Heterogeneity in Meta-analysis: Concepts, Quantification, Exploration, and Clinical Interpretation

Abstract

Heterogeneity is a central consideration in meta-analysis because included studies commonly differ in populations, interventions, comparators, outcome definitions, study conduct, and risk of bias. This narrative methodological review summarizes the conceptual basis, statistical assessment, exploration, and reporting of heterogeneity in pairwise meta-analysis. Clinical and methodological diversity should be assessed before pooling, whereas statistical heterogeneity is evaluated from the study-specific results during quantitative synthesis. Cochran’s Q statistic, the I2 statistic, between-study variance (τ2), and prediction intervals provide complementary rather than interchangeable information. Fixed-effect models assume a common underlying effect, whereas random-effects models estimate an average effect across a distribution of true effects and require careful interpretation. Heterogeneity should be anticipated, interpreted in a clinical context, and transparently reported rather than treated solely as a statistical nuisance. Meta-analysts should avoid rigid I2 thresholds, report τ2 and prediction intervals where feasible, distinguish effect modification from bias, and consider whether a pooled estimate remains clinically meaningful.