Abstract
OBJECTIVE: Lymph nodal yield (LNY) is increasingly used in oral squamous cell carcinoma (OSCC) research as a prognostic factor and quality metric but remain to be validated across independent OSCC cohorts. STUDY DESIGN: Retrospective cohort study. SETTING: Primary OSCC patients (n = 1080) who underwent a neck dissection, collected from three sites: Brescia, Cork, and Heidelberg. METHODS: The accuracy of LNY in predicting 3- and 5-year mortality was evaluated using the area under the ROC curve (AUC). Binomial regression models, adjusted for LNY and various confounders, were used to cross-validate predictive performance across independent cohorts. A model estimating the population-average treatment effect of LNY was used to estimate the effect of what would have been observed if the OSCC patients had been randomly assigned to have specific ranges of LNYs removed, irrespective of their individual characteristics. Mean LNY was estimated between the three cohorts. RESULTS: ROC curve analysis demonstrated that LNY did not improve prognostic accuracy in predicting 3- and 5-year mortality risks. Cross-validation across independent cohorts showed that adding LNY to the binomial regression models did not improve predictive accuracy. Using the LNY range of 33 to 44 as a reference, the average treatment effect model found no significant difference in the 3-year risk of death across LNY groups (LNY: 2-17, 17-25, 25-33, and 44-119) for pN-positive patients. CONCLUSION: We found no improvement in the prognostic accuracy for mortality when considering LNY in different settings. Substantial inter-center variability and lack of consistent survival benefit challenge the use of LNY as a prognostic or surgical quality metric.