OBJECTIVES: Economic evaluations of competing interventions often estimate mean overall survival (OS) as a measure of intervention effect. New treatments offer some patients the possibility of being “cured” of their disease, in that they become long-term survivors whose risk of death is the same as a disease-free person. Grouping cured and non-cured patients together and reporting one mean value for OS may provide a biased assessment of a therapy that cures a proportion of patients. In this study, we compared standard survival analysis versus an approach that accounts for the fraction of patients cured. METHODS: We used clinical trial data from advanced melanoma patients treated with ipilimumab (n=137) versus gp100 (n=136) and applied statistical methodology for mixture cure models. We used logistic regression to model the probability that a patient was cured and a Weibull regression model to estimate the excess mortality for non-cured patients. Both cured and non-cured patients were subject to background mortality not related to cancer; we calculated this using age- and gender-matched mortality data from US Social Security life tables. RESULTS: Ignoring a cured proportion, ipilimumab had an estimated mean OS that was 8 months longer than gp100. Cure model analysis showed that the proportion of cured patients drove this difference, with 20% cured on ipilimumab compared to 6% with gp100. The mean OS among non-cured patients was 5 months on ipilimumab versus 4 months on gp100.  The mean OS among cured patients was 26 years on both arms. After adjusting for covariates, ipilimumab had an improved cure proportion compared to gp100 (OR=2.01, 95% CI (1.00, 4.06)), but there were no significant differences in survival among non-cured patients (HR=1.05, 95% CI (0.80, 1.38)). CONCLUSIONS: This analysis supports using cure modeling in health economic evaluation in advanced melanoma, since it may reduce bias in OS estimates.