Cost utility of tumour necrosis factor-α inhibitors for rheumatoid arthritis: an application of Bayesian methods for evidence synthesis in a Markov model.

BACKGROUND: Rheumatoid arthritis (RA) is a chronic autoimmune disease that affects approximately 1.5 million people in the US. Tumour necrosis factor (TNF)-α inhibitors have been shown to effectively treat and maintain remission in patients with moderately to severely active RA compared with conventional agents. The high acquisition cost of TNF-α inhibitors prohibits access, which mandates economic investigations into their affordability. The lack of head-to-head comparisons between these agents makes it difficult to determine which agent is the most cost effective.
OBJECTIVE: This study aimed to determine which TNF-α inhibitor was the most cost-effective agent for the treatment of moderately to severely active RA from the US healthcare payer's perspective.
METHODS: A Markov model was constructed to analyse the cost utility of five TNF-α inhibitors (in combination with methotrexate [+MTX]) versus MTX monotherapy using Bayesian methods for evidence synthesis. The model had a cycle length of 3 months and an overall time horizon of 5 years. Transition probabilities and utility scores were based on published studies. Total direct costs were adjusted to year 2009 $US using the medical component of the Consumer Price Index. All costs and QALYs were discounted at a rate of 3% per year. Patient response to the different strategies was determined by the American College of Rheumatology (ACR)50 criteria. One-way and probabilistic sensitivity analyses (PSAs) were performed to test the robustness of the base-case scenario. The base-case scenario was changed to ACR20 criteria (scenario 1) and ACR70 criteria (scenario 2) to determine the model's robustness. Cost-effectiveness acceptability curves and cost-effectiveness frontiers were used to estimate the cost-effectiveness probability of each treatment strategy. A willingness-to-pay (WTP) threshold was defined as three times the US GDP per capita ($US139,143 per additional QALY gained). Primary results were presented as incremental cost-effective ratios (ICERs).
RESULTS: Etanercept+MTX was the most cost-effective treatment strategy in the base-case scenario up to a WTP threshold of $US2 185,497 per QALY gained. At a WTP threshold of greater than $US2 185,497 per QALY gained, certolizumab+MTX was the most cost-effective treatment strategy. One-way analyses showed that the base-case scenario was sensitive to the probability of achieving ACR50 criteria for MTX and each TNF-α inhibitor, and changes in the utility score for patients who achieved the ACR50 criteria. With the exception of infliximab, all of the TNF-α inhibitors were sensitive to drug cost per cycle. In the scenario analyses, certolizumab+MTX was a dominant treatment strategy using ACR20 criteria, but etanercept+MTX was a dominant treatment strategy using ACR70 criteria.
CONCLUSIONS: Etanercept+MTX was a cost-effective treatment strategy in the base-case scenario; however, the model was sensitive to parameter uncertainties and ACR response criteria. Although Bayesian methods were used to determine transition probabilities, future studies will need to focus on head-to-head comparisons of multiple TNF-α inhibitors to provide valid comparisons.