REPEATED SIGNIFICANCE TESTING WITH CENSORED RANK STATISTICS IN INTERIM ANALYSIS OF CLINICAL-TRIALS

A simple class of stopping rules is introduced for time-sequential ran k tests to compare time to failure between two treatment groups, such as in the case of a clinical trial in which patients enter serially an d in which interim analyses of the data are performed periodically so that the trial may be stopped early when one treatment is found to be significantly better than the other. These time-sequential rank tests are shown to achieve both savings in study duration and increase in po wer over their nonsequential counterparts, and provide a simple but st atistically efficient method to circumvent the difficulty of ''calenda r time'' versus ''information time'' in the design of group sequential trials with failure-time endpoints.