To develop and validate a novel risk stratification model for locally advanced nasopharyngeal carcinoma (LA-NPC) based on the "glyco-lipid-viral" axis, addressing the limitations of traditional TNM staging and single-modality biomarkers. A total of 486 LA-NPC patients were retrospectively enrolled, with a metabolomics subcohort (n = 121) selected. PET/CT-derived glycolytic parameters (MTV, TLG), serum lipid biomarkers, and pretreatment EBV DNA were integrated via recursive partitioning analysis (RPA) to construct the model. Its performance was compared with TNM staging and single-modality markers. A multimodal model was built via RPA in 121 LA-NPC patients, using Lipsig-Score (lipid metabolite signature score), nasoMTV (PET/CT), and pretreatment EBV DNA. Three risk groups were defined: low-risk (Lipsig-Score < 0. 0127 + EBV DNA < 500 copies/mL), intermediate-risk (Lipsig-Score < 0. 0127 + EBV DNA ≥ 500 copies/mL, or Lipsig-Score ≥ 0. 0127 + nasoMTV < 5. 5), and high-risk (Lipsig-Score ≥ 0. 0127 + nasoMTV ≥ 5. 5). Kaplan–Meier analysis showed significant differences in PFS (log-rank P < 0. 0001) and OS (log-rank P < 0. 001) among groups, with the model outperforming TNM staging and single-modal markers. The integrated model captures LA-NPC’s multi-dimensional biology, enabling precise risk stratification and providing a reliable basis for individualized treatment.
Zhou et al. (Mon,) studied this question.