OBJECTIVES: The aim of this analysis was to assess predictors of health-related quality of life over time, and to estimate utilities for patients with treatment-naïve advanced melanoma in the randomized CheckMate 066 trial comparing nivolumab with dacarbazine for use in a cost-effectiveness model (CEM). METHODS: The EQ-5D was administered at baseline and every 6 weeks in CheckMate 066 and was used to generate index utility scores using the UK time trade-off (TTO) method. Covariates were based on a combination of prior analyses from large trial datasets, including patient demographic and clinical characteristics, quantitative metrics of fit, qualitative/clinical plausibility, and relevance to the CEM. Several longitudinal mixed linear models were explored using different covariate sets. RESULTS: This analysis included 288 patients and 1,125 visits where the EQ-5D was administered. Mean baseline utility score was 0.75 for nivolumab, 0.69 for dacarbazine, and 0.72 across both treatment arms. The final model included baseline utility (to adjust for imbalance between treatment arms), progression status (pre/post), days until death or end of follow-up (<30 days/30+ days), and treatment arm. Parameter estimates in the model were 0.603 for baseline utility (P<0.001), −0.074 for progression status (P<0.001), −0.022 for <30 days left (P=0.092), and −0.069 for treatment arm (dacarbazine vs nivolumab; P=0.008). When implemented in the CEM, the utility estimate for the pre- and post-progression states were 0.802 and 0.728, respectively (applying nivolumab as the treatment arm). A decrement for the month preceding death was applied using the estimate for <30 days until death or end of follow-up (−0.022). CONCLUSIONS: These results showed that baseline utility, progression, <30 days until death or end of follow-up, and treatment arm are all predictors of utility over time, which is consistent with prior work in melanoma. As data mature, these analyses will be replicated in this and other nivolumab trials.