To establish a clinical prediction model for specific and overall survival in advanced salivary gland malignant tumors, providing a potential reference tool for personalized clinical care and adjunct treatment decision-making.
Retrospective data from the Surveillance, Epidemiology, and End Results (SEER) 8.4.3 version were utilized. Clinical variables collected included patient age, gender, race, diagnosis year, SERR historic stage A, histological types(The histological classification of advanced salivary gland malignant tumors in this study mainly include acinic cell carcinoma, adenoid cystic, mucoepidermoid carcinoma, salivary duct carcinoma, carcinoma ex pleomorphic adenoma, squamous cell carcinoma NOS, salivary gland carcinoma NOS and other histological types), T stage, N stage, M stage, treatment modalities (including surgery, radiotherapy, and chemotherapy), marital status, cancer-specific survival (CSS), overall survival (OS), and survival status. Patients diagnosed with TNM III/TNM IV stage salivary gland malignant tumors between 2010 and 2015 were allocated to the training set, while those diagnosed between 2004 and 2009 served as the test set. Chi-square test compared clinical variables between the training and test sets. Univariate and multivariate COX regression models identified prognostic factors, Kaplan-Meier analysis depicted survival curves, and predictive models for patient OS and CSS were constructed using R-studio environment, with calibration curves and C-index calculated. All statistical analyses were conducted using SPSS 25.0 and R-studio, with P < 0.05 considered significant.
A total of 1477 late-stage salivary gland malignant tumor patients were included. Independent prognostic factors for OS included age, tumor histology, histologic types, T stage, M stage, surgery, and radiotherapy. CSS shared similar prognostic factors with OS. Predictive nomograms based on clinical variables showed high accuracy for 1, 2, 3, 5, and 10-year OS and CSS, with C-indexes of 0.748 and 0.783, respectively. External validation confirmed the models’ accuracy, with well-fitted calibration curves between predicted and observed survival rates.
Nomograms constructed from clinical data can effectively predict OS and CSS for advanced salivary gland malignant tumor patients, providing a potential reference tool for personalized clinical care and adjunct treatment strategies.

© 2025. The Author(s).