OBJECTIVES In oncology, the impact of interventions on health-related quality of life (HRQL) is traditionally modelled based on disease progression status. The aim of this analysis was to assess if more meaningful patterns exist in HRQL data, based on other clinically important events that should be considered in modelling utility. METHODS HRQL data from the CA184-024 trial of ipilimumab plus dacarbazine in previously untreated patients with unresectable malignant melanoma were analysed. European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire Core-30 (EORTC QLQ-C30) responses were mapped to a generic, preference-based measure (EORTC-8D) by means of a published and validated mapping algorithm. The utility observations available for each patient were used to examine the relationships between HRQL and a number of disease- and time-based variables, including treatment effect, progression status and time to death, via a mixed effects regression model.  RESULTS Progression status was found not to be significantly predictive of utility (p=0.29). Of the variables considered, the strongest relationship was with time to death, the mixed effects model for which was significantly predictive of utility (p≤0.001). HRQL dropped as patients approached death; patients treated with ipilimumab had a utility of 0.86 if time to death was more than 1 year, which reduced to 0.61 during the final month of life. The ipilimumab treatment variable was associated with a small negative coefficient (-0.02), accounting for the adverse event profile of the drug when added to dacarbazine (p=0.06).  CONCLUSIONS Analysis of the CA184-024 HRQL data showed that time to death rather than progression status was significantly predictive of utility. Hence, modellers should carefully examine primary data to determine if a time to event approach or a progression based approach is appropriate to model utility best reflecting the pathology of the disease.