Sample sizes for phase II and phase III clinical trials: an integrated approach

In this paper the following problem of clinical research is explored. Several potential new treatments are available for use against a certain disease. These are evaluated in a series of pilot studies which will constitute phase II clinical trials. The most promising will then be compared with a standard treatment in a phase III trial. Of interest will be the number of patients needed for the complete research programme, the proportions of these that should be involved in each phase, and the number of treatments which should be tried. Optimal strategies are found which maximize the probability that the overall programme identifies a treatment which is significantly better than the standard.