Paule-Mandel estimators for network meta-analysis with random inconsistency effects

Area of Science:

• Biostatistics
• Evidence Synthesis
• Comparative Effectiveness Research

Background:

• Network meta-analysis (NMA) synthesizes evidence from multiple randomized controlled trials.
• Inconsistency, where direct and indirect evidence conflict, poses a challenge in NMA.
• Existing models for NMA with random inconsistency effects include method of moments and maximum likelihood estimation.

Purpose of the Study:

• To extend the Paule and Mandel estimator for use in network meta-analysis models with random inconsistency effects.
• To evaluate the performance of the extended Paule and Mandel method against existing estimation techniques.
• To provide a robust and accurate estimation method for complex NMA.

Main Methods:

• Extension of the Paule and Mandel method for univariate meta-analysis to NMA with random inconsistency effects.
• Application and comparison of three estimation methods: extended Paule and Mandel, method of moments, and maximum likelihood estimation.
• Simulation study using a highly heterogeneous dataset to assess method performance.

Main Results:

• The extended Paule and Mandel method demonstrated satisfactory performance, yielding more accurate inferences than the method of moments.
• The Paule and Mandel method showed advantages over likelihood-based methods, being semiparametric and not requiring convergence diagnostics.
• The proposed methodology is a fully viable alternative to existing estimation methods, including restricted maximum likelihood.

Conclusions:

• The extended Paule and Mandel method is a valuable addition to the toolkit for network meta-analysis.
• This method offers improved accuracy and practical advantages for analyzing complex and heterogeneous treatment networks.
• It provides a robust alternative for estimating average treatment effects in the presence of inconsistency.