Linear energy transfer optimization mitigates biological model discrepancies in carbon ion radiotherapy for pelvic osteosarcoma

In the combined treatment of osteosarcoma with a carbon ion radiotherapy (CIRT) boost, large pelvic tumors are associated with significantly lower tumor control compared to craniofacial localization. This could be linked to lower dose-averaged linear energy transfer (LET_d) values in the target and larger uncertainties related to the model used for relative biological effectiveness (RBE) calculation. This study aimed to bridge the difference between RBE-weighted doses from two clinically used models, the local effect model I (LEM-I) and the modified microdosimetric kinetic model (mMKM), using LET_d optimization.

Fifteen pelvic osteosarcoma CIRT plans, initially optimized with LEM-I, were re-optimized to achieve higher target LET_d while preserving target and organ at risk (OAR) dose objectives. Both dose- and LET-optimized plans were recalculated with mMKM, and dose metrics were compared across the groups. Robustness against setup and range uncertainties and changes in patient anatomy (using control CTs) was investigated.

LET_d optimization resulted in target LET_d escalation without compromising LEM-I target coverage. The LET-optimized group also demonstrated better mMKM dose coverage and a significant decrease (up to 9% of prescription) in the disparity between LEM-I and mMKM target dose. OAR doses remained largely unaffected by model choice or optimization strategy. The robustness loss with LET_d optimization was minor (intra-patient dose variability increased by 0.3-2.3%) compared to the reduction in model-to-model variation.

When adequately balanced with physical uncertainties, LET_d optimization could become a useful strategy for bridging RBE model predictions and increasing the efficacy of CIRT.