Evaluation of interventions

The first modern randomized controlled trial was published in 1948, and featured randomly allocated treatment groups, blinded outcome assessment, and a sufficient number of patients. Randomized controlled trials are now accepted as the best possible way to assess the effects of clinical interventions, but good trials sometimes disagree. Although the elements of a credible trial have been defined, the relative importance of these elements to results is not known. Information from clinical trials may be difficult to use in the care of patients if trial outcomes are not patient centered or the conditions of the trial are very different from the situation in clinical practice. Observational studies are often, but not always, an adequate substitute from randomized trials. Systematic reviews with meta-analyses may reveal patterns of results when individual trials do not, and so complement the information content of large trials. When trials disagree, resolution may be found not by from more trials but also from critical appraisal of individual trials, weighing the totality of the evidence, and Bayesian reasoning. (C) 2002 Elsevier Science Inc. All rights reserved.