Recent methodological contributions to clinical trials.

This paper is concerned with the maximum that can be achieved in clinical trials using optimal methodologies if such exist, i.e., with strategy and not tactics. It might be thought that strategy so defined is too vague and too general to be fruitfully discussed, but I hope to convince you of the contrary. Considerations of strategy will involve us in controversial questions about the foundations of statistics, but this prospect shall not inhibit us any more than it would have Reed or Frost. A point of view about statistical methods that was once universal and is still very common is that it provides unique, unequivocal answers to problems of analysis and design—that all statisticians faced with the same problem will, or at least should, provide the same answer. R. A. Fisher's writing, in particular, reflects this view, with continual glimpses furnished by such phrases as "uniquely superior" methods of "absolute validity," "canons of valid inference," "rigorous but uncertain conclusions," etc. His Design of Experiments was dazzling in its self-assurance and clarity and gave statisticians in many fields the inspiration and confidence to apply these unequivocal procedures, and to develop new ones. The extensions to clinical trials have been numerous and important, the key figure being Bradford Hill. In the opinion of the President of the Royal College of Surgeons they constitute as crucial a contribution to medicine as the discovery of penicillin (1). One would naturally have expected that continuing methodological advances would have accompanied these extensions. The statistical literature on