Meta-analyses and overviews of randomised trials.

The number of meta-analyses or overviews appearing in the literature is rapidly increasing. Because of the problems inherent in non-randomised studies, which are not diminished by using meta-analysis techniques, this paper only refers to meta-analyses of randomised trials. Examples are given of the value of meta-analyses of randomised trials in many different circumstances and the different methods that have been used are discussed. The most reliable type of meta-analysis uses individual patient data and includes all trials fitting the pre-defined criteria, published and unpublished. As with a single randomised trial, large numbers of randomised patients are needed to give reliable results. Well-conducted meta-analyses or overviews are the best method of summarising all available unbiased evidence on the relative effects of treatments.