Egger’s test is commonly used to assess potential publication bias in a meta-analysis via funnel plot asymmetry (Egger’s test is a linear regression of the intervention effect estimates on their standard errors weighted by their inverse variance). The performance of Egger’s and related tests has been extensively studied for binary outcomes, but not for continuous ones. Comparative continuous outcomes are commonly measured on an absolute (mean) difference scale, and it is not uncommon for the magnitude of effect to be related to response in the control arm (i.e. baseline risk). When this is the case, funnel plots can appear highly asymmetric, even when publication bias is not present, since correlations between outcome and both effect size and its standard error exist. Through application to a motivating collection of meta-analyses of post-operative analgesics, and simulation studies, we will show that Egger’s test is potentially misleading for continuous outcomes and a test which regresses the residuals from a meta-regression model, including baseline risk as a study-level covariate, against inverse sample size has better statistical properties. 

These videos, originally part of the Cochrane Learning Live webinar series, are highly relevant to reviewers, editors and statisticians with interests in dealing with bias in meta-analysis. Although there is some statistical content, concepts are explained visually where possible keeping much of the material accessible.

The webinar was delivered in July 2020 and below you will find the videos from the webinar, together with accompanying slides to download [PDF].

Part 1: Introduction to publication bias
Part 2: Example: postoperative pain
Part 3: Simulation study, limitations and conclusions
Part 4: Questions and answers

Presenter Bios

Suzanne Freeman is an NIHR Research Fellow and member of the NIHR Complex Reviews Support Unit. Her research interests include network meta-analysis, individual participant data meta-analysis, synthesis of time-to-event and continuous outcomes and diagnostic test accuracy meta-analysis.

Alex Sutton has had a long-standing interest in methodology for evidence synthesis and has worked on developing methods for publication bias, diagnostic test accuracy meta-analysis and network meta-analysis. He is currently particularly interested in methods for visual communication of information and its application to synthesis methods.

Part 1: Introduction to publication bias

Part 2: Example: postoperative pain

Part 3: Simulation study, limitations and conclusions

Part 4: Questions and answers

Additional materials

Download the slides from the webinar [PDF]