Machine learning based prognostic model for oral squamous cell carcinoma using SEER data and external validation

Oral squamous cell carcinoma (OSCC) is characterized by an insidious onset, pronounced aggressiveness, and a substantial impact on patient survival. Current prognostication relies heavily on the TNM staging system, which often lacks precision. To address this, we developed and validated a machine learning (ML)-based prognostic nomogram. Analyzing 8,927 patients from the SEER database, we employed ML algorithms (LASSO, XGBoost, random forest (RF), support vector machine (SVM)) to identify key determinants, including age, race, sex, grade, and TNM stage, and constructed a Cox-based risk model. Crucially, benchmarking analysis demonstrated that our model achieved a C-index of 0.714, significantly outperforming the traditional TNM staging system (C-index: 0.665, p < 0.001). These findings indicate that integrating demographic and pathological features with ML techniques yields superior risk stratification compared to anatomical staging alone. This nomogram offers a precise tool for personalized survival assessment and clinical decision-making.