A statistical method for removing unbalanced trials with multiple covariates in meta-analysis

Area of Science:

• Biostatistics
• Medical Research Methodology
• Quantitative Synthesis

Background:

• Existing meta-analysis literature lacks robust quantitative methods for assessing trial 'combinability'.
• Covariate balance is a critical but often overlooked prerequisite for valid meta-analyses.
• Inaccurate trial selection can compromise the statistical integrity and generalizability of meta-analysis results.

Approach:

• Proposes a novel four-stage statistical method to identify and eliminate unbalanced randomized controlled trials (RCTs).
• Employs the combined Anderson-Darling test on Empirical Cumulative Distribution Functions (ECDFs) of meta-arms.
• Validates the method using datasets from established meta-analyses.

Key Points:

• The method effectively identifies trials with significant covariate imbalance.
• Addresses the neglect of quantitative and statistical assessments in meta-analysis literature.
• Provides a practical procedure for enhancing the quality and reliability of meta-analytic data.

Conclusions:

• The proposed method offers a simple yet powerful tool for improving meta-analysis rigor.
• Enhances the 'combinability' of trials by ensuring covariate balance.
• Contributes to more dependable and accurate synthesized evidence in medical research.