed procedure strongly controls the family-wise error rate. The proposed procedure is uniformly more powerful than some existing procedures as suggested by simulations.