The impact of correction methods on rare-event meta-analysis

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Evidence Synthesis

Background:

• Meta-analyses are vital for synthesizing clinical trial evidence, particularly for rare or adverse outcomes.
• Common meta-analysis methods struggle with zero events in trial arms, impacting statistical power.
• Continuity corrections (CCs) are numerical adjustments to address issues with zero events in meta-analyses.

Purpose of the Study:

• To evaluate the performance of various continuity correction methods in meta-analyses with rare events.
• To assess how CCs interact with different pooling methods and heterogeneity variance estimators.
• To guide researchers in selecting appropriate meta-analysis methods for rare-event data.

Main Methods:

• A simulation study was conducted to compare multiple continuity correction methods.
• The study examined combinations of CCs with various common meta-analysis pooling methods and heterogeneity estimators.
• Key meta-analysis outcomes, including overall treatment effect, confidence interval coverage, and Type I error rate, were analyzed.

Main Results:

• No single continuity correction and pooling method combination is universally optimal.
• When heterogeneity variance is absent, most pooling methods perform well with appropriate CCs.
• Excluding studies with zero events is effective without heterogeneity; excluding single-zero studies is not recommended with heterogeneity.

Conclusions:

• The choice of continuity correction and pooling method significantly impacts meta-analysis results for rare events.
• Researchers should carefully consider the presence of heterogeneity and zero events when selecting methods.
• A web application is available to assist in method selection for rare-event meta-analyses.