Bayesian inference for network meta-regression using multivariate random effects with applications to cholesterol...

Area of Science:

• Cardiovascular Medicine
• Biostatistics
• Pharmacology

Background:

• Low-density lipoprotein cholesterol (LDL-C) is a key factor in atherosclerosis and coronary heart disease.
• Statins and Ezetimibe are primary therapies targeting LDL-C through different mechanisms: liver synthesis inhibition and intestinal absorption inhibition, respectively.
• Synthesizing evidence from multiple clinical trials is crucial for evaluating cholesterol-lowering drug efficacy and safety.

Purpose of the Study:

• To propose and validate a new Bayesian inference approach for arm-based network meta-regression.
• To develop a novel strategy for grouping random effects variances based on treatment mechanisms.
• To apply the methodology to analyze aggregate network meta-data from statins +/- Ezetimibe trials.

Main Methods:

• Utilized aggregate data from 29 double-blind, randomized, controlled clinical trials involving statins +/- Ezetimibe.
• Developed a Bayesian meta-regression framework with a new variance grouping strategy for random effects.
• Employed Bayesian model comparison to select optimal treatment group variances.
• Implemented a Markov chain Monte Carlo (MCMC) sampling algorithm for posterior computations.

Main Results:

• The proposed Bayesian meta-regression approach was successfully applied to analyze data from 29 trials with 11 treatment arms.
• The novel variance grouping strategy demonstrated utility, particularly for treatment arms appearing in single trials.
• The methodology facilitated the synthesis of results from diverse cholesterol-lowering drug trials.

Conclusions:

• The developed Bayesian arm-based network meta-regression approach offers a robust method for synthesizing evidence from complex clinical trial networks.
• The variance grouping strategy enhances the analysis of treatment effects, especially in scenarios with limited data per arm.
• This methodology provides valuable insights into the comparative effectiveness of cholesterol-lowering therapies.