Abstract

The introduction of carefully controlled clinical trials in the 1940s was a welcome advance. The use of controls, randomisation, blinding techniques and cross-over designs improved the credibility of the data, and the result of any such trial was usually sufficient to determine ‘rational therapy’. Since then the limitations of using data from a single randomised controlled trial have been recognised and to overcome these, attempts have been made to combine results of several such trials. Systematic overview, or meta-analysis, has become a standard statistical approach for examining the efficacy of treatments. In brief, systematic overviews involve the objective, systematic analysis of the information from all the randomised, controlled trials of a particular treatment that have been performed. The trials need not be the same in all respects; they should broadly seek to answer the same clinical question, but details such as particular patient characteristics and drug doses used may differ. Meta-analysis makes it possible to estimate with increased confidence whether or not a particular treatment improves outcome, the probable size of such an improvement, and whether further studies are needed.