Development and validation of a nomogram for predicting cancer-specific survival in pediatric melanoma
10.3760/cma.j.cn.115807-20210324-00096
- VernacularTitle:儿童恶性黑色素瘤患者特异性生存率的预测模型建立及其效果评价
- Author:
Rujing PAN
1
;
Xuan XUAN
;
Mingfen LYU
;
Jingjing LIU
Author Information
1. 温州医科大学附属第一医院皮肤科 325000
- Keywords:
Cancer-specific survival;
Nomogram;
Pediatric melanoma
- From:
Chinese Journal of Endocrine Surgery
2021;15(6):572-577
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To identify prognosis factors and construct a nomogram for the prediction of cancer-specific survival in surgically resected pediatric melanoma.Methods:A total of 912 patients aged 0-19 years were extracted from the Surveillance, Epidemiology, and SEER database and randomly divided into training cohort (n=640) and validation cohort (n=272) (A ratio of 70:30) . Univariable and multivariable cox analysis were used to determine prognostic factors, and these factors were used to construct a nomogram for predicting cancer-specific survival in patients with resected pediatric melanoma. Model performance was evaluated by Harrell’s concordance index (C-index) ,the area under the time-dependent receiver operating characteristic curve (AUC) and calibration plots.Results:As revealed by the multivariable cox analysis, age, tumor location, ulceration, and lymph node status were all associated with melanoma-specific survival in pediatric patients. On the basis of these factors, a nomogram was constructed. Both the training cohort and the validation cohort had a concordance index of 0.9, which validated the accuracy of our nomogram. The nomogram showed significant discriminative power in both training cohort (3-year AUC: 0.87, 5-year AUC: 0.88, 10-year AUC: 0.85) and validation cohort (3-year AUC: 0.87, 5-year AUC: 0.87, 10-year AUC: 0.89) . Also, the nomogram displayed a good calibration.Conclusions:These results suggest that the new model has superior predictive performance in predicting cancer-specific survival of pediatric melanoma. This individualized prediction model can provide reference for tailoring treatment and clinical counseling of pediatric melanoma.
