OBJECTIVES: In 2020, World Cancer Research Fund International reported that endometrial cancer is the sixth most common cancer in women worldwide with over 400,000 new cases per year. Health state utilities play a crucial role in assessing the quality of life and treatment outcomes in patients in clinical trials, and are essential components of cost effectiveness models and budget impact models in health technology assessment. Historically, oncology utility analyses have centered around progression status, but recently time-to-death (TTD) utility analysis have gained importance. The objective of this analysis is to compare the health-state utility prediction models based on progression status and/or TTD using EQ-5D-5L data from endometrial cancer patients enrolled in the dostarlimab monotherapy GARNET trial (Clinicaltrials.gov identifier: NCT02715284).

METHODS: Data on EQ-5D-5L responses at baseline, every 3 to 6 weeks during treatment, at end-of-treatment visit, safety follow-up, and every 90 days during the post-treatment follow-up period were utilised. Utilities were derived based on the Netherlands reference value set. Using the Generalized Estimating Equation (GEE), the utilities were predicted using different combination of predictors. Models were compared using quasi-likelihood information criteria (QIC), generalized R-square, mean absolute error (MAE) and root mean square error (RMSE). Statistical significance of the predictor variables was also assessed.

RESULTS: Patients experienced reduction in utilities post progression and at time closer to death. Compared to progression status alone, the model with TTD showed a lower QIC. The model containing both progression and TTD indicated statistical significance of their interaction. Addition of baseline covariates improved prediction as confirmed by higher R-square and lower MAE and RMSE.

CONCLUSIONS: Using data from endometrial cancer patients, different models were fitted to predict health utilities. The model including progression status, TTD, their interaction term, and baseline covariates was found to be the better fit and may improve cost effectiveness evaluation.