Incorporation of stromal tumor-infiltrating lymphocytes into an early death prediction model significantly improves net reclassification for outcome estimation in advanced buccal cancer

Abstract

We explored the impact of stromal tumor-infiltrating lymphocytes (sTILs) on the prognostic value of an early death model for advanced buccal cancer. We assessed 121 patients with advanced buccal cancer who underwent primary tumor resection at a medical center. Predictors of early death and 5-year overall survival (OS) were analyzed using Cox regression models. Performance of models was evaluated with the Harrell C and Akaike information criterion. The net reclassification improvement of the early death model was also calculated relative to the 5-year OS model for one-year all-cause mortality. A total of 121 patients with advanced buccal cancer were recruited. Mean age was 56.1 ± 9.8 years; 117 (96.7%) patients were male. sTILs ≤30%, clinical nodal disease, pathological nodal disease, poor differentiation, lymphovascular invasion, perineural invasion, WPOI 5, and no adjuvant radiotherapy were risk factors for early death in univariate analysis. In multivariate analysis, clinical TNM, sTILs, clinical nodal disease, poor differentiation, lymphovascular invasion, and no adjuvant RT were independent factors for early death. sTILs, pathological nodal disease, poor differentiation, lymphovascular invasion, and no adjuvant RT were independent factors for early death in the multivariate model with pathological TNM. The discriminatory ability was better for early death model for 1-year all-cause mortality. Finally, incorporation of sTILs into the early death model increased net reclassification by 21% for the clinical TNM model and 28% for the pathological TNM model. Addition of sTILs improved the early death model, which may help physicians to identify high-risk patients for more intensive treatment and follow-up.