BACKGROUND & AIMS: Patients with steroid-refractory ulcerative colitis face a difficult treatment decision between colectomy and therapy with infliximab or cyclosporine. The aim of this study was to understand how individual patient preferences for the various treatment outcomes influence the optimal treatment decision for a given patient. METHODS: A Markov model was used to simulate treatment with total colectomy with an ileo pouch-anal anastomosis (TC/IPAA), cyclosporine (CSA), infliximab (INFLX), and infliximab followed by cyclosporine for treatment failures (INFLX-->CSA). Utility weights for treatment outcomes were elicited from 48 patients using both time trade-off and visual rating scale methods. Preference sets were applied to the model to identify the therapy that maximized quality-adjusted life years (QALYs) for each patient. Sensitivity analyses were performed to assess model robustness. RESULTS: Optimal treatment was highly variable among patients (INFLX-->CSA = 42%, 20/48; TC/IPAA = 37%, 18/48; CSA = 21%, 10/48; INFLX = 0%, 0/48). However, when average preference weights from our sample were applied to the model, medical treatments were superior to TC (CSA = .26 QALYs gained vs TC/IPAA; INFLX-->CSA = .25 QALYs gained vs TC/IPAA). CONCLUSIONS: Patient preferences have a clear impact on the optimal treatment for steroid-refractory ulcerative colitis. Although averaged preferences support the use of medical interventions, a third of individual patients may benefit most from proceeding directly to colectomy. Failure to fully assess individual preferences may result in suboptimal treatment for these patients.