Allocation of Subjects in Medical Experiments

Abstract

Randomized clinical trials have been a subject of much controversy for ethical reasons. When alternative experimental designs are judged on the basis of expected outcomes to the entire patient population, including the experimental subjects, it can be demonstrated that classical randomized designs need not be optimal. Consideration of individual patient preferences in a decision-analytic framework may lead to rejection of randomized trials in favor of alternative, perhaps nonrandomized, designs. Moreover, the ethical problems of experimentation can be rendered less severe if adaptive designs are used — i.e., designs that make use of information obtained during the investigation, thus permitting meaningful inferences to be drawn while exposing fewer patients to the inferior treatment. For both statistical and ethical reasons, adaptive designs should be used more often, and strictly randomized designs should be used sparingly. (N Engl J Med 291:1278–1285, 1974)