Prognostic Value of Radiologic Extra-nodal Extension and Its Potential to Refine the N Classification for Nasopharyngeal Carcinoma

T. Lu,Y. Hu,Y. Xiao,Q. Guo,S. H. Huang,Y. Fang,J. Zong,Y. Chen,B. O'Sullivan,S. Lin,J. Pan
DOI: https://doi.org/10.1016/j.ijrobp.2019.06.1567
2019-01-01
Abstract:The 8th edition TNM (TNM8) includes extra-nodal extension (ENE) for non-viral head and neck cancer. The prognostic value of ENE has not been emphasized in nasopharyngeal carcinoma (NPC). We evaluated the prognostic value of various grades of radiologic ENE (rENE) and their potential role in refining the N classification for NPC. All cN+ NPC patients (pts) treated with IMRT in our institution between 2005 and 2011 were included. Pre-treatment MRIs were reviewed by a specialized neuroradiologist. rENE was defined as a lymph node (LN) with irregular borders with (high-grade, HG) or without (low-grade, LG) adjacent structure invasion. If ≥2 LNs abutted together with rENE (loss of internodal fat plane) but without adjacent structure invasion, it was termed inter-mediate grade rENE (IG-rENE). Distant metastases free survival (DMFS), regional relapse free survival (RRFS), and overall survival (OS) were compared among rENE grades vs. cN+ without rENE (No-rENE). Multivariable analyses (MVA) confirmed the incremental prognostic value of the various grades of rENE. Refined N classification incorporating rENE grades was compared with TNM8 using Akaike information criterion (AIC) and Harrell’s concordance index (c-Index). A total of 1390 of 1616 (86%) NPC were cN+, in whom rENE was detected in 826 (59%) pts: 256 (18.4%) LG-rENE, 487 (35%) IG-rENE, and 83 (6%) HG-rENE. Median follow-up was 53 months. Actuarial rates of 5-year DMFS and OS decreased with higher grade of rENE (No-/LG-/IG-/HG-rENE): 88%/82%/74%/56% (p<0.001), 87%/81%/77%/60% (p<0.001), respectively. MVA (adjusted for age, gender, T, N, chemotherapy, LDH) confirmed that IG-rENE and HG-rENE had increased HRs for DM (HR: 1.97 and 3.00, p<0.001 and <0.001) and death (HR: 1.69 and 2.45, p=0.002 and <0.001), while LG-rENE was not prognostic (DMFS: HR=1.280, p=0.224; OS: HR=1.249, p=0.259). 5-year RRFS rates differed with rENE grades (96%/91%/93/91%, p=0.042) but was non-prognostic in MVA (p=0.286). N2 with No/LG-rENE had similar DMFS (87% vs. 85%, p=0.600) and OS (86% vs. 83%, p=0.630) vs N1 with No/LG pts. N2 with IG-rENE had similar DMFS (69% vs. 65%, p=0.509) and OS (71% vs. 71%, p=0.672) compared to N3 with No-/LG-rENE; HG-rENE pts had the most detrimental DMFS compared to the other groups (all p<0.05). We propose a refined N for future TNM to include IG-rENE and HG-rENE: New N1: N1 or N2_No/LG-rENE; New N2: N1_IG-rENE; New N3: N3 or N2_IG-rENE or any N_HG-rENE. Compared to the TNM8 N, the new N classification had a lower AIC and higher c-index for DMFS (AIC: 3809.6 vs 3830.9; c-index: 0.700 vs. 0.677) and OS (AIC: 3693.8 vs. 3705.9; c-index: 0.735 vs. 0.725). IG- and HG-rENE are independently prognostic for DMFS and OS in NPC. Compared to the TNM8 N classification, an N classification incorporating IG- and HG-rENE improves prognostication of risk of DM and death.
What problem does this paper attempt to address?