Development and internal validation of a prognostic nomogram for cancer-specific survival in Ewing sarcoma: a SEER-based study
1Shandong Medical and Pharmaceutical University, Binzhou, China.
Background:
Ewing sarcoma is a rare and aggressive malignancy with substantial survival heterogeneity. Conventional tumor-node-metastasis (TNM) staging provides limited individualized prognostic assessment, and previous Surveillance, Epidemiology, and End Results (SEER)-based nomograms have rarely reported optimism-corrected performance or clinical utility. This study aimed to develop and internally validate a prognostic nomogram for cancer-specific survival (CSS) in patients with Ewing sarcoma.
Methods:
Using the SEER database, 720 patients with pathologically confirmed Ewing sarcoma diagnosed between 2004 and 2015 were identified. Candidate predictors were selected using least absolute shrinkage and selection operator (LASSO) regression and incorporated into a multivariable Cox proportional hazards model. Model performance was assessed using Harrell's C-index, time-dependent area under the curve (AUC), calibration, and decision curve analysis. Internal validation was performed with 1,000 bootstrap resamples to estimate optimism-corrected performance. Multiple imputation by chained equations was used for sensitivity analysis.
Results:
Among 720 patients, 258 CSS events occurred during a mean follow-up of 89.3 months. The final model included age, sex, race, primary tumor site, T stage, N stage, M stage, surgery, radiation, and chemotherapy. The apparent C-index was 0.736, with a bootstrap-corrected C-index of 0.721. The nomogram demonstrated good discrimination at 3, 5, and 10 years (AUCs: 0.795, 0.784, and 0.780, respectively) and acceptable calibration. Compared with TNM staging alone, the nomogram showed improved discrimination (C-index 0.680 vs. 0.736, P<0.001) and greater clinical net benefit. Sensitivity analysis using multiple imputation showed generally consistent predictor effects.
Conclusions:
We developed and internally validated a nomogram for individualized CSS prediction in Ewing sarcoma. The model showed moderate discrimination and improved performance compared with TNM staging. However, external validation is required before clinical application.
