A confidence interval robust to publication bias for random-effects meta-analysis of few studies

TL;DR

This paper introduces a new confidence interval method for random-effects meta-analysis that is more robust to publication bias, especially effective when analyzing few studies, improving coverage probabilities over existing methods.

Contribution

It proposes a variation of Henmi and Copas's method with an improved heterogeneity estimator tailored for small-sample meta-analyses.

Findings

The new method outperforms competitors in coverage probability.

It shows significant improvement over previous methods in simulations.

Application to a pediatric immunosuppression meta-analysis demonstrates practical utility.

Abstract

Systematic reviews aim to summarize all the available evidence relevant to a particular research question. If appropriate, the data from identified studies are quantitatively combined in a meta-analysis. Often only few studies regarding a particular research question exist. In these settings the estimation of the between-study heterogeneity is challenging. Furthermore, the assessment of publication bias is difficult as standard methods such as visual inspection or formal hypothesis tests in funnel plots do not provide adequate guidance. Previously, Henmi and Copas (Statistics in Medicine 2010, 29: 2969--2983) proposed a confidence interval for the overall effect in random-effects meta-analysis that is robust to publication bias to some extent. As is evident from their simulations, the confidence intervals have improved coverage compared with standard methods. To our knowledge, the properties of their method has never been assessed for meta-analyses including fewer than five studies. In this manuscript, we propose a variation of the method by Henmi and Copas employing an improved estimator of the between-study heterogeneity, in particular when dealing with few studies only. In a simulation study, the proposed method is compared to several competitors. Overall, we found that our method outperforms the others in terms of coverage probabilities. In particular, an improvement compared with the proposal by Henmi and Copas is demonstrated. The work is motivated and illustrated by a systematic review and meta-analysis in paediatric immunosuppression following liver transplantations.