Meta-Analysis of Rare Binary Adverse Event Data.

Area of Science:

• Biostatistics
• Clinical Research Methodology
• Pharmacovigilance

Background:

• Meta-analysis of binary adverse event data is crucial for drug safety.
• Rare adverse events present unique challenges in statistical analysis.
• Existing moment-based meta-analytic methods may be biased with rare events.

Purpose of the Study:

• To evaluate fixed-effects and random-effects moment-based meta-analytic methods for binary adverse event data.
• To specifically address the challenges posed by rare adverse events.
• To develop and assess novel methods for improved estimation and hypothesis testing.

Main Methods:

• Analysis of fixed-effects and random-effects moment-based meta-analytic methods.
• Derivation of three new methods: simple average treatment effect estimator, new heterogeneity estimator, and parametric bootstrapping test for heterogeneity.
• Simulation studies to compare statistical properties of traditional and new methods.

Main Results:

• Moment-based estimators for treatment effects and heterogeneity are generally biased, with bias increasing as events become rarer.
• The newly developed methods significantly reduce, but do not entirely eliminate, this bias.
• Comparison of estimators and hypothesis testing methods using a real-world dataset.

Conclusions:

• Traditional moment-based meta-analysis methods exhibit bias in estimating treatment effects and heterogeneity, particularly for rare adverse events.
• The proposed novel methods offer improvements in reducing bias for rare event data.
• Further research and validation are needed for optimal application in clinical practice.