Data visualisation for meta-analysis

Graphical displays play a pivotal role in understanding data sets and disseminating results. For meta-analysis, they are instrumental in presenting findings from multiple studies. This report presents guidance to authors wishing to submit graphical displays as part of their meta-analysis to a clinical cardiology journal, such as Heart. When using graphical displays for meta-analysis, we recommend the following: • Use a flow diagram to describe the number of studies returned from the initial search, the inclusion/exclusion criteria applied and the final number of studies used in the meta-analysis. • Present results from the meta-analysis using a figure that incorporates a forest plot and underlying (tabulated) statistics, including test for heterogeneity. • Use displays such as funnel plot (minimum 10 studies) and Galbraith plot to visually present distribution of effect sizes or associations in order to evaluate small-study effects and publication bias). • For meta-regression, the bubble plot is a useful display for assessing associations by study level factors. • Final checks on graphs, such as appropriate use of axis scale, line pattern, text size and graph resolution, should always be performed.