Background Randomized clinical trials (RCTs) are often positioned at the top of evidence hierarchies. Meta-analyses of RCTs aim to integrate the state of knowledge on a given scientific question, particularly for rare drug-related outcomes. However, although RCTs are valuable tools in our armamentarium, they are rarely designed to evaluate drug safety and are thus susceptible to limitations that may hamper the ability of both RCTs and meta-analyses to fully characterize the safety profiles of drugs. Their potential limitations might be exacerbated in the study of rare outcomes, often encountered in drug safety assessment, when even minor deviations from the intended randomization could impact the stability of the risk estimates. Purpose This article considers the methodological caveats of both RCTs and meta-analyses of RCTs pertinent to the study of drug-related harms. It is intended to stimulate discussion about the impact of these caveats on interpreting findings of RCTs and meta-analyses for drug safety, which would foster more robust, critical evaluations, and thus enhance clinical and regulatory decision-making. Methods Pertinent issues that can influence the interpretation of drug-related harms discussed in this article were based on authors' expertise and review of the literature. Results Investigators and clinicians should be cognizant of the potential limitations of the secondary use of RCTs and meta-analyses in the assessment of drug-related harms and, when applicable, should consider potential remedies to overcome these limitations. Limitations Only few practical examples are included in the article due to the fact that many of the discussed caveats are not examined and/or reported in many publications. In addition, the confidential nature of data reviewed at a regulatory agency forestalls an in depth discussion of examples pertaining to specific drugs. Furthermore, our ability to quantify the extent of encountering, or the actual impact of, the caveats addressed in this review on the RCTs findings is limited. It is worth noting that the mere encounter of a given caveat does not mean that it will obviate the utility of drug safety information from a given trial. The extent of its impact is expected to vary based on the specifics of the trial, the drugs studied, the indications, and the nature of the adverse events. Conclusions Although some of the limitations described are inherent in RCTs, some of the sources of bias highlighted in this article could be minimized by careful RCT design, planned follow-up, and improved collection of information on adverse events. As future research sheds more light on pertinent knowledge gaps and issues, the ability to maximize the use of RCTs and meta-analyses of RCTs to address drug safety questions of interest will be greatly enhanced. Clinical Trials 2011; 8: 559-570. http://ctj.sagepub.com
