Yuwei Wang, Kaiyun You, Xiaowen Lan, Peiyan Liang, Yujun Hu, Yanhui Jiang, Han Zhou, Xingsheng Qiu
Pretreatment MRI-defined necrotic nodal phenotypes provide additional risk stratification beyond LNN alone and may identify a subgroup of N3 NPC patients with increased risk of distant metastasis. These findings require external validation before clinical implementation.
BACKGROUND: Distant metastasis remains a major cause of treatment failure in nasopharyngeal carcinoma (NPC) treated with definitive intensity-modulated radiotherapy (IMRT), particularly among patients with advanced nodal disease. Although lymph node necrosis (LNN) is an adverse imaging feature, its prognostic significance may vary according to accompanying nodal characteristics.
METHODS: This retrospective cohort study included 800 patients with newly diagnosed node-positive non-metastatic NPC treated with definitive IMRT between 2016 and 2023. Pretreatment MRI was reviewed for LNN-related features, including radiologic extranodal extension (rENE), necrotic nodal burden, and anatomic distribution. Associations with distant metastasis-free survival (DMFS) were assessed using Cox regression and overlap weighting.
RESULTS: Among 800 patients, 450 (56.3%) had LNN. After overlap weighting, LNN remained independently associated with inferior DMFS (5-year DMFS, 87.0% vs. 94.2%; weighted HR, 2.26; 95% CI, 1.24-4.12; P = 0.008). Among patients with LNN, severe rENE, higher necrotic nodal burden, and multi-area necrosis were associated with increased metastatic risk. The predefined adverse necrotic nodal phenotype did not further stratify N2 disease but identified significant heterogeneity within N3 disease, with inferior DMFS in N3b compared with N3a patients (HR, 2.52; 95% CI, 1.18-5.39; P = 0.017).
CONCLUSIONS: Pretreatment MRI-defined necrotic nodal phenotypes provide additional risk stratification beyond LNN alone and may identify a subgroup of N3 NPC patients with increased risk of distant metastasis. These findings require external validation before clinical implementation.