Depicting distant metastatic risk by refined subgroups derived from the 8th edition nasopharyngeal carcinoma TNM

Tumor-nodal-metastasis (TNM) is the most important survival predictor in nasopharyngeal carcinoma (NPC). Distant metastasis (DM) is the predominant failure pattern of NPC in the intensity-modulated radiotherapy (IMRT) era. The DM risk appears to be different for T-N subsets within the same clinical stage. Appropriately depicting DM risk has emerged as an important issue in tailoring individualized treatment and underpins the reason for this study.

A total of 1616 non-metastatic (M0) NPC patients treated with IMRT were included. All were re-staged according to the 8th edition AJCC/UICC TNM (TNM-8). DM-free survival (DMFS) was calculated and compared among T-N subsets within each stage and DM risk groups were derived by Recursive-partitioning analysis (RPA) based on ordinal T and N categories.

Significant heterogeneity in DM risk was evident among T-N subsets within cTNM-8 stages II-IV. The RPA algorithm classified patients into four DM risk groups: RPA-I (T1N0-1 and T2-3N0), RPA-II (T2-3N1), RPA-III (T4N0-1 and T1-3N2) and RPA-IV (T4N2 and T1-4N3), with 5-year DMFS of 93.4% (95% CI: 91.3-96.1), 84.3% (80.8-87.8), 78.9% (75.4-82.4) and 63.6% (56.3-70.9), respectively (p < 0.001). Compared to cTNM-8 stage grouping, RPA grouping had a lower Akaike information criterion (AIC) and higher Harrell's concordance index (c-index) for DMFS.

Significant heterogeneity in DM risk exists among T-N subsets within cTNM-8 stages. The RPA groups demonstrated improved intra-group hazard consistency compared to cTNM-8 stage groups. While further validation is warranted, these RPA prognostic groupings provide a strong anatomic foundation to augment DM prediction for optimal targeting in future clinical trials.