ESTRO 2024 - Abstract Book

S4995

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

847

Poster Discussion

Voxel-based analysis of symptomatic brainstem necrosis following paediatric proton therapy

Andreas H Handeland 1,2 , Daniel J Indelicato 3 , Grete May Engeseth 1,4 , Kristian S Ytre-Hauge 2 , Helge Henjum 2 , Hardev Grewal 3 , Johannes Tjelta 1,2 , Erlend Lyngholm 2 , Camilla H Stokkevåg 1,2 1 Haukeland University Hospital, Department of Oncology and Medical Physics, Bergen, Norway. 2 University of Bergen, Department of Physics and Technology, Bergen, Norway. 3 University of Florida, Department of Radiation Oncology, Jacksonville, Florida, USA. 4 Oslo Metropolitan University, Department of Life Sciences and Health, Oslo, Norway

Purpose/Objective:

Brainstem necrosis is a serious side effect following high doses of radiotherapy in the treatment of paediatric brain tumours. Whereas proton therapy currently uses a constant relative biological effectiveness (RBE), there is solid preclinical evidence of variability in the RBE across the proton field depending on factors such as linear energy transfer (LET). The LET increases towards the distal edge of the proton beams, possibly increasing the RBE and risk of toxicity in organs close to the tumour volume. Moreover, clinical reports throughout recent years have been inconclusive regarding the correlation between brain/brainstem necrosis and LET. A recent review emphasised the need for appropriate statistical analysis and, in particular, to account for inter-patient variability [1]. To enable the development of clinical risk mitigation methods during proton therapy, it is crucial to improve our understanding and clinical evidence of LET effects in relation to brainstem necrosis. In this study, we therefore used multiple methods of voxel based statistical analyses, herein accounting for inter-patient variability, to investigate proton dose and LET in the brainstem of paediatric brain tumour patients with symptomatic brainstem necrosis. From a cohort of 954 paediatric brain tumour patients treated with passively scattered proton therapy between 2006 2017, 36 patients diagnosed with ependymoma or craniopharyngioma were retrospectively selected for a matched case-control study. 9 patients with symptomatic brainstem necrosis (CTCAE grade 2+) following treatment were each matched to three controls based on age, gender, adjuvant therapy and (RBE1.1-weighted) brainstem dose. T1 weighted contrast-enhanced follow-up MRIs exhibiting radiographic abnormalities corresponding with brainstem necrosis symptom onset were collected for the 9 brainstem necrosis patients and registered to the planning CTs. The regions exhibiting lesions in the follow-up MRIs were subsequently contoured, with both matching and contouring performed in the Eclipse treatment planning system (Varian Medical Systems, Palo Alto, California, USA) (Figure 1a). All voxels in the brainstem of each patient were included in the subsequent analyses, wherein voxels within the lesions in the brainstem necrosis patients were referred to as events while remaining voxels outside the lesions and all voxels of the control patients were considered non-events. Dose and dose-averaged LET (LETd) distributions were calculated using an implementation of the treatment nozzle in the FLUKA Monte Carlo code (Figure 1b-c). LETd considered primary and secondary protons and was converted to LETd-to-water. Dose and LETd were both rescaled to have a mean value of 0 and a standard deviation of 1. Statistical analyses were performed using R (Version 4.1.0, R Foundation for Statistical Computing, Vienna, Austria). Initially, voxel-wise univariate and bivariate logistic regression models were fitted taking both dose and LETd as predictors for the full cohort using the ‘glm’ function. A bivariate mixed-effect logistic regression model was then fitted with the ‘glmer’ function from the lme4 library [2] to account Material/Methods: