Identifying the prognostic value of MRI-based tumor response and predicting the risk of radio-resistance in re-radiotherapy for locally recurrent nasopharyngeal carcinoma

To investigate the prognostic value of tumor response (TR) for locoregionally recurrent nasopharyngeal carcinoma (lrNPC) patients at the end of re-radiotherapy (re-RT) and develop a risk score model to predict patient's radiosensitivity to re-RT.A total of 594 patients with lrNPC from 2010 to 2020 were retrospectively reviewed as the total cohort. Among these, 310 patients with complete first-line treatment data were reviewed as a secondary cohort. Overall survival (OS) was the primary endpoint. Locoregional control (LRC) was the secondary endpoint. Multivariate Cox analysis was performed to investigate the prognostic value of TR at the end of re-RT (rTR). A risk score model for predicting rTR was obtained by logistic regression analysis, and its effectiveness was compared using receiver operating characteristic (ROC) analysis.Patients with complete response (CR) to rTR had higher 5-year OS and LRC rate than non-CR patients in both the total and secondary cohort. rTR was an independent prognostic factor for OS (P = 0.002) and LRC (P = 0.008). We developed a risk score model including four significant risk factors (relapse T stage, relapse gross tumor volume, time to recurrence, and initial TR). The area under the curve of the risk score model was 0.73 (95% CI: 0.678 to 0.780), which was significantly higher than that of each variable alone. Patients with the highest risk scores may be insensitive to re-RT and had a residual tumor risk of 89.9% after rRT.rTR was an independent prognostic factor for OS and LRC in lrNPC patients. We developed a risk score model for predicting patients' sensitivity to re-RT to screen for radiosensitive patients. This can serve as a treatment decision-making tool for clinicians.