Mapping EORTC QLQ-C30 into EQ-5D-5L index for Korean patients with gastric cancer

Abstract Purpose This study aimed to develop and validate a mapping algorithm to derive EQ-5D-5L utility values from EORTC QLQ-C30 responses in Korean patients with gastric cancer. Methods Secondary data from a multicenter cohort of 399 patients across 24 hospitals in Korea were analyzed, with utility calculated using the Korean EQ-5D-5L valuation set. Five regression approaches (OLS, TPM, ALDVMM, CLAD, and Beta regression) were applied under two specifications: Model 1, without control variables, and Model 2, with age and sex as control variables. Models were evaluated using RMSE and MAE, with 5-fold cross-validation and time point-specific external validation (6, 12, and 24 months). Results Among the 399 patients (mean age, 61.42 ± 9.36 years; male, 71.9%), the mean baseline EQ-5D-5L utility was 0.883 ± 0.115, with a ceiling effect observed in 39.35% of patients. TPM showed the best predictive performance among the five regression approaches in both Model 1 (R²=0.546, RMSE = 0.077, MAE = 0.060) and Model 2 (R²=0.549, RMSE = 0.077, MAE = 0.060), consistent with 5-fold cross-validation results. In time point-specific external validation, however, the model with the lowest MSE varied by time point. Conclusions This study offers a validated mapping algorithm for converting EORTC QLQ-C30 scores into EQ-5D-5L utility values in Korean gastric cancer patients, supporting its application in clinical research and healthcare policymaking.