Distant lymph node metastasis in head and neck cancer: incidence, predictors, and survival outcomes in a population-based cohort

Abstract

Background Distant lymph node metastasis (DLM) is rare in head and neck cancer (HNC), occurring in only 0.76% of cases, underscoring the importance of understanding its clinical behavior. However, population-based data on the incidence, predictors, and prognostic implications of DLM relative to other metastatic patterns remain limited. This study aimed to determine the incidence of DLM at diagnosis, identify its independent predictors, and evaluate the impact of DLM and its treatment on survival in a population-based cohort of patients with HNC. Methods We performed a retrospective cohort study using the Surveillance, Epidemiology, and End Results (SEER) database [2016–2022] and identified 42,908 patients with HNC. Multivariable logistic regression was used to determine factors associated with DLM at diagnosis. Overall survival (OS) was estimated using the Kaplan-Meier method. Cancer-specific mortality was assessed using a competing-risks analysis, while treatment effects within the DLM-positive cohort were studied using a 3-month landmark analysis to reduce immortal-time bias. Results DLM was present at diagnosis in 325 patients (0.76%). Independent prognostic factors were age 60 years or older, hypopharyngeal tumor primary site, advanced tumor (T) and node (N) stage, neuroendocrine carcinoma, and melanoma (all P<0.05). Model discrimination was good [area under the receiver operating characteristic curve (AUC) =0.835]. Patients with DLM had significantly shorter median OS than those without distant metastasis (8.0 vs. 78.0 months, P<0.001). Among the DLM-positive population, isolated nodal disease correlated with longer median OS than DLM in the presence of visceral spread (12.0 vs. 6.0 months; P<0.001), and this benefit persisted in a competing-risks analysis for cancer-specific death. In the landmark analysis, combined chemoradiotherapy remained associated with improved survival compared with no active treatment [adjusted hazard ratio (HR) =0.41, 95% confidence interval (CI): 0.26–0.63; P<0.001]. Conclusions These findings suggest that subclassifying metastatic disease could enhance risk stratification and guide more personalized treatment strategies.