Modeling Sarcopenia to Predict Survival for Patients With Nasopharyngeal Carcinoma Receiving Concurrent Chemoradiotherapy

Background The present study aimed to construct a prognostic nomogram including Epstein-Barr virus DNA (EBV-DNA) and sarcopenia in patients with nasopharyngeal carcinoma (NPC) receiving concurrent chemoradiotherapy (CCRT). Methods In this retrospective analysis, we studied 1,045 patients with NPC who had been treated with CCRT between 2010 and 2014. Sarcopenia was determined using routine pre-radiotherapy computed tomography scans of the third cervical vertebrae. A new S-E grade was constructed using a receiver-operating characteristic (ROC) curve analyses determined cutoff values of sarcopenia and plasma EBV-DNA. The nomogram was developed base on the sarcopenia-EBV (S-E) grade and traditional prognostic factors. A calibration curve, time-dependent ROC, decision curve analysis, and the concordance index (C-index) determined the accuracy of prediction and discrimination of the nomogram, and were compared with TNM staging system and a traditional nomogram. Results Patient survival was significantly different when sarcopenia (P < 0.001) or EBV-DNA (P = 0.001) were used and they continued to be independent prognostic factors for survival upon univariate (P < 0.001, P = 0.002, respectively) and multivariate (P < 0.001, P = 0.015, respectively) analyses. Predicting overall survival (OS) was more accurate using the S-E grade than using TNM staging and sarcopenia or EBV-DNA alone. Nomogram B (model with sarcopenia) or nomogram A (model without sarcopenia) were then developed based on the identified independent prognostic factors. Comparing nomogram prediction with actual observation showed good agreement among the calibration curves for probability of 1-, 3-, and 5-year OS. Predicted survival (C-index = 0.77) of nomogram B was statistically higher than that of nomogram A (0.676, P = 0.020) and TNM staging (0.604, P < 0.001). Risk group stratification could distinguish between survival curves within respective TNM stages (all stages, P < 0.001; stage III, P < 0.001; stage IV, P = 0.002). Conclusions The sarcopenia-EBV DNA nomogram allowed more accurate prediction of prognosis for patients with NPC receiving CCRT.