Methods to calculate uncertainty in the estimated overall effect size from a random‐effects meta‐analysis

Abstract

Meta-analyses are an important tool within systematic reviews to estimate the overall effect size and its confidence interval for an outcome of interest. If heterogeneity between the results of the relevant studies is anticipated, then a random-effects model is often preferred for analysis. In this model, a prediction interval for the true effect in a new study also provides additional useful information. However, the DerSimonian and Laird method-frequently used as the default method for meta-analyses with random effects-has been long challenged due to its unfavorable statistical properties. Several alternative methods have been proposed that may have better statistical properties in specific scenarios. In this paper, we aim to provide a comprehensive overview of available methods for calculating point estimates, confidence intervals, and prediction intervals for the overall effect size under the random-effects model. We indicate whether some methods are preferable than others by considering the results of comparative simulation and real-life data studies.

Physical Activity and the Risk of Liver Cancer: A Systematic Review and Meta-Analysis of Prospective Studies and a Bias Analysis

“…We used the Hartung-Knapp/Sidik-Jonkman random-effects meta-analysis approach (whose advantages were demonstrated previously [26,27]) to combined studyspecific log HR, with the empirical Bayes estimator for the between-study variance (τ²) [26], for estimation of a mean HR, with corresponding 95% confidence intervals and 95% prediction intervals [28]. The 95% CI from a random-effects model contains highly probable values for the mean HR [28,29].…”

Comparative effectiveness of common therapies for Wilson disease: A systematic review and meta‐analysis of controlled studies

“…In the case that at least six studies without zero events could be included in the meta-analysis, 36,37 we performed inverse-variance random effects meta-analyses using the Paule-Mandel between study heterogeneity estimator with modified Hartung-Knapp confidence intervals (CIs). 38,39 For consistency, we used the same model for sensitivity analyses irrespective of the number of studies.…”

“…The primary outcomes were mortality and asymptomatic/improved, the secondary outcomes side effects, early neurological deterioration, treatment discontinuation and OLT. In the case that at least six studies without zero events could be included in the meta‐analysis, we performed inverse‐variance random effects meta‐analyses using the Paule‐Mandel between study heterogeneity estimator with modified Hartung‐Knapp confidence intervals (CIs) . For consistency, we used the same model for sensitivity analyses irrespective of the number of studies.…”

The number needed to treat in pairwise and network meta-analysis and its graphical representation

“…For the NNT CI calculation, an appropriate method should be chosen to calculate CIs for the selected effect measure. For a review of methods to obtain CIs for the estimated overall effect from a random-effects meta-analysis see Veroniki et al [31]. If the chosen effect measure is the RD, then the NNT CI is simply obtained by inverting and exchanging the corresponding RD confidence limits.…”