Computation of the Properties of Multi-Stage Clinical Trial Design Based on SCPRT

Abstract

Multi-stage clinical trial, also termed group sequential clinical trial, is commonly used design for clinical trials to evaluate a new treatment against existing one(s). Such design allows the trial to stop early for pronounced treatment effect or the lack of it thereof based on data accumulated at the intermediate stage. The sequential conditional probability ratio test (SCPRT) approach is derived based on the concept of discordance probability, namely, the probability that the decision to accept or reject the null hypothesis based on interim data would reverse should the trial continue to the planned end. This probability is controlled at a preset small level. It is one of the intuitively appealing procedures along with stochastic curtailing but the procedure has been shown to be more efficient than the procedures based on stochastic curtailing. However, in the existing SCPRTs, the discordance probability, type I error and power are not easy to compute. Here we investigate a simulation based method to compute these quantities and apply them to a real data problem as an illustration.