Methods to estimate the between-study variance and its uncertainty in meta-analysis.

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Meta-analyses commonly estimate overall effects but also aim to infer between-study variability.
• The DerSimonian and Laird method for estimating between-study variance is widely used but has faced criticism.

Purpose of the Study:

• To identify methods for estimating between-study variance and its uncertainty.
• To summarize existing simulation and empirical evidence comparing these methods.

Main Methods:

• Literature search to identify estimators for between-study variance and methods for confidence intervals.
• Qualitative evaluation of simulation and empirical studies comparing these methods.

Main Results:

• Identified 16 estimators for between-study variance and 7 methods for confidence intervals.
• Simulation studies indicate Paule-Mandel (dichotomous/continuous) and REML (continuous) estimators are better alternatives.
• Q-profile and generalized Cochran methods are recommended for confidence intervals.

Conclusions:

• Paule-Mandel and REML estimators show promise for between-study variance estimation.
• Q-profile and generalized Cochran methods are recommended for confidence interval calculation.
• Further extensive simulation studies are needed for definitive evidence-based recommendations.