The role of randomization in clinical studies: myths and beliefs

Abstract

On the basis of a survey of the methodological literature, we analyze widespread views on randomization and the advantage of randomized over nonrandomized studies. These views follow from theoretical considerations and at least three types of empirical investigations into the results of published studies. Randomization is often credited with advantages that it does not possess or confer. Several popular theoretical arguments in favor of randomization are shown to be either incorrect or imprecise. The published empirical comparisons of randomized with nonrandomized studies have methodological weaknesses and do not give any convincing information about the value of carefully designed and conducted nonrandomized studies. Six arguments, most of which are pragmatic rather than epistemological, are given to support our belief that randomization should not be avoided without compelling need. We conclude that although there are good arguments in favor of randomization, these are not the ones usually found in the literature. The very negative view on nonrandomized studies sometimes encountered in biostatistics and medicine may be comprehensible from a historical, pragmatic, or educational viewpoint, but it is not well founded on epistemological grounds.