An introduction to the statistical design of phase III cancer clinical trials

Unless a phase III clinical trial is properly designed, it may be impossible to draw valid conclusions concerning the relative efficacy of the treatments being studied. The purpose of this paper is to present the basic principles of trial design as related to treatment assignment by randomization, the stratification for prognostic factors, the number of treatments to be compared and the determination of the sample size. Emphasis is placed on the distinction between the number of events and the number of patients required in time to event analyses and between trials designed to detect a difference and trials attempting to show treatment equivalence.