Subdivision of metachronous metastatic nasopharyngeal carcinoma based on the framework of TNM-9: A recursive partitioning analysis of real-world data

Purpose: To validate the prognostic value of the M1 category in version nine of the AJCC/UICC TNM staging system (TNM-9) in metachronous metastatic nasopharyngeal carcinoma (mmNPC) and to develop a refined M1 classification for improved risk stratification. Methods: Patients with newly diagnosed mmNPC (2015–2024) were analyzed and restaged per TNM-9 M1 criteria. Independent prognostic factors from multivariate Cox analysis were integrated using recursive partitioning analysis (RPA) to establish an RPA-based M1 classification (RPA-M1). Model performance was assessed using Harrell's C-index, time-dependent AUC, and Akaike information criterion (AIC). Results: Among 218 patients, 111 (50.9%) were M1a and 107 (49.1%) M1b. The M1a subgroup demonstrated significantly higher 3-year overall survival (OS) than the M1b subgroup (63.2% vs. 40.8%, P < 0.001). Both the TNM-9 M1 category and disease-free interval (DFI) were independent prognostic factors for OS (both P < 0.05). The RPA model stratified patients into distinct prognostic groups: RPA-M1a (≤ 3 metastatic lesions and DFI > 12 months) and RPA-M1b (> 3 lesions or DFI ≤ 12 months), with 3-year OS rates of 74.6% and 40.6%, respectively (P < 0.001). The RPA-M1 classification showed superior prognostic discrimination compared with the TNM-9 M1 stage, evidenced by higher C-index, time-dependent AUCs, and lower AIC. Metastasis-directed therapy improved OS only in the RPA-M1a subgroup, while intensified systemic therapy primarily benefited the RPA-M1b subgroup. Conclusion: The TNM-9 M1 category is applicable to mmNPC. The novel RPA-M1 classification, which integrates metastatic burden and DFI, offers superior risk stratification and enables more precise prognostication and individualized management.