ESTRO 2024 - Abstract Book

S930

Clinical - CNS

ESTRO 2024

Between December 2018 and February 2022, 264 patients started treatment for brain tumours at our Institution with PBS-PBT. Clinical and treatment-related characteristics are reported in Table 1. Of note, 14 patients had reported alopecia at baseline, likely resulting from induction chemotherapy. Of these, 13/14 gradually subsided and resolved in the 12 months following PBT and therefore were not considered in the analysis, but one who persisted, after initial improvement, at 12 months and it was included in the group of patients with permanent RIA. After accounting for baseline alopecia, rates of G1+G2 acute (≤90 days after PBT completion), late (>90 days) and permanent RIA (persisting for >12 months) in our cohort were 61.7%, 24.6% and 14.4%, respectively. Most of the permanent RIA (35/38; 92%) were preceded by acute or late RIA, or both (Figure 1). The performance of the original LKB- and MLR-NTCP models (2) for clinician- rated G2 alopecia was good (0.769≤ AUC≤ 0.860) in our independent cohort, the only exception being the acute RIA with MLR analysis (Figure 2). The acute RIA MLR model performs poorly in our cohort presumably due to differences in the median age between the testing and validation cohorts (56 y vs 17 y, respectively), suggesting this model cannot extrapolate suitable NTCP values for younger age groups and is unsuitable for use on paediatric/ young adults. We do not have sufficient adults (25y+) in our population to validate the acute RIA MLR model on the appropriate age group, therefore we cannot comment on the performance of this in the adult population.

Conclusion:

The validation of the NTCP models for G2 alopecia was overall successful in our independent validation cohort. NTCP models represent the key to maximising the benefit of technological advances in radiotherapy, such as PBS PBT. The implementation of NTCP models for different toxicity endpoints can allow better prediction and prevention of toxicities, for consent purposes and, whenever possible, planning optimisation. A crucial step in integrating prediction models into clinical practice is model validation. However, differences (e.g in the demographics) between the cohorts used for models creation and validation may affect their generalizability. This appears to be especially true for models that are based on dosimetric and clinical data, whereas models that exclusively rely on dosimetric parameters exhibit enhanced robustness and transferability.

Keywords: proton therapy, alopecia, cosmetic outcomes

References:

(1) Freites-Martinez A, Shapiro J, Goldfarb S, Nangia J, Jimenez JJ, Paus R, Lacouture ME. Hair disorders in patients with cancer. J Am Acad Dermatol. 2019 May;80(5):1179-1196. doi: 10.1016/j.jaad.2018.03.055. Epub 2018 Apr 14. PMID: 29660422; PMCID: PMC6186204. (2) Palma G, Taffelli A, Fellin F, D'Avino V, Scartoni D, Tommasino F, Scifoni E, Durante M, Amichetti M, Schwarz M, Amelio D, Cella L. Modelling the risk of radiation induced alopecia in brain tumor patients treated with scanned proton beams. Radiother Oncol. 2020 Mar;144:127-134. doi: 10.1016/j.radonc.2019.11.013. Epub 2019 Dec 2. PMID: 31805517.

2629

Digital Poster