Interpreting EORTC CAT Core short-form scores in metastatic breast cancer: EQ-VAS-associated cut-points and comparison with published thresholds for clinical importance

Abstract Background Published thresholds for clinical importance (TCIs) for the European Organisation for Research and Treatment of Cancer CAT Core were derived using domain-specific relevance anchors, which are not routinely collected. We estimated EQ-VAS-associated cut-points for EORTC CAT Core short-form scores in patients with metastatic breast cancer, compared them with published TCIs, and examined patient-reported Eastern Cooperative Oncology Group (ECOG) performance status as a secondary reference classification. Methods This post hoc cross-sectional analysis used data from the PRO B trial. EQ-VAS ≤ 70 was used to define lower global health status, with EQ-VAS ≤ 65 examined in sensitivity analysis; patient-reported ECOG ≥ 2 defined impaired performance status in secondary analysis. Receiver operating characteristic (ROC) analyses based on continuous CAT Core T-scores identified Youden-optimal cut-points with 5,000 bootstrap replications for estimating confidence intervals and optimism-corrected performance. Results Among 855 patients with CAT Core data available for the baseline analysis (mean age 51 years, SD 11), 841 had EQ-VAS and 854 had ECOG data. Using EQ-VAS ≤ 70, AUCs ranged from 0.57 to 0.82; eight of 14 domains had AUC ≥ 0.70, five had AUC 0.60–0.69, and diarrhea had AUC 0.57. Four domains did not meet the criterion used for interpretation of |ρ| ≥0.30 and were therefore interpreted as exploratory. Published TCIs showed different sensitivity-specificity profiles when evaluating against the EQ-VAS reference classification, illustrating the dependence of cut-point performance on the reference construct. Conclusions This study identified EQ-VAS-associated cut-points for EORTC CAT Core short-form scores in metastatic breast cancer. Their performance varied by domain, and they should be interpreted as context-specific reference values rather than replacements for published relevance-based TCIs.