A Bayesian approach to evaluating net clinical benefit allowed for parameter uncertainty

Although randomized controlled trials (RCTs) are conducted to establish whether novel interventions work on average in the patient population, there is a growing desire to move to a more individualized approach to evaluation. The potential benefits and harms of a treatment policy may differ between individuals. If these benefits and harms are not evaluated distinctly, and in a quantitative framework, transparency can be lost in the decision-making process.

Glasziou and Irwig have outlined the concept of net clinical treatment benefit for identifying the patients for whom the potential benefits of treatment outweigh the possible side effects. This study revisits the decision whether to use warfarin to treat atrial fibrillation. In this analysis, RCT and various sorts of observational data are synthesized.

This reanalysis brings into question the conclusions of the original analysis on who would benefit from warfarin; however, caution is advised, due to limitations in the quality of life data available.

A fully realized Bayesian implementation of the model is presented. This provides a framework for including uncertainty related to the estimation of all model parameters, and permits both direct probability statements and credible intervals for specific patient groups to be expressed.