Related Experiment Video
Updated: Mar 24, 2026
Competing-Risk Nomogram for Predicting Cancer-Specific Survival in Multiple Primary Colorectal Cancer Patients after Surgery
Published on: September 27, 2024
Development and internal validation of a Cox-based nomogram for predicting cancer-specific survival in central
Zhiling Tan1, Liyi Deng2, Youzhong Ye2
1Department of Neurosurgery, The First Affiliated Hospital of Jinan University, Guangzhou, 510632, Guangdong, China; Department of Neurosurgery, The First People's Hospital of Chenzhou, 423000, Hunan, China.
Background:
Central nervous system (CNS) ependymoma is a heterogeneous tumor with highly variable clinical behavior. Most existing prognostic models primarily focus on overall survival, whereas robust tools specifically predicting cancer-specific survival (CSS) in CNS ependymoma are lacking. We aimed to develop a Surveillance, Epidemiology, and End Results (SEER)-based nomogram for individualized prediction of CSS in patients with CNS ependymoma.
Methods:
We identified patients with primary CNS ependymoma (2010-2022) from the Surveillance, Epidemiology, and End Results (SEER) database. The cohort was randomly split in a 7:3 ratio into training and validation cohorts; patients were assigned using a computer-generated random number sequence. Multivariable Cox proportional hazards regression was used to identify independent prognostic factors. A nomogram was constructed to predict 3-, 5-, and 8-year CSS. Model performance was evaluated using the concordance index (C-index), time-dependent receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA), and was compared with a SEER summary stage-only model.
Results:
The study included 1744 patients with CNS ependymoma. During a median follow-up of 63 months, the 3-, 5-, and 8-year CSS rates were 92.4%, 90.5%, and 86.7%, respectively. Multivariable analysis identified age, sex, histologic subtype, primary tumor site, and SEER summary stage as independent predictors of CSS. The nomogram demonstrated superior discrimination, with C-indices of 0.818 in the training cohort and 0.827 in the validation cohort, significantly outperforming the stage-only model (0.588 and 0.547, respectively). Similarly, the nomogram yielded higher areas under the curve (AUCs) at all time points (e.g., 5-year AUC: 0.850 vs. 0.600 in the training cohort). Calibration plots showed excellent agreement between predicted and observed CSS, and DCA indicated a higher net clinical benefit for the nomogram than for the stage-only model across a broad range of threshold probabilities.
Conclusions:
We developed and internally validated a robust, CSS-oriented nomogram for CNS ependymoma using a contemporary, population-based cohort. This tool outperformed conventional staging systems and may provide a practical approach for individualized risk stratification and clinical decision-making in patients with CNS ependymoma.
