A non-iterative extension of the multivariate random effects meta-analysis

Area of Science:

• Biomedical Research
• Statistical Methods
• Meta-Analysis

Background:

• Multivariate meta-analysis methods are increasingly used in biomedical research.
• Existing techniques, like the multivariate DerSimonian and Laird method, face computational challenges.
• Non-iterative and iterative approaches have been developed to address these issues.

Purpose of the Study:

• To propose an extension of the Hartung and Makambi (2002) and Makambi (2001) methods for multivariate meta-analysis.
• To evaluate the performance of the proposed method against the multivariate DerSimonian-Laird approach.

Main Methods:

• Development of a novel multivariate meta-analysis method based on existing work.
• Conducting a simulation study to compare bias and mean square error.
• Application of the proposed method to a real-world example.

Main Results:

• The proposed multivariate meta-analysis approach demonstrated superior performance in specific circumstances compared to the multivariate DerSimonian-Laird method.
• Simulation results indicated favorable bias and mean square error characteristics for the new method.

Conclusions:

• The proposed extension offers a viable and potentially improved alternative for multivariate meta-analysis in biomedical research.
• The method provides a valuable tool for researchers dealing with complex multivariate data in meta-analyses.