扁桃体鳞状细胞癌患者生存的社会经济与临床预测因素

Background: Tonsillar cancer represents a distinct and understudied subset of head and neck malignancies with rising incidence, largely driven by human papillomavirus (HPV), associated oropharyngeal cancer. Understanding the roles of clinical characteristics and socioeconomic factors is critical for optimizing early detection and improving outcomes in this population. This study aimed to identify key demographic, clinical, and socioeconomic predictors of treatment delay and overall survival (OS) among patients with tonsillar cancer. Methods: We conducted a population-based retrospective cohort study using the Surveillance, Epidemiology, and End Results (SEER) database (2010-2021), including individuals aged 20-85+ years with primary tonsil cancer. Data were extracted using SEER*Stat version. The primary endpoint was OS. Multivariable Cox proportional hazards regression and machine learning models enhanced with Shapley additive explanations (SHAP) were used to assess prognostic factors and model interpretability. Model performance was compared with random survival forest (RSF) analyses. Results: Among 6,148 patients with Stage I-IV tonsillar squamous cell carcinoma, 1,457 (23.7%) deaths occurred during follow-up (median overall survival [OS], 46 months; interquartile range [IQR], 18-85 months). Nearly half (47.5%) presented with stage IV disease, and most were male (84.3%). Regarding treatment, 2,574 (42%) received chemoradiation alone, while 1,381 (22.5%) underwent combined chemoradiation and surgery. The cohort was predominantly non-Hispanic (NH) White (81%), followed by Hispanic (7.1%), NH Black (6.6%), and Other/Unknown (5.3%). In adjusted Cox models, poorer survival was independently associated with stage IV (HR, 2.21; 95% CI, 1.72-2.47), tumor size >40 mm (HR, 2.09; 95% CI, 1.73-2.54), older age (HR, 2.71; 95% CI, 2.22-3.31), distant metastases (HR, 3.47; 95% CI, 2.78-4.32), rural residence (HR, 1.28; 95% CI, 1.10-1.50), low household income (HR, 1.17; 95% CI, 1.04-1.33), and NH Black race (HR, 1.57; 95% CI, 1.33-1.88). The random survival forest demonstrated good discrimination (C-index 0.715), comparable to that of the Cox model (0.759). SHAP analysis identified metastasis, stage, nodal involvement, income, rurality, and race/ethnicity as key predictors. Conclusion: In this large, population-based study of squamous cell carcinoma of the tonsil, nearly half of the patients presented with advanced disease, and socioeconomic disadvantage was strongly associated with poorer survival. Machine learning analysis identified metastasis, disease stage, nodal involvement, and social factors as key predictors, suggesting that integrating clinical and social context into prognostic models may guide targeted early detection and precision survivorship strategies to reduce disparities.