AUTHOR=Zheng Miyun , Xu Maodong , You Mengxing , Huang Zhiqing 

TITLE=Development of a prognostic nomogram for ocular melanoma: a SEER population-based study (2000–2021)

JOURNAL=Frontiers in Medicine

VOLUME=Volume 12 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2025.1494925

DOI=10.3389/fmed.2025.1494925

ISSN=2296-858X

ABSTRACT=IntroductionOcular melanoma (OM) is a rare but lethal subtype of melanoma. This study develops a prognostic nomogram for OM using machine learning and internal validation techniques, aiming to improve prognosis prediction and clinical decision-making.MethodsIndependent prognostic variables were identified using univariate and multivariate COX proportional hazard regression models. Significant variables were then incorporated into the nomogram. The predictive accuracy of the nomogram was evaluated using receiver operating characteristic (ROC) curves, calibration plots, decision curve analysis (DCA), and 10-fold cross-validation. The performance of the nomogram was compared with that of a machine learning model.ResultsThirteen variables, including age, sex, tumor site, histologic subtype, stage, basal diameter size, tumor thickness, liver metastasis, first malignant primary indicator, marital status, and treatment modalities (surgery/radiotherapy/chemotherapy) were identified as independent prognostic factors for overall survival (OS) and were included in the nomogram (all P < 0.05). The nomogram showed a concordance index of 0.712. The areas under the curve (AUC) for predicting 3-, 5-, and 10-year survival rates were 0.749, 0.734, and 0.730, respectively. Calibration plots for 3-, 5-, and 10-year survival were in close agreement with the ideal predictions, and DCA indicated a superior net benefit. The average AUC from 10-fold cross-validation was 0.725. The machine-learning model identified liver metastasis as the most significant predictor of survival, followed by age, radiotherapy, stage, and other factors that were incorporated into the nomogram. The machine-learning model achieved a predictive AUC score of 0.750.ConclusionsA robust nomogram incorporating 13 significant clinicopathological variables was developed. The combined use of ROC curve analysis, calibration plots, DCA, 10-fold cross-validation, and machine learning confirmed the strong predictive performance of the nomogram for survival outcomes in patients with OM.