Accurate confidence intervals for risk difference in meta-analysis with rare events

Area of Science:

• Biostatistics
• Clinical Trials
• Statistical Modeling

Background:

• Meta-analysis is crucial for estimating treatment effects from multiple studies.
• Traditional methods struggle with rare binary outcomes, such as safety data in clinical trials.

Purpose of the Study:

• To propose and evaluate a novel method for computing confidence intervals for risk difference in meta-analysis involving rare events.
• To assess the performance of the proposed importance sampling method against existing techniques.

Main Methods:

• Utilized importance sampling to calculate confidence intervals for risk difference in meta-analysis.
• Compared proposed intervals with existing fixed- and random-effects models and the Tian et al. (2009) interval.
• Conducted extensive simulation studies under homogeneity and heterogeneity assumptions.

Main Results:

• The proposed accurate interval, based on the random-effects model for sample space ordering, demonstrated satisfactory performance under heterogeneity.
• Traditional fixed-effects model intervals performed well under study homogeneity.
• Simulation studies evaluated coverage probability and average interval length.

Conclusions:

• The proposed importance sampling method provides a valuable tool for meta-analysis with rare events.
• The choice of method (fixed- vs. random-effects) depends on the homogeneity or heterogeneity of study effects.
• Accurate confidence intervals are crucial for reliable interpretation of rare event data in clinical research.