Can 3D pseudo-continuous arterial spin labeling perfusion imaging be applied to predict early response to chemoradiotherapy in patients with advanced nasopharyngeal carcinoma?

BACKGROUND AND PURPOSE: Chemoradiotherapy (CRT) has been widely applied in patients with advanced nasopharyngeal carcinoma (ANPC). However, limited imaging modality exists on the evaluation of early response to CRT. The purpose of this study was therefore to investigate whether 3D pseudo-continuous arterial spin labeling (3D pCASL) perfusion imaging could predict early response to CRT in ANPC patients. MATERIALS AND METHODS: Seventy ANPC patients who received CRT underwent pre-treatment MRI including 3D pCASL perfusion measurements, and were categorized into response group (RG) and no-response group (NRG) according to RECIST 1.1. Pre-treatment 3D pCASL derived cerebral blood flow (CBF) values in tumors were compared between RG and NRG patients. Receiver-operating characteristic (ROC) analysis was performed to determine the optimal diagnostic cutoff value for CBF in predicting tumor response to CRT. Clinicopathological variables were also analyzed by using univariate and binary logistic regression. The corresponding obtained variables with statistical significance were further applied to create a nomogram in which the bootstrap resampling method was used for calibration. RESULTS: Forty-eight patients in RG had significantly higher pre-treatment CBF values in tumors compared with 22 patients in NRG (P < 0.001). CBF showed the high area under the ROC curve (AUC = 0.843) in distinguishing RG from NRG patients. The corresponding cutoff value for CBF was 103.68 ml/100 g/min, with respective accuracy, sensitivity and specificity of 82.86%, 87.50% and 72.73%. The nomogram was generated by binary logistic regression results, incorporating three variables: CBF value, clinical stage and pathological type. The AUC, accuracy, sensitivity and specificity of the nomogram was respectively 0.893, 84.28%, 81.25% and 90.91% in predicting tumor response to CRT. Moreover, as shown in the calibration curve, a strong agreement was observed between nomogram prediction probability and actual clinical findings (P = 0.309). CONCLUSIONS: 3D pCASL derived CBF in tumor could act as a noninvasive effective biomarker to predict tumor response to CRT in ANPC patients before clinical treatment. Furthermore, the nomogram combining CBF and clinicopathological variables could serve as a novel clinical analysis tool for treatment response prediction.