The Risk of Bias in Randomized Clinical Trials

Medical studies continue to exert a profound influence on future medical policy and medical decision making. This is how it should be. Recognition that not all comparative medical studies are equally reliable, that in fact some designs are better than others, has led to the development of evidence-based medicine. The... read morerandomized clinical trial is, rightly, considered to be the best study design, in terms of providing the most reliable evidence. Though the very act of randomizing by itself removes certain biases, and though it is both tempting and convenient to assume that all biases are thereby removed, the reality is rather more complicated. Even masked randomized trials have a substantial risk of bias, and this risk of bias remains even when the trial is conducted by a prestigious institution and published in a prestigious journal. Our aim is to connect specific design features to a greater or lesser risk of bias, so as to elucidate how best to conduct randomized trials, and also to inform reviewers of trials regarding what types of bias to look for and how to determine their risk in any given trial.
This series will cover key determinants of risk of bias in randomized trials, including improper randomization, imperfect masking, failed allocation concealment, the improper use of surrogate endpoints, the improper use of a pre-randomization run-in phase, statistical analyses whose validity is predicated on unrealistic assumptions, and other similar issues.