Retrospective analysis of dose and dose-averaged LET combined effect on local tumour control in adenoid cystic carcinoma treated with carbon-ion radiotherapy.

Abstract

To investigate the association between RBE modeling, RBE-weighted dose, and dose-averaged LET (LET_d) distributions and local control (LC) in adenoid cystic carcinoma (ACC) patients treated with carbon-ion radiotherapy (CIRT). 74 ACC patients treated with CIRT (2016-2023) were retrospectively analysed (median follow-up: 47 months). Treatment was planned using the local effect model (LEM-I) with prescriptions of 68.8 or 65.6 Gy(RBE) in 16 fractions. Patients were stratified as LC or local relapse (LR). Clinical plans were recalculated with the modified microdosimetric kinetic model (mMKM). Median CTV DVHs were analyzed using dose-at-volume metrics, and intergroup differences were assessed with Mann-Whitney tests (p < 0.05) for both RBE models. Combined dose-LET_d (DL) effects were explored using cumulative DL-volume histograms (DLVHs). Discriminative DL regions were identified via ROC analysis (AUC > 0.75) and clustered using DBSCAN to derive discriminative DL-volume (DLV) metrics, validated by Kaplan-Meier analysis (log-rank p < 0.05). 23 patients (31%) developed LR. At 68.8 Gy(RBE), no DVH differences emerged between groups. At 65.6 Gy(RBE), significant differences in target coverage metrics appeared only after mMKM recalculation, whereas LEM-I metrics showed no separation. DLVH analysis identified discriminative regions at high LET_d (>40 keV/µm). The derived DLV metric corresponded to 64.0 Gy(RBE)-44 keV/µm (90% volume) for LEM-I and 56.3 Gy(RBE)-43 keV/µm (90% volume) for mMKM, both significantly stratifying LC. Outcome differences indicated RBE model-dependent sensitivity to prescription doses. A combined high-dose and -LET_d volume metric provided clinically relevant LC stratification, supporting multi-model RBE evaluation and DL-guided CIRT optimization.