Tumor residue in patients with stage II-IVA nasopharyngeal carcinoma who received intensity-modulated radiation therapy: development and validation of a prediction nomogram integrating postradiotherapy plasma Epstein-Barr virus deoxyribonucleic acid, clinical stage, and radiotherapy dose

Background: To develop and validate a predictive nomogram for tumor residue 3-6 months after treatment based on postradiotherapy plasma Epstein-Barr virus (EBV) deoxyribonucleic acid (DNA), clinical stage, and radiotherapy (RT) dose in patients with stage II-IVA nasopharyngeal carcinoma (NPC) treated with intensity-modulated radiation therapy (IMRT).

Methods: In this retrospective study, 1050 eligible patients with stage II-IVA NPC, who completed curative IMRT and underwent pretreatment and postradiotherapy (-7 to +28 days after IMRT) EBV DNA testing, were enrolled from 2012 to 2017. The prognostic value of the residue was explored using Cox regression analysis in patients (n=1050). A nomogram for predicting tumor residues after 3-6 months was developed using logistic regression analyses in the development cohort (n=736) and validated in an internal cohort (n=314).

Results: Tumor residue was an independent inferior prognostic factor for 5-year overall survival, progression-free survival, locoregional recurrence-free survival and distant metastasis-free survival (all P<0.001). A prediction nomogram based on postradiotherapy plasma EBV DNA level (0 vs. 1-499 vs. ≥500 copies/ml), clinical stage (II vs. III vs. IVA), and RT dose (68.00-69.96 vs. 70.00-74.00 Gy) estimated the probability of residue development. The nomogram showed better discrimination (area under the curve (AUC): 0.752) than either the clinical stage (0.659) or postradiotherapy EBV DNA level (0.627) alone in the development and validation cohorts (AUC: 0.728).

Conclusions: We developed and validated a nomogram model integrating clinical characteristics at the end of IMRT for predicting whether tumor will residue or not after 3-6 months. Thus, high-risk NPC patients who might benefit from immediate additional intervention could be identified by the model, and the probability of residue can be reduced in the future.