Designs for efficient clinical trials

Abstract

This article reviews statistical designs for the conduct of efficient Phase II and Phase III clinical trials. Two stage designs for Phase II trials are tabulated which minimize the number of patients entered when the experimental treatment is inactive or not promising. Randomized Phase II trials for selecting the most promising of several experimental regimens are also described. Phase III designs are presented with simple early stopping rules for dropping experimental regimens. Inappropriate sequential analysis of interim data is a likely cause of many spurious claims of treatment effectiveness. General methods for the valid calculation of confidence limits for differences in treatment effectiveness based on sequentially accumulating data are described. Factorial designs for answering "two questions for the price of one" are also discussed.