Some statistical considerations for the interpretation of trials of combined androgen therapy

Since 1989, there have been strongly conflicting opinions regarding the validity of the proposition that an antiandrogen added to standard hormonal therapy could improve the outcome in patients with newly diagnosed metastatic prostate cancer. To some degree, these conflicting opinions arose from misinterpretations of a number of "negative" clinical trials. The misinterpretations happened because these "negative" trials had insufficient statistical power to refute the reported "positive trial", and information illustrating their statistical power was absent from their reports. Thus, readers were not given the necessary information to distinguish between negative and inconclusive results. In addition, many of the negative reports were early analyses of ongoing clinical trials, a practice that always should be avoided. This article reviews and illustrates the statistical principles that are applicable to the interpretation of low-powered negative clinical trials and discusses the design and conduct of clinical trials.