OBJECTIVES: To estimate utility values by health states for regorafenib and placebo-treated subjects from the phase-3 GIST – Regorafenib In Progressive Disease (GRID) trial, and test the assumption that utility values remained constant over successive cycles of treatment in the same health state. METHODS: The GRID study included a double-blinded phase, plus an open-label regorafenib phase for those whose disease progressed.  The EQ-5D index was evaluated using paired-samples comparison as the primary analysis, and repeated measures as a sensitivity analyses. RESULTS: 185 subjects were included; 63% males, with an overall average age of 58 years.  55% received study treatment as 3rd-line, the rest as 4th-line or later.  67% were randomized to receive regorafenib as initial double-blind therapy. Average utility at baseline was 0.767 units. There were no differences in baseline characteristics or EQ-5D for either treatment arm, or those whose disease progressed. The paired-samples analysis compared progression-free EQ-5D index versus any first, post-progression assessment. Of those with available data (N=77) there was a difference of -0.120 units (p=0.001). In the repeated analysis, the Δ-EQ-5D between progression-free disease and disease progression (in double-blind phase) was -0.041 units (p=0.051).  The mean EQ-5D index following discontinuation of open-label treatment due to secondary progression was much lower, with a difference of -0.231 units (p<0.001). Whilst adjusting for disease status and treatment, the cycle number did not significantly influence the EQ-5D index (p=0.341). CONCLUSIONS: Heath-related utility remained stable over successive treatment cycles after controlling for disease status and treatment type, suggesting that for subjects treated with regorafenib who remained progression-free, that active treatment did not lead to deterioration in utility. Due to the cross-over design, the repeated measures analysis did not contain a homogenous, group of people whose disease had progressed. Thus, the paired-sample analysis provides a better estimate of utility.