Bias in the analysis and reporting of randomized controlled trials.

The most reliable information on any type of medical intervention is provided by the results of randomized clinical trials (RCTs). In response to increasing pressure to make effective use of limited resources, increasing numbers of health professionals rely on the medical literature, in particular reports of RCTs. However, RCTs may be influenced by a number of factors that introduce bias during the conduct, analysis, and reporting of the trial. Trials may be described as random, when in fact only quasi-random means of treatment allocation have been used; patients may be selectively removed from the analysis; and the report may restrict presentation to or give undue emphasis to only the analyses that yield positive results. The implications of such bias are discussed with particular reference to the effect that they may have on reviews and meta-analyses.