• Top
• Oral presentation

Oral presentation

Random effects meta-analyses offer a means of combining results from clinical trials in the presence of heterogeneity. The treatment effects in the different trials are assumed to be non-identical, but related through a distribution, usually a normal distribution. The results of a random effects meta-analysis include estimates of the mean and standard deviation of this distribution. Often in practice, however, only the mean and its confidence interval are reported. Indeed, the software for generating analyses in Cochrane reviews does not present the standard deviation. We believe this is misguided, and demonstrate empirically that it can yield misleading results. The confidence interval for the mean incorporates uncertainty about the mean due to the heterogeneity, but does not reflect the extent of heterogeneity among the trials.

We have developed simple alternative ways of presenting random effects meta-analyses that explicitly describe the extent of the heterogeneity by concentrating on the entire distribution of treatment effects. These are most easily calculated from Bayesian meta-analyses. However, we concentrate here on classical versions that would usefully supplement random effects meta-analyses currently appearing in Cochrane reviews.

Have something to say? Post a comment on this article!