Control of error in randomized clinical trials

Abstract

The randomized clinical trial is the method of choice for comparing the effects of alternative care options, both in its own right and as the cornerstone of systematic reviews of the subject. Errors in such trials, therefore, have major consequences for health care. This paper provides a brief introduction to the major sources of such errors, whether they be systematic or chance errors. It addresses selection bias, due to either biased entry in or biased exclusion from the trial, bias in assessing outcomes, and biases due to contamination or co-intervention. Random errors, including type I and type II errors, are discussed along with ways in which they can be minimized. Small clinical trials, in particular, provide a major problem not only by themselves, but also if they become incorporated in systematic reviews without appropriate consideration of the phenomenon of publication bias.