Multivariate meta-analysis – is it worth the extra effort

Traditionally, meta-analysis models combine summary estimates of a single quantitative endpoint, taken from different studies, to produce a single pooled result to aid evidence-based clinical decision-making. Multiple pooled results are required when there are multiple endpoints of interest across studies, such as multiple outcomes (e.g. overall and disease-free survival), multiple time-points (e.g. in longitudinal studies) and multiple treatment groups. In this situation, a pooled result for each endpoint is usually obtained by applying a separate univariate metaanalysis to each endpoint independently. However, an alternative option is a multivariate meta-analysis model that jointly synthesises the multiple endpoints and obtains all the pooled results collectively [1-4]. This approach allows the correlation between endpoints to be taken into account, both within-studies and betweenstudies, but unfortunately it is rarely used in practice. The reasons for this may include tradition, the increased complexity of the multivariate approach, and perhaps a lack of understanding as to when multivariate meta-analysis is beneficial.