OBJECTIVES: In analysis of health state utilities derived from a nivolumab phase III clinical trial a statistically significant difference in baseline utilities was noted between nivolumab (NIVO) and dacarbazine (DTIC) treatment arms.  The aim of this study was to analyse and assess possible drivers of this difference in mean baseline utility. METHODS: Patient demographics and baseline disease characteristics were compared between treatment arms to identify any variables which may be responsible for the observed differences in baseline utilities.  All identified variables were included in a generalised linear model and tested for significance differences in a stepwise manner to determine the final predictive utility model. RESULTS: Population demographics and baseline disease characteristics were generally well balanced between treatment arms, however patients randomised to NIVO were generally of better ECOG performance status (70.5% and 28.6% for PS 0 and 1 respectively) than those randomised to DTIC (58.2% and 40.4%).  A generalised linear regression model of utility therefore included ECOG performance status in addition standard variables age, sex, weight, smoking history and metastatic stage.  After stepwise elimination of non-significant variables, ECOG performance status was found to be a strongly significant predictor of health state utility (p<0.000) as was sex (p<0.000) and age (p=0.006).  Whilst metastatic stage remained in the final model specification as these were pre-defined stratification factors, neither was a significant predictor of health state utility (p=0.724, p=0.493 respectively). CONCLUSIONS: ECOG performance status was found to be an independent predictor of health state utility in addition to age and sex.  As such it is critical that any economic evaluation of an intervention in this patient population identifies any differences in performance status between intervention and comparator arms and seeks to establish the impact any difference may have on subsequent results.