Data-based predictive distributions for between-study heterogeneity

In small meta-analyses, a conventional random-effects meta-analysis is problematic because between-study heterogeneity is imprecisely estimated, and this imprecision is not taken into account. A Bayesian meta-analysis allows researchers to incorporate external evidence on the likely extent of between-study heterogeneity in their particular research setting.

Cochrane Scientific Committee recommendation (see the meeting minutes and supporting document from June 2018):
The Committee recommends that Cochrane Reviewers are encouraged to add Bayesian meta-analysis alongside the traditional techniques included in RevMan to supplement and improve their review. Particularly where there is a very high or low heterogeneity estimate. Therefore, in these situations an additional Bayesian analysis will have the greatest impact. This will be included in the new updated Handbook chapter.

If you have any questions, contact the Cochrane Methods Team (methods@cochrane.org).