PCN295 Health State Utility Values By Time to Death in Recurrent or Metastatic Squamous-Cell Carcinoma of the Head and Neck

Abstract

Cost-utility analysis is the preferred form of economic evaluation for the majority of reimbursement agencies. This analysis aimed to estimate utilities based on time-to-death (TTD) periods associated with recurrent or metastatic squamous-cell carcinoma of the head and neck (SCCHN). Utilities were generated using patient-level data of the CheckMate 141 trial (NCT02105636; December 2015 data cut) of nivolumab versus investigator’s choice using the three-level EQ-5D (EQ-5D-3L) and UK value set. TTD periods were defined in line with prior approaches to TTD-based utility values and modified based on data availability (Model 1: ≥183 days, 92-182 days, 0-91 days; Model 2: 57-91 days, 29-56 days, 0-28 days). Utility values were estimated using linear mixed models with a fixed effect of TTD period and random intercept for each patient. Treatment-independent and treatment-specific (incorporating fixed effects of treatment and its interaction with TTD period) utilities were estimated. 502 observations from 273 patients could be assigned to a TTD period. Estimated treatment-independent utilities were: ≥183 days (n=135, 0.694 [95% CI 0.652, 0.736]), 92-182 days (n=81, 0.651 [95% CI 0.594, 0.707]), 0-91 days (n=118, 0.510 [95% CI 0.461, 0.560]). Splitting the final 3 months prior to death yielded the following utilities: 57-91 days (n=63, 0.569 [95% CI 0.484, 0.655]), 29-56 days (n=40, 0.487 [95% CI 0.378, 0.596]), 0-28 days (n=33, 0.422 [95% CI 0.302, 0.541]). Treatment-specific utilities showed a similar trend of decreasing utilities closer to death in a near-linear fashion; however, sample sizes became small for the final TTD periods. This study estimated TTD-based utility values in SCCHN. Utility values decreased in a monotonic trend in accordance with closer TTD, especially in the final 3 months prior to death. These values can support cost-utility analyses as a supplementary alternative to the more commonly applied progression-based models.