We retrospectively analyzed consecutive OSCC patients treated between January 1, 2018, and June 1, 2024. Inclusion criteria were: preoperative ¹⁸F-FDG PET/CT within 4 weeks before surgery showing no distant metastases, curative resection of the primary tumor with neck dissection, pathological confirmation of OSCC, and ≥12 months of follow-up for surviving patients. Exclusion criteria included tumor recurrence, prior malignancy, treatment before PET/CT, blood glucose >200 mg/dL during PET/CT, or death within 30 days post-surgery. Patients were stratified by pathological lymph node status (pN0 vs. pN+).

Patients fasted for ≥6 hours before ¹⁸F-FDG injection. ¹⁸F-FDG was administered intravenously at 0.1 mCi/kg, PET/CT scans were conducted using the Ingenuity TF (Philips Medical Systems, Cleveland, USA) at Shanghai Ninth People’s Hospital (11). Images were acquired 60 minutes post-injection, skull base to mid-thigh/foot. Non-contrast CT (120 kVp; 150–200 mA; 3 mm thickness; 512×512 matrix) followed by PET (90 seconds/bed). PET data reconstructed on-scanner: OSEM, 3 iterations, 33 subsets. Tumor volumes of interest (VOIs) were auto-delineated via 40% SUVmax threshold (12), manually adjusted to exclude adjacent physiological ¹⁸F-FDG-avid structures. T-SUVmax, T-SUVmean, and T-MTV auto-calculated; T-TLG = T-SUVmean × T-MTV. In pN+ patients, the nodal maximum standardized uptake value (N-SUVmax) measured at most metabolically active lymph node; if no nodes exceeded background SUVmax, longest-diameter node on positive side selected.

Non-contrast CT images from PET/CT were analyzed using SliceMatic 5.0 software (Tomovision, Montreal, Canada). A cross-sectional CT image at the third lumbar (L3) level, clearly showing the L3 transverse processes, was selected. Paraspinal and parietal muscles were semi-automatically segmented using Hounsfield unit (HU) thresholds (-29 to +150). L3-SMI was calculated as cross-sectional muscle area (cm²) normalized to height² (m²). Chinese population-specific sarcopenia cutoffs were applied (men: <55.0 cm²/m²; women: <36.6 cm²/m²) (13).

Primary tumor features (AJCC/UICC TNM staging, 8th edition) included tumor grade, PNI, lymphovascular invasion (LVI), surgical margin status, depth of invasion (DOI; categorized as ≤5 mm, >5–10 mm, or >10 mm), pathological T stage, and pathological N stage. For pN+ patients, lymph node characteristics evaluated included extranodal extension (ENE), lymph node yield (LNY; total nodes removed), number of positive nodes, and lymph node density [LND; ratio of positive to total nodes removed, also termed lymph node ratio (LNR)].

Adjuvant therapy recommendations were determined by a multidisciplinary tumor board (MDT) involving oncologists, surgeons, radiologists, and nuclear medicine physicians, with treatment initiated within 6 weeks post-surgery. Radiotherapy (56–60 Gy total dose at 2.0 Gy per fraction via linear accelerator) and/or cisplatin-based chemotherapy were administered based on clinical status, pathological findings, and nodal involvement.

Statistical analysis was performed using SPSS version 17.0, (Chicago, IL, USA) and MedCal19.6.1 (Ostend, Belgium). No missing data were identified in this study. Continuous variables were expressed as mean ± standard deviation or median (range), while categorical variables were presented as frequencies (percentages). Between-group comparisons were performed using Student’s t-test, Mann-Whitney U test, χ² test, or Fisher’s exact test as appropriate. Survival outcomes were analyzed using Cox proportional hazards regression models, with results expressed as hazard ratios (HRs) and 95% confidence intervals (CIs). Primary endpoints were progression-free survival (PFS) and overall survival (OS), defined from PET/CT date to OSCC-related recurrence/all-cause death (censored at June 1, 2025). Receiver operating characteristic curve analysis with Youden index identified optimal cutoffs for key variables (Supplementary Table 1), using OS as the endpoint. Univariate Cox models were followed by multivariate analysis (backward stepwise) with P < 0.05 variables. T-MTV was prioritized over T-TLG due to multicollinearity (VIF = 9.418). Kaplan-Meier curves and log-rank tests visualized survival differences. To evaluate the stability of the findings and quantify potential bias, internal validation of prognostic factors was conducted through bootstrap resampling with 1000 repetitions by using R version 4.1.0. A two-sided P < 0.05 was considered statistically significant.

