The reliable detection of moderate differences in major health outcomes that arise as a result of treatment requires large-scale randomized evidence (and the appropriate interpretation of this evidence once it has been generated). This may take the form of a single mega-trial or, exceptionally, a meta-analysis of many smaller randomized trials may provide worthwhile information. Small or non-randomized studies cannot generally be trusted to distinguish reliably between a moderate benefit, a moderate hazard, and a negligible difference in major outcomes. Simple design, streamlined data collection, and use of the "uncertainty principle' to guide eligibility would all encourage the recruitment of larger samples in randomized trials. Future trials need to adopt these methods in order to detect any moderate improvements in major outcomes that may await discovery.



Br J Clin Pharmacol

Publication Date





349 - 353


Clinical Trials, Phase III as Topic, Ethics, Medical, Humans, Meta-Analysis as Topic, Outcome Assessment (Health Care), Randomized Controlled Trials as Topic, Reproducibility of Results, Risk Factors, Sample Size, Selection Bias, Treatment Outcome