Development Of A Clinicomolecular Risk Stratification System For Nonmetastatic Nasopharyngeal Carcinoma Using Epstein-Barr Virus Dna And Tnm Stage: A "Big Data" Analysis Of 9,160 Endemic Cases.

Fo-Ping Chen,Guan-Qun Zhou,Ying-Shan Luo,Kuan Rui Lloyd Tan,Sze Huey Tan,Yu Zheng,Li Lin,Ying Guo,Jun Ma,Melvin L.K. Chua,Ying Sun

DOI: https://doi.org/10.1200/JCO.2019.37.15_suppl.6043

IF: 45.3

2019-01-01

Journal of Clinical Oncology

Abstract:6043 Background: To construct a clinicomolecular index integrating circulating Epstein-barr virus (cfEBV) DNA with T- and N- categories for better prognostication in nasopharyngeal carcinoma (NPC). Methods: Clinical and treatment records of 9,160 biopsy-proven, non-metastatic NPC cases were identified from an institutional “Big-data” platform. Decision tree modeling (DTM), recursive partitioning analysis (RPA) and adjusted hazard ratio (AHR) methods were used to generate clinicomolecular risk models. Outcome prediction of the models were compared against 8th edition TNM stage and two RPA-original models. Results: We observed linearity between cfEBV DNA and DFS; cfEBV DNA of > 2,000 copies was consistent for risk discretisation (HR > 1.0) for DFS, OS and DMFS in our cohort of 9,160 patients. DTM, RPA-new and AHR modelling using a two-tiered stratification by cfEBV DNA (≤2,000 and > 2,000 copies) and T- and N- categories yielded five risk groups with significantly disparate DFS (P < 0.001 for all subgroup comparisons). AHR model outperformed all other models and the TNM stage classification with better hazard consistency, hazard discrimination, explained variation, sample size balance and likelihood difference. Importantly, our clinicomolecular AHR groupings were significantly associated with the efficacy of different therapeutic regimes. Outcomes were comparable between concurrent chemoradiotherapy and intensity-modulated radiotherapy (IMRT) and IMRT alone for AHR1-2 (3-y DFS [AHR1] = 96.2% [IMRT] vs 95.3% [chemo-IMRT]; 3-y DFS [AHR2] = 91.1% vs 90.4%). Neoadjuvant chemotherapy and chemo-IMRT was superior to chemo-IMRT alone for AHR4-5 (AHRDFS 0.77[0.65-0.93], P = 0.005; 0.77[0.61-0.97], P = 0.027), but not for AHR3 (AHRDFS 1.07[0.86-1.34]). Conclusions: Here, we present a robust clinicomolecular risk stratification system that outperforms the TNM stage classification in non-metastatic NPC. Our clinicomolecular model is associated with the efficacy of different therapeutic regimes.