A total of 147 OSCC patients were included, with demographic/clinical characteristics in Table 1. Median follow-up was 43 months (12–83): 42.5 months for pN0 and 44 months for pN+. During follow-up, 42 patients (29%) died, including 12 in the pN0 and 30 in the pN+ group. Disease progression or recurrence occurred in 52 patients (35%), with 15 in the pN0 and 37 in the pN+ group. Two patients developed second primary malignancies: one in the pN0 group (kidney carcinoma at 4 months) and one in the pN+ group (esophageal carcinoma at 36 months). Both were alive without OSCC recurrence at the end of follow-up. Adjuvant therapy was recommended for 39.2% of pN0 patients (29/74, all completed) and all pN+ patients (73/73), with 84.9% (62/73) completing it. Among pN+ patients who did not complete adjuvant therapy (9 due to physical status, 2 due to patient refusal), there were 7 males [L3-SMI: 35.71 (29.83–53.91)] and 4 females [L3-SMI: 35.71 (30.29–44.25)]; 81.8% (9/11) of these patients had sarcopenia.

A comparison of prognostic variables between pN0 and pN+ patients is presented in Table 2. Higher values of T-SUVmax, T-MTV, T-TLG, as well as younger age, were observed in pN+ patients compared to pN0 patients. Furthermore, DOI >5mm, PNI and poorly differentiated tumor grade were more frequently observed among pN+ patients.

In pN0 patients, T-SUVmax, along with T-MTV and T-TLG, emerged as significant prognostic factors for OS and PFS in the univariate analysis. In the multivariate analysis, T-SUVmax remained an independent predictor for OS (HR:10.240, 95%CI:2.161-48.512, P = 0.003) and PFS (HR:8.229, 95%CI:2.224-30.444, P = 0.002) (Table 3). This finding was confirmed by Kaplan-Meier survival analysis, as shown in Figure 1.

In pN+ patients, the univariate analysis for OS identified T-MTV, T-TLG, N-SUVmax, L3-SMI, the number of positive nodes, and LND as significant prognostic factors. Among these, the number of positive nodes (HR:4.151, 95%CI:1.887-9.133, P = 0.000) and L3-SMI (HR:0.133, 95%CI:0.018-0.975, P = 0.047) were independently associated with OS in the multivariate analysis. Similarly, for PFS, the univariate analysis identified T-MTV, N-SUVmax, L3-SMI, the number of positive nodes, and pT stage as significant prognostic factors. N-SUVmax (HR:2.343, 95%CI:1.151-4.769, P = 0.019), L3-SMI (HR:0.177, 95%CI:0.039-0.799, P = 0.024), pT stage (HR:3.739, 95%CI:1.367-10.230, P = 0.010), and the number of positive nodes (HR:1.971, 95%CI: 1.003-3.874, P = 0.049) emerge as independent predictors in the multivariate analysis (Supplementary Table 2). Notably, L3-SMI and the number of positive nodes were the two factors consistently associated with both OS and PFS, as confirmed by Kaplan-Meier survival analysis (Figure 2).

Bootstrap validation confirmed the prognostic value of key factors in both cohorts. In pN0 patients, T-SUVmax >13.27 was a relative stable predictor for both OS (median HR = 11.80, 95%CI: 3.38-41.95) and PFS (median HR = 10.36, 95%CI: 3.38-40.16). In pN+ patients, L3-SMI was a protective factor (median OS HR = 0.17, 95% CI: 0.07-0.51; median PFS HR = 0.20, 95%CI: 0.06-0.61), while a nodal burden of ≥3 nodes was a powerful risk factor (median OS HR = 4.36, 95%CI: 2.15-11.17; median PFS HR = 2.69, 95%CI: 1.41-5.56).