Sample Size: Profound Implications of Mundane Calculations

Controversies in the design of clinical trials excite a small but feisty group of academicians. On one side of the controversy is a group I call the "rigorists," who argue that clinical medicine must be based only on facts that are established by scientifically rigorous clinical trials.1,2 Arguments for scientific rigor have an attractive purity about them that is both mathematical and moral. The simple certainties of statistical science can lead one to believe that, if only we would live by them, we would achieve not only low P values but universal peace and harmony as well. Thus, it may be considered not just good science but an ethical requirement of good research that studies be of sufficient size to give statistically significant results. No moral argument can confront such assertions head on. Nobody argues, for example, that studies ought to be of insufficient size to give meaningful results. Instead, on the other side of this debate we find a group whom I call "apologists," who make arguments about why, even though scientific rigor is the highest good, other considerations sometimes dictate that medical practice be based on knowledge that falls short of the highest scientific standards. These arguments can be based on the physician's primary loyalty to his or her patient,3-5 to the difficulties in deciding whether or not a sufficient state of uncertainty about the relative merits of two treatments exists to ethically justify clinical trials,6,7 or on the special difficulties in obtaining informed consent or randomizing patients in particular populations.8-10