Inter‐fraction dose and LET robustness in LET‐optimized proton therapy for pediatric, adolescent and young adult intracranial tumors

This is the authors' abstract. We don't add key points for this paper.

Background Although pediatric, adolescent and young adults (AYA) patients are known to be eligible for proton therapy, the presence of high linear energy transfer (LET) in critical structures, such as the brainstem, could lead to an increased risk of toxicity. To address this issue, dose‐averaged LET (LET d )‐optimized (LO) planning strategies have been proposed as a means to mitigate LET‐related risks while maintaining dose quality. However, the balance between dose distribution quality, LET d optimization, and plan robustness must be evaluated across the entire treatment course. As the treatments here investigated are delivered in multiple fractions, the assessment of the benefits of LET d optimization and its robustness against the anatomical and setup variations during treatment course, is crucial. Purpose This study investigates the inter‐fraction robustness of LO plans compared to standard (STD) dose‐driven plans in pediatric and AYA patients with intracranial tumors. Both dose and LET d distributions were evaluated using re‐evaluation CTs acquired during treatment. Additionally, recalculations with variable relative biological effectiveness (RBE) models were performed to investigate plan sensitivity to biological uncertainties. Methods Twenty patients (prescription dose ≥ 54 Gy) were retrospectively included. For each patient, STD and LO plans were generated and clinically approved by a medical doctor. Both plan types were optimized with robust optimization parameters, including 2 mm setup and 3.5% range uncertainties. Reference plans were subsequently recalculated using variable RBE models, in addition to the fixed RBE = 1.1 assumption. To evaluate robustness, all plans were recalculated on re‐evaluation CTs acquired throughout the treatment course. Dosimetric endpoints included D 98% and D 1% for the clinical target volume (CTV) and D 1% for the brainstem. LET d robustness was assessed by quantifying the brainstem volume receiving > 50 Gy and LET d above thresholds of 4, 3.5, and 3 keV/µm (V50@LET d x). Results LO and STD plans exhibited comparable dose distributions, with no statistically significant differences in CTV coverage or brainstem sparing. In terms of LET, LO plans achieved a significant reduction in V50@LET d x volumes ( p < 0.01). Furthermore, inter‐fraction variation in LET d ‐sensitive brainstem volumes was lower in LO plans, indicating enhanced LET d robustness across the treatment course. As expected, due to the optimized LET d values in the brainstem, plan recalculations with variable RBE models showed a smaller D 1% deviation in LO compared to STD plans when referenced to the fixed RBE = 1.1 model. These results demonstrate that LO not only maintains conventional dose robustness but also reduces LET d ‐related uncertainties. Conclusions LO and STD plans exhibited comparable dose distributions, with no statistically significant differences in CTV coverage/brainstem sparing. LO plans achieved a significant reduction in V50@LET d x ( p < 0.01). Furthermore, inter‐fraction variation in LET d for brainstem was lower in LO plans, indicating enhanced LET d robustness throughout the treatment course. The plan recalculations with variable models showed a minor D 1% deviation in LO versus STD plans, when compared to fixed 1.1 RBE model.