Nomogram model to predict postoperative relapse after mandibular osteoradionecrosis surgery

Osteoradionecrosis of the mandible (ORNM) is one of the most devastating complications following radiotherapy. Postoperative relapse (POR) occurs with high incidence even if a radical resection is performed. The current investigation was designed to identify prognostic factors for POR and to establish a nomogram model to estimate the risk for the onset of POR of ORNM.A retrospective study was conducted in ORNM patients during the period from 2003 to 2016. Predictive factors for POR were preliminarily filtered by Kaplan-Meier analysis and were further confirmed by Cox regression model. A nomogram model was established to predict the risk for the onset of POR, and the performance was estimated by receiver operating characteristic (ROC) and calibration curve. POR was defined as the primary outcome variable and was measured using univariate and multivariate analyses.A total of 213 patients were analyzed, and the total incidence of POR was 24.4% (52/213). In the Cox regression analysis, radiation doses ≥80 Gy (versus＜80 Gy, OR = 3.528, P＜0.001, 95% CI: 1.759-7.076), location of ORNM (lesion only in mandibular body versus that involving mandibular body, angulus and ramus versus, OR = 2.900, P = 0.007, 95% CI: 1.345-6.253), S classification (S2 versus S0, OR = 8.926, P = 0.001, 95% CI: 2.487-32.036), and surgical treatment (sequestretomy versus ER + reconstruction, OR = 3.299, P = 0.012, 95% CI: 1.294-8.411) were significantly associated with POR. The current nomogram model can effectively evaluate the hazard risk and survival rate of POR. The discrimination capability was tested by the ROC curve with an area under the curve of 0.813, revealing highly predictive abilities. The calibration curve showed sufficient fitness.The current nomogram model was effective in predicting the risk of POR in ORNM patients.