Improved inference for fixed-effects meta-analysis of 2 × 2 tables

Area of Science:

• Biostatistics
• Medical Research Methodology

Background:

• Meta-analysis of 2x2 tables is crucial for analyzing adverse events and survey data.
• Traditional fixed-effects inference assumes homogeneity, which is often unrealistic in practice.
• Existing methods may yield unreliable estimates when study heterogeneity is present.

Purpose of the Study:

• To derive improved confidence intervals for common meta-analysis estimators under heterogeneity.
• To provide a more interpretable and accurate approach for meta-analyzing binary data.

Main Methods:

• Developed novel confidence intervals for widely-used meta-analysis estimators that are robust to heterogeneity.
• Validated the proposed methods through simulation studies, assessing coverage levels under various conditions.
• Applied the new methods to a real-world meta-analysis of sclerotherapy trials.

Main Results:

• The derived confidence intervals provide coverage closer to the nominal level in the presence of heterogeneity for both small and large samples.
• Conventional confidence intervals derived under homogeneity assumptions can be overly conservative or anti-conservative.
• The proposed methods offer more accurate characterization of estimator uncertainty.

Conclusions:

• The developed methods offer a more accurate and interpretable approach to meta-analyzing binary data, especially when study heterogeneity exists.
• These improved confidence intervals enhance the reliability of findings in fields like clinical trials and epidemiological research.
• The study demonstrates the practical utility of robust statistical methods in addressing common challenges in meta-analysis.