OBJECTIVE: To identify the costs and utility of symptomatic treatment using a shared model of care (SC) and aggressive treatment using a hospital model of care (HT) for established stable rheumatoid arthritis (duration > 5 years) and estimate the relative cost acceptability of shared care. METHODS Data was obtained from 466 adults with established RA, who were enrolled into a randomised controlled trial of SC and HT and followed for 36-months. A societal perspective was used; costs and outcomes were discounted at 3.5%, the recommended UK Treasury rate. The primary outcome was quality adjusted life years (QALY's) measured by the EQ-5D questionnaire and population utility tariffs. Direct costs were measured as resource use multiplied by published national unit costs. An incremental cost utility ratio (ICER) was estimated. Bootstrapping techniques were used to derive net benefit statistics and cost acceptability curves and determine the probability that SC was cost effective compared to HT. Sensitivity analysis was used to assess the impact of alternative discount rates, alternative methods of imputing missing and censored data and alternative sources of unit costs. RESULTS: SC is associated with higher QALY's and higher costs than HT, giving a net cost of £1517/QALY gained. The probability that SC was cost-effective was 0.89, over a £0 to £50,000 range of cost/QALY threshold values. The probability that SC was cost-effective ranged from 0.5 if decision makers are prepared to pay £2000/QALY gained, to 0.8 for a threshold of £13,000/QALY gained, to 0.9 for a threshold of £42000/QALY gained. Sensitivity analysis showed that the probability SC was cost effective was reduced if protocol driven visits were included (p = 0.84) or censored data was not imputed (p = 0.70). CONCLUSION: The primary and sensitivity analyses indicate that shared care is likely to be more cost-effective in the UK than hospital treatment in 70%-90% of cases.