Interim analyses and stopping rules in cancer clinical trials.

A clinical trial conducted according to a schedule of interim analyses written into the protocol, and stopped according to a predetermined rule, is known to statisticians as a sequential clinical trial. This methodology is becoming more widely used in trials concerning life-threatening diseases because of its ability to adjust the sample size to the emerging information on treatment efficacy. When treatments under comparison differ appreciably, small samples will be sufficient; for more subtle differences larger numbers of patients need to be recruited. Sequential methods have already been used in certain cancer clinical trials, and they are especially appropriate for such studies. In this paper the principles of sample size determination are reviewed, and the essential aspects of designing sequential trials are described. The necessity for a special form of statistical analysis following a sequential trial is explained, and the consequences of early or late stopping on the analysis are investigated. Compromises which have to be made between the formal requirements of theory and the practical realities of trial conduct are discussed.