OBJECTIVES: Rheumatoid arthritis (RA) is a chronic autoimmune disorder associated with substantial health and economic burden. We describe a cost-utility model developed to assess the cost-utility of Abatacept in the treatment of RA in the UK (NHS). METHODS: A probabilistic patient level simulation model was developed to estimate long-term costs and health outcomes of abatacept versus methotrexate (MTX) in RA patients who failed anti-TNF therapy. The model predicted patients' HAQ (Health Assessment Questionnaire) scores over time based on the initial response to treatment (% change in HAQ at six months). Patients with an inadequate response (failure to achieve a specified reduction in HAQ score) ended treatment at this point. Responding patients continued treatment with a reduced rate of HAQ progression until long-term treatment failure – modelled as an exponential process. On long term treatment failure, patients' HAQ was assumed to worsen by amount equal to their initial improvement. Efficacy and adverse events rates were obtained from a randomized clinical trial. Costs, utility and annual mortality rates (AMR) were estimated as a function of HAQ. RESULTS: Compared to MTX, Abatacept treatment results in 1.6 additional QALYs at an additional cost of £40,371, giving an ICER of £25,395/QALY, a value in line with other biologic treatments recommended for use in the UK. Results were stable under a range of sensitivity analyses. CONCLUSION: Compared to MTX, Abatacept is a cost-effective treatment for patients who failed anti-TNF therapy. The use of a patient level simulation allows costs and utilities to be estimated as non-linear functions of HAQ and for the AMR to be conditioned on patient's HAQ scores. The model was implemented in R and was sufficiently fast for probabilistic analysis. This allowed two key requirements for decision-making to be met: unbiased estimates of costs and effects and an assessment of the decision uncertainty.