Incorporating data from various trial designs into a mixed treatment comparison model

Area of Science:

• Health economics
• Biostatistics
• Clinical trial methodology

Background:

• Accurate relative treatment efficacy estimates are vital for healthcare decision-making.
• Bayesian mixed treatment comparison (MTC) models are increasingly used for healthcare economic evaluations.
• Current MTC models often exclude data from non-randomized controlled trials, limiting evidence synthesis.

Purpose of the Study:

• To propose and evaluate novel methods for incorporating diverse trial designs into Bayesian MTC models.
• To address limitations in current MTC methodology by including data from various trial designs.
• To demonstrate the application of these methods in a rheumatoid arthritis case study.

Main Methods:

• Developed three alternative methods for data integration in MTC: naive pooling, prior adjustment using observational data, and a three-level hierarchical model.
• Naive pooling: simple integration without design differentiation.
• Prior adjustment: uses observational data to adjust for trial design bias.
• Three-level hierarchical model: flexible approach allowing bias adjustment and heterogeneity modeling between trial designs.

Main Results:

• The three-level hierarchical model offers the most flexibility, enabling bias adjustment and accounting for heterogeneity across different trial designs.
• Naive pooling is the simplest but does not account for potential biases introduced by varying trial designs.
• Utilizing observational data as prior information provides a method to adjust for bias related to trial design.

Conclusions:

• The proposed methods enhance Bayesian MTC models by enabling the inclusion of data from various trial designs.
• These advanced techniques improve the accuracy and robustness of relative treatment efficacy estimates.
• The application in rheumatoid arthritis demonstrates the practical utility of these methods for healthcare decision-making.