BackgroundThe present study aims to evaluate dosimetric and clinical risk factors for the development of maxillary osteoradionecrosis (ORN) in head and neck adenoid cystic carcinoma (ACC) patients treated with carbon ion radiotherapy (CIRT).MethodsClinical data and treatment plans of ACC patients, consecutively treated from January 2013 to September 2016 within the phase II clinical trial CNAO S9/2012/C, were retrospectively reviewed. ORN and other treatment-related toxicity were graded according to the Common Terminology Criteria for Adverse Events (CTACE), version 4.0. The maxillary bone was contoured on the planning CT, and only patients receiving more than 10% of the prescription dose at their maxilla were considered for the analysis (67 patients). The volumes of maxilla receiving doses from 10 Gy (RBE) to 60 Gy (RBE) (V_D), with an increment of 10 Gy (RBE), and additional clinical factors were correlated to the incidence of ORN with univariate analysis (Chi-square test). The logistic regression model was subsequently applied for multivariate analysis. Treatment plans calculated with a local effect model (LEM)-based optimization were recalculated with the modified microdosimetric kinetic model (MKM), and compared with literature data from the Japanese experience.ResultsThe median time interval from the start of CIRT to ORN appearance was 24 months (range, 8–54 months). Maxillary ORN was observed in 11 patients (16.4%). Grade 1 ORN was observed in 2 patients (18.1%), G2 in 4 (36.3%), G3 in 4 (36.3%) and G4 in 1 (9.3%). From univariate analysis, the site of the tumor, the presence of teeth within the PTV and acute mucositis correlated with the development of maxillary ORN. V_D were significantly higher for all the dose levels tested in patients with maxillary ORN than patients without necrosis, according to both radiobiological models. The multivariate analysis showed that V60 significantly correlated with ORN risk.ConclusionThe volume of maxilla irradiated with high dose values was relevant for ORN development in our cohort of ACC patients. These results are in line with previously published data obtained with a different radiobiological model. Our findings might be helpful to prevent the risk of ORN in patients receiving CIRT.