There is increasing evidence in the literature that alternative N-classification systems offer a simpler and more precise risk stratification than the current N- classification in patients with oral squamous cell carcinoma. The purpose of this study is to compare three broadly proposed models incorporating lymph node ratio, log odds of positive lymph nodes and number of positive lymph nodes regarding disease-free and overall survival.

This is a retrospective study of patients treated in a single center between 2013 and 2019. Cox proportional hazard models, Kaplan Meier curves, the long rank test and the area under the curve were implemented to compare the risk-stratification ability of the three models. Moreover, a hazard ratio plot was calculated to investigate the association between nodal yield and disease-free survival.

231 patients were included. All three proposed N-models were significantly correlated to the patient's prognosis in the multivariate analysis. Pairwise comparisons between the groups showed that the current pN classification offers the worst stratification and that the model incorporating the number of positive lymph nodes had a better performance for predicting both endpoints. Finally, a decrease in hazard ratio was observed with each additional lymph node removed up to the number of 39 lymph nodes.

The model incorporating the nodal burden offered a better predictive ability. The current N-classification cannot adequately identify patients with different prognosis. A non-linear relationship was found between nodal yield and survival, which could be the disadvantage of the lymph node ratio and log odds models.