An alternative parameterization of Bayesian logistic hierarchical models for mixed treatment comparisons

Area of Science:

• Biostatistics
• Health Economics
• Clinical Epidemiology

Background:

• Mixed treatment comparison (MTC) models are crucial for synthesizing evidence from multiple randomized clinical trials (RCTs).
• Current MTC models rely on specific parameterizations that can be complex to implement and interpret.
• Simplifying MTC models could enhance their accessibility and application in evidence synthesis.

Purpose of the Study:

• To introduce and evaluate a novel treatment-based parameterization for MTC models.
• To compare the performance of the proposed model against traditional MTC approaches using simulations and real-world data.
• To assess the model's simplicity, speed, and ease of implementation.

Main Methods:

• Development of a treatment-based parameterization estimating outcomes at both study and treatment levels.
• Comparison via simulation studies.
• Validation using three distinct RCT datasets from systematic reviews (cirrhosis bleeding, antihypertensive drugs in diabetes, smoking cessation).

Main Results:

• Simulation results indicated that the treatment-based MTC model performed similarly or better than conventional MTC models.
• Real data analyses showed minimal differences in inferences between the proposed and commonly used MTC models.
• The proposed MTC approach demonstrated comparable or superior performance, with added benefits of simplicity and speed.

Conclusions:

• The proposed treatment-based parameterization offers a viable and potentially advantageous alternative to existing MTC models.
• This simplified approach is easier to implement in standard statistical software, facilitating broader use in evidence synthesis.
• The model provides robust and reliable results, comparable to or exceeding traditional methods.