An Individualized Prognostic Model in Patients with Locoregionally Advanced Nasopharyngeal Carcinoma Based on Serum Metabolomic Profiling

**Background:** Nasopharyngeal carcinoma (NPC) is prevalent in Southern China and Southeast Asia, with ~75% of patients presenting at locoregionally advanced stages. Despite standard treatment, ~30% experience treatment failure. The TNM staging system inadequately captures tumor heterogeneity, highlighting the need for novel biomarkers. Metabolomics offers a promising approach to reflect tumor biology, but its application in NPC prognosis has been limited. This study aimed to develop a serum metabolomic signature to predict progression-free survival (PFS) in LA-NPC.

**Methods:** A total of 320 LA-NPC patients (stage III–IVa) treated with platinum-based chemotherapy and intensity-modulated radiotherapy at Sun Yat-sen University Cancer Center (2013–2016) were enrolled. Pretreatment fasting serum samples underwent widely targeted metabolomic profiling using UPLC–MS/MS (Metware database), detecting 746 endogenous metabolites. Patients were randomly split 7:3 into training (n=224) and validation (n=96) sets. Univariate Cox regression (FDR < 0.05) identified 27 PFS-associated metabolites. Multivariate Cox regression with AIC-based stepwise selection yielded a 9-metabolite signature. A metabolic risk score (Met score) was calculated using Cox regression coefficients. Patients were stratified into high- and low-risk groups by the median Met score (cutoff = 0.92). Performance was assessed via Kaplan–Meier survival analysis, time-dependent ROC curves, and C-index. A nomogram integrating Met score, sex, overall stage, and EBV-DNA level was developed.

**Key Results:** Median follow-up was 64.0 months (training) and 65.7 months (validation). The 9-metabolite signature significantly stratified PFS: in the training set, 5-year PFS was 53.7% (95% CI 45.12–63.86) for high-risk vs. 83.0% (95% CI 76.31–90.26) for low-risk (HR = 3.40, 95% CI 2.13–5.40, log-rank p < 0.001). In the validation set, 5-year PFS was 58.6% (95% CI 46.4–74.1) for high-risk vs. 90.9% (95% CI 82.8–99.8) for low-risk (HR = 3.68, 95% CI 1.68–8.08, p = 0.003). AUC for 5-year PFS was 0.76 (training) and 0.75 (validation). In multivariate analysis, Met score (HR 1.18, 95% CI 1.11–1.26, p < 0.001), overall stage (HR 1.93, 95% CI 1.19–3.15, p = 0.008), sex (HR 2.44, 95% CI 1.25–4.80, p < 0.001), and EBV-DNA (HR 2.33, 95% CI 1.44–3.79, p < 0.001) were independent predictors. The combined nomogram showed improved C-index vs. the traditional model (training: 0.77 vs. 0.71; validation: 0.72 vs. 0.67). KEGG enrichment identified tyrosine metabolism as the most significantly enriched pathway (p < 0.05).

**Clinical Implications:** This study provides the first large-scale metabolomics-based prognostic model for LA-NPC. The 9-metabolite signature offers a non-invasive, convenient, and potentially cost-effective tool to complement TNM staging and EBV-DNA for risk stratification. The nomogram could help identify high-risk patients who may benefit from intensified treatment strategies. Limitations include single-institution design, need for external validation, and unclear biological mechanisms of several metabolites. Future work should include prospective, multi-center studies and multi-omics integration.