Association of Pathologic Nodal Metastasis Count With Oncologic Outcomes in Head and Neck Cancer: A Secondary Analysis of RTOG 9501, RTOG 0234, and EORTC 22931.

Abstract

PURPOSE/OBJECTIVE(S)\nCurrent pathologic lymph node (LN) staging for head and neck squamous cell carcinomas (HNSCC) is based on LN size, laterality, and presence of extracapsular extension. However, some retrospective registry and single institution studies have suggested that the number of positive (+) LNs is the dominant nodal factor driving survival in HNSCC. Given the limitations of these datasets, it is unclear whether number of +LNs increases mortality through increased locoregional recurrence (LRR), distant metastases (DM), or both. In this study, we evaluate the prognostic value of number of +LNs in HNSCC using an international cohort of patients enrolled on three prospective randomized controlled trials (RCTs).\n\n\nMATERIALS/METHODS\nSecondary analysis was performed of RTOG 9501, RTOG 0234, and EORTC 22931, 3 prospective RCTs of patients undergoing surgery followed by post-operative radiation (RT) with or without systemic therapy. Univariate and multivariable models with restricted cubic splines were constructed to evaluate the association between number of +LNs and overall survival (OS), disease-free survival (DFS), LRR, and DM. Competing risks analysis was used for LRR and DM. Non-linear restricted cubic splines were used to model the relationship between +LN number and outcomes.\n\n\nRESULTS\nOverall, 947 patients were included in this secondary analysis. OS independently decreased with increasing number of +LNs without plateau, which was most pronounced up to a change point of 5 +LNs (hazard ratio [HR], 1.19 per LN; 95% confidence interval [CI], 1.10-1.29; P < 0.001). Mortality risk continued to increase after 5 +LNs, but more slowly (HR per LN, 1.03; 95% CI, 1.01-1.05; P < 0.001). Similarly, DFS decreased sharply up to 5 LNs (HR per LN, 1.19; 95% CI, 1.1-1.29; P < 0.001), and more gradually beyond this (HR per LN, 1.03; 95% CI, 1.01-1.05; P < 0.001). DM displayed similar behavior to OS and DFS, with risk increasing sharply with each +LN up to 5 (HR per LN, 1.14; 95% CI, 1.01-1.28; P\u202f=\u202f0.04), and more slowly beyond this (HR per LN, 1.05; 95% CI, 1.02-1.07; P\u202f=\u202f0.002). By contrast, although LRR increased sharply up to 5 +LNs (HR per LN, 1.25; 95% CI, 1.11-1.39, P < 0.001), risk plateaued beyond this (HR per LN, 1.00; 95% CI, 0.967-1.04; P\u202f=\u202f0.91). There were marked differences in 5-year OS (54% vs 42%, P < 0.001), DFS (47% vs 27%, P < 0.001), LRR (22% vs 32%, P\u202f=\u202f0.003), and DM (22% vs 38%, P < 0.001) when comparing patients with ≤5 and > 5 +LNs, respectively.\n\n\nCONCLUSION\nNumber of +LNs was validated as an independent factor associated with worse OS and DFS without plateau in a secondary analysis of 3 RCTs of patients with HNSCC undergoing surgery and post-operative RT. For patients with ≤5 +LNs, increased mortality was driven by both increased LRR and DM risk. However, beyond 5 +LNs, additional mortality increases were driven entirely by increasing DM. These changing patterns of recurrence as a function of quantitative nodal burden can help inform future clinical trials attempted to improve outcomes through treatment intensification.