Comparative validation of prognostic indices in brain metastases receiving cranial radiotherapy

Introduction. This study aimed to externally validate widely used prognostic indices in patients with brain metastases (BM) treated with cranial radiotherapy and to evaluate the combined prognostic impact of clinical variables at the time of BM diagnosis and treatment-related factors on overall survival (OS). Material and methods. A total of 116 patients who received cranial radiotherapy for BM between 2020 and 2024 were retrospectively analyzed. OS was estimated using the Kaplan–Meier method. Univariate and multivariate Cox regression analyses were performed separately for clinical variables at the time of BM diagnosis and treatment-related factors, followed by a combined analysis including all significant predictors. The prognostic performance of four indices — Recursive Partitioning Analysis (RPA), Graded Prognostic Assessment (GPA), Diagnosis-Specific GPA (ds-GPA), and Basic Score For Brain Metastases (BSBM) — was assessed using receiver operating characteristic (ROC) curves and area under the curve (AUC) values. Results. The median OS was 9.0 months, with 6- and 12-month survival rates of 63.8% and 34.5%, respectively. In the final multivariate analysis, higher performance status [Karnofsky Performance Score (KPS) ≥ 70], lung primary, lower extracranial metastasis burden, receipt of chemotherapy, and surgical resection were independent predictors of improved survival, while radiotherapy modality (sterotactic radiosurgery vs. whole brain radiotherapy) showed no significant difference. All four indices effectively stratified OS, with ds-GPA and BSBM demonstrating the highest predictive accuracy (AUC = 0.931 and 0.899). Conclusions. This study confirms the validity of established prognostic indices and highlights that achieving the survival predicted at the time of BM diagnosis largely depends on the effective delivery of systemic and local therapies. These findings underscore the importance of individualized and multidisciplinary management in optimizing real-world outcomes.