Dose-Gated, Letd-Triggered Signatures of Mandibular Osteoradionecrosis (ORN) after Head-and-Neck Intensity Modulated Proton Therapy
Abstract
Purpose
Linear energy transfer (LETd) has been implicated in mandibular osteoradionecrosis (ORN) following proton therapy; however, its strong spatial coupling with dose has limited interpretation of whether LETd contributes independently to ORN risk. This study aims to (1) characterize the dose-gated nature of mandibular ORN, (2) decouple dose and LETd at the voxel level, and (3) determine whether LETd further modulates ORN likelihood within high-dose mandibular subvolumes.
Methods
Voxel-level physical dose and dose-averaged LET (LETd) were analyzed in six head-and-neck proton therapy patients with mandibular ORN. ORN voxels were segmented within the mandible, and non-ORN control voxels were sampled excluding a 3-mm expanded ORN margin. Physical dose was normalized to the patient-specific maximum mandibular dose. Analyses were restricted to Dose_norm ≥ 0.5–0.7, corresponding approximately to the doses of ~35–50 Gy. Dose–LET separation was evaluated using a dose-controlled logistic regression model. ORN and control voxels were subsequently matched one-to-one within narrow Dose_norm bins, enforcing identical dose distributions relative to Dmax prior to LETd comparison. Voxel-level discrimination was assessed using ROC analysis.
Results
ORN voxels consistently localized to high-dose mandibular regions, confirming a dose-gated signature. LETd remained a significant independent modifier of ORN risk within this high-dose regime. Dose matching achieved negligible residual dose imbalance, demonstrating effective physical dose–LET decoupling. Within dose-matched analyses, ORN voxels exhibited systematically higher LETd than matched controls, with substantial inter-patient heterogeneity. Voxel-level discrimination was highest for the combined Dose_norm + LETd model (AUC = 0.845), compared with Dose_norm alone (AUC = 0.773) and LETd alone (AUC = 0.534).
Conclusion
Mandibular ORN exhibits a dose-gated signature, with LETd providing additional, dose-independent risk modulation within high-dose mandibular subvolumes. By combining statistical and physical dose–LET separation, this framework clarifies when LETd is clinically relevant and supports LET-aware refinement of mandibular avoidance, without assuming a universal LETd threshold.