metapreg

metapreg performs meta-analysis, network meta-analysis, and meta-regression of binomial proportions using generalized linear models in Stata.

Key Features:

• Binomial models: Implements binomial distribution-based models for meta-analysis of proportions without requiring normal approximation.
• Logistic models: Employs logistic regression models for analyzing binomial proportions in meta-analytic frameworks.
• Logistic-normal models: Utilizes logistic-normal models to account for over-dispersion and between-study heterogeneity.
• Network meta-analysis: Supports network meta-analysis of binomial proportions for comparing multiple treatments or interventions.
• Meta-regression: Performs meta-regression to explore sources of heterogeneity across studies.
• No continuity correction: Analyzes binomial data without requiring ad-hoc continuity correction or data imputation.

Scientific Applications:

• Evidence synthesis of binomial data: Synthesizes proportion data from multiple studies to estimate population-averaged proportions.
• Meta-analysis of proportions: Analyzes binary outcome proportions across multiple studies with improved statistical efficiency.
• Comparative effectiveness research: Conducts network meta-analyses to compare multiple interventions using binomial outcome data.

Methodology:

metapreg implements generalized linear models including binomial, logistic, and logistic-normal models for meta-analysis, network meta-analysis, and meta-regression of binomial proportions in Stata. The tool was evaluated through a simulation study comparing its estimators with existing estimators in metaprop and metan under conditions including high over-dispersion and small meta-analysis.