Prognostic association of postoperative pre-radiotherapy neutrophil-to-lymphocyte ratio with disease-free survival in oral squamous cell carcinoma patients undergoing adjuvant volumetric modulated arc therapy: an exploratory analysis

The prognostic value of postoperative pre-radiotherapy neutrophil-to-lymphocyte ratio (NLR) in oral squamous cell carcinoma (OSCC) patients receiving adjuvant volumetric modulated arc therapy (VMAT) remains inadequately characterized. This study evaluated the association of NLR with disease-free survival (DFS), explicitly testing its functional form and treating any data-derived cutoff as exploratory. We retrospectively analyzed 102 OSCC patients who underwent curative surgery followed by adjuvant VMAT between January 2020 and December 2023. Postoperative pre-VMAT NLR was calculated from blood counts obtained within two weeks before VMAT initiation. NLR was analyzed as a continuous variable using restricted cubic splines and Cox regression, with the proportional-hazards assumption tested. A receiver operating characteristic (ROC)-derived cutoff was evaluated only as an exploratory secondary analysis. Disease-free survival (DFS) was assessed with the Kaplan-Meier method and Cox regression and overall survival (OS) was analyzed descriptively. A parsimonious model was built transparently and internally validated with 500-sample bootstrap resampling, with discrimination summarized by the optimism-corrected concordance index (C-index) and censoring-aware time-dependent area under the curve (tdAUC). Calibration and decision curve analysis (DCA) were performed. The cohort comprised 64 males (62.7%) and 38 females (37.3%), with a median age of 59.0 years (IQR 51.0–64.0). The median NLR was 2.03 (IQR 1.49–2.92). Over a median follow-up of 32.5 months, 24 DFS events occurred (5 locoregional recurrences, 18 distant metastases, and 1 death without prior recurrence). NLR showed no evidence of a nonlinear association with log-hazard (nonlinearity P = 0.374). As a continuous variable, higher NLR was associated with worse DFS (adjusted HR 1.16 per unit, 95% CI 1.03–1.30, P = 0.011) after adjustment for age and sex. The exploratory ROC-derived cutoff of 1.70 (AUC = 0.656) separated Kaplan-Meier curves (log-rank P = 0.011). The multivariable model achieved an apparent C-index of 0.738 and a bootstrap optimism-corrected C-index of 0.710, with tdAUCs of 0.822 (1-year) and 0.790 (2-year). Higher postoperative pre-radiotherapy NLR was associated with inferior DFS in OSCC patients undergoing adjuvant VMAT. However, modest and optimism-prone discrimination, the absence of key pathological covariates (extranodal extension, perineural and lymphovascular invasion, depth of invasion), and the small number of events limit definitive conclusions. These hypothesis-generating findings require confirmation in larger, prospectively collected cohorts before NLR can inform individualized management.