ABSTRACT
As pointed out in Wallis (1980), sequential analysis was born in response to demands for more efficient testing of anti-aircraft gunnery during World War II, culminating in Wald's development of the sequential probability ratio test which had an immediate impact on weapons testing. Since the medical community widely accepts hypothesis testing as a means of assessing the reproducibility of the results of an experiment, it was recognized within a few years of Wald's seminal work that the sequential approach might provide a more efficient design of clinical trials to test new medical treatments. Despite substantial methodological development in this direction, sequential analysis received little attention from the biomedical community until the early 1980's follow-
ing early termination of the Beta-Blocker Heart Attack Trial (BRAT). The main reason for this lack of interest was that the fixed sample-size ("sample-size" meaning the number of patients accrued) for a typical trial was too small to allow further reduction by a sequential design while still maintaining reasonable power at the alternatives of interest. On the other hand, BRAT's survival endpoint had to be monitored over a period of 4 years, with all patients accrued within the first 27 months and with periodic reviews of the data by a Data and Safety Monitoring Board. The trial was terminated at one of these reviews, 8 months prior to its prescheduled ending. Since interim reviews of the data are usually incorporated in the design and execution of long-term clinical trials at least for the purpose of monitoring safety of the treatments, these trials (of which BRAT is an illuminating example) are particularly suited to group sequential designs.
