Tumor size and treatment factors as correlates of 10-year mortality in grade III spinal ependymomas: a nationwide analysis

Grade III spinal ependymomas are rare central nervous system tumors with poor prognoses and limited treatment guidance. This study aimed to identify predictors of long-term mortality using a national dataset. The National Cancer Database was queried for patients diagnosed with grade III spinal ependymomas from 2004 to 2017. Demographic, clinical, and treatment variables were analyzed with traditional statistics and machine learning models. Model performance was evaluated with C-index and Integrated Brier Score (IBS), and SHAP analysis was used for feature interpretation. A total of 2,011 patients were included. Surgical resection was performed in 83% of cases, with gross total resection in 96% of those. Radiation was given to 33% (mean 41.4 Gy), and chemotherapy to 18%. Ten-year survival was 75%. Long-term survivors were younger (42 vs. 52 years, P < 0.001), had fewer comorbidities (P < 0.001), and were more often privately insured. Mortality was higher among those receiving radiation (56% vs. 24%, P < 0.001) or chemotherapy (37% vs. 12%, P < 0.001). Random Survival Forest was the best-performing model (C-index 0.724; IBS 0.176), identifying tumor size, radiation, chemotherapy, and surgery as key correlates of mortality. Kaplan-Meier analysis showed reduced survival with radiation and chemotherapy, but improved survival with surgery. Stratified analyses revealed diminished surgical benefit in pediatric patients and small tumors. Tumor size emerged as the strongest correlate of mortality. Surgery improves survival, though benefit varies by age and tumor features. Chemotherapy and radiation correlated with higher mortality, underscoring the need for individualized treatment planning and prospective validation.