A Bayesian model that jointly considers comparative effectiveness research and patients' preferences may help inform GRADE recommendations: an application to rheumatoid arthritis treatment recommendations

Objectives The objective of the study was to estimate the preferred treatment for early rheumatoid arthritis using a novel Bayesian approach that jointly considers patients’ preferences and comparative effectiveness research. Study Design and Setting We estimated the preferred treatment using patients' preferences measured in a discrete-choice experiment to apply weights to benefit and harm outcomes from a network meta-analysis and other considerations (dosing, rare adverse events). Using Bayesian analyses, we considered the variability in patients' preferences and the imprecision in both patients’ preferences and the treatment effects; all key considerations in the Grading of Recommendations Assessment, Development, and Evaluation approach. Results We estimated that most patients in our population would prefer triple therapy as initial treatment (78%) or after an inadequate response to methotrexate (62%). The probability of choosing triple therapy as initial treatment was further from 50% (the point of indifference) for more patients, making our prediction more confident, and suggesting a stronger recommendation could be made. After an inadequate response to methotrexate, the choice was more split, suggesting a decision aid may be helpful. Conclusion Using a novel approach, we estimated that many patients with early rheumatoid arthritis may prefer triple therapy to other treatment options, in contrast to existing guidelines. This offers an approach that may help inform Grading of Recommendations Assessment, Development, and Evaluation treatment recommendations.