Borrowing of strength and study weights in multivariate and network meta-analysis

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Multivariate and network meta-analysis allow borrowing strength across related effects.
• Current assessment of this 'borrowing of strength' is informal.
• There is a need for formal, quantitative methods to assess borrowing strength.

Purpose of the Study:

• To introduce novel mathematical definitions for quantifying 'borrowing of strength' in meta-analysis.
• To develop a method for calculating study weights within a unified statistical framework.
• To provide tools for more rigorous assessment of evidence synthesis in complex scenarios.

Main Methods:

• Decomposition of the score statistic to compare precision between multivariate and univariate models.
• Derivation of a new framework for calculating study weights.
• Application of proposed methods to meta-analyses with correlated effects.

Main Results:

• Formal definitions of 'borrowing of strength' are presented, emulating informal assessments.
• A method for calculating study weights is integrated into the borrowing of strength framework.
• Demonstrated utility in meta-analyses of multiple outcomes, risk factors, and treatments.

Conclusions:

• The proposed definitions and methods offer a quantitative approach to assessing evidence synthesis.
• This framework enhances the transparency and rigor of multivariate and network meta-analyses.
• Study weights can now be formally reported alongside borrowing strength metrics.