Development and internal validation of dual overall and cancer-specific survival nomograms for primary spinal and sacropelvic Ewing sarcoma: a population-based study of 346 patients.

Background: Primary spinal Ewing sarcoma (ES) is exceptionally rare, and existing prognostic evidence is largely derived from mixed-site cohorts that fail to capture the spine-specific clinical and biological profile. Accurate, site-specific prognostication remains unavailable, limiting consistency in clinical decision-making. This study aimed to define independent prognostic determinants for spinal ES and to develop and internally validate dedicated nomograms and a risk-stratification system for both overall survival (OS) and cancer-specific survival (CSS). Methods: Patients diagnosed with primary spinal ES between 1995 and 2015 were identified from the Surveillance, Epidemiology, and End Results (SEER) database. Prognostic variables for OS and CSS were assessed using Cox proportional hazards regression with backward stepwise selection optimized by the Akaike information criterion (AIC). Independent predictors were integrated into separate OS and CSS nomograms. Discrimination, calibration, and clinical utility were evaluated using Harrell's concordance index (C-index), time-dependent area under the receiver operating characteristic curve (AUC), calibration curves with bootstrap resampling, and decision curve analysis (DCA). A three-tier risk classification system was constructed based on total nomogram scores. Results: A total of 346 eligible patients were included and randomly allocated to a training cohort (n=262) and a validation cohort (n=84) in a 3:1 ratio; 158 deaths from any cause and 142 ES-specific deaths were observed. Chemotherapy omission, distant disease and Black race/ethnicity were independent predictors of worse OS and CSS, larger tumor size was an independent predictor of worse OS, and married status was associated with lower mortality risk; the racial association, based on 10 Black patients and none in validation, is hypothesis-generating. The OS and CSS nomograms showed moderate discrimination in the training [OS C-index 0.726, 95% confidence interval (CI): 0.696-0.756; CSS 0.727, 0.697-0.757] and validation cohorts (OS 0.699, 0.640-0.758; CSS 0.680, 0.622-0.738); after 1,000-resample bootstrap internal validation the optimism-corrected C-index was 0.695 (OS) and 0.702 (CSS), with calibration slopes of 0.917 and 0.938, respectively. Calibration curves showed close agreement between predicted and observed survival. The three-tier risk classification produced well-separated survival curves for both OS and CSS, with median CSS of 197 months, 37 months, and 13 months in the low-, intermediate-, and high-risk groups, respectively. Conclusions: This study provides the first dual-endpoint, site-specific prognostic framework for primary spinal and sacropelvic ES. Discrimination was moderate and validation internal only; the nomograms should be regarded as hypothesis-generating, and external validation in independent contemporary cohorts is required before clinical use.