Background: We aimed to compare the predictive performance of pN-categories in oral squamous cell carcinoma (OSCC) encompassing the most recent 8th edition (TNM8), its predecessor (TNM7), and a newly proposed algorithm (pN-N+), which classifies patients according to the number of positive lymph nodes and extranodal extension.

Methods: Consecutive, primary OSCC patients from seven previously published cohorts were included and classified according to the three pN-classifications: TNM7, TNM8 and pN-N+. Overall survival probabilities were summarised with the Kaplan -Meier method. We added each of the three metrics to a Cox regression adjusted for pT-category, lymph nodal yield, age, sex, radiotherapy and chemotherapy, and trained these models in one institution. We evaluated the predictive performance in the remaining six institutions and assessed the predicted 5-year risk of death using the area under the receiver operating characteristics curve (AUC) and Brier scores.

Results: All 1,905 included patients were classified according to TNM7 and pN-N+. A subset of 1,575 patients was additionally classified according to TNM8, leading to upstaging in 27.0%. The pN-N+ ranked overall best determined by the obtained AUC and Brier scores. In contrast to pN-N+, TNM7 and TNM8 both suffered from disproportionate patient distribution across pN-categories and poor pN-categorial discrimination on overall survival.

Conclusions: The TNM8 pN-classification designates a larger subset to more advanced disease stages but failed to show improvement of its predictive performance compared to TNM7. The pN-categories of TNM7/8 are disproportionate and inconsistently discriminated. The pN-N+ conveyed the best measures of prognosis and should be considered in future TNM iterations. (C) 2021 Elsevier Ltd. All rights reserved.