S2446
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
15fractions) at our institution. In contrast to the training set[1] current validation cohort, being temporally consecutive, included patients treated using more advanced irradiation techniques such as TomoDirect and VMAT (n=395/683), reflecting the technological evolution in radiotherapy practice over time.Clinical and technical data, along with DVHs of the skin—defined as a 5mm inward expansion from the body contour—were extracted. Reporting at least one G2+ RTOG event among fibrosis, atrophy, telangiectasia, pain (FATP) within 66 months (minimum follow-up: 42 months) was considered as the end- point. A multivariable logistic regression model integrating skin DVH metrics and clinical variables was previously developed on a cohort of 1066 patients treated with 3DCRT tangential field technique in the period 2009-2017. The model included two clinical features (aromatase inhibitor and cosmesis), and two dosimetric variables (skin V20Gy and skin V42Gy). A model including only dosimetry variables (V20Gy/V42Gy) was also developed. The performances of both models onto the new cohort were tested through logistic regression and calibration plots. Results: The incidence of FATP was 2.1% (14 patients), compared to 4% in the training set (p=0.024). The incidences for 3DCRT and VMAT+TomoDirect were 9/299 and 5/395 respectivey(p=0.10); if pooling the techniques in “static” (3DCRT+TomoDirect) against arc (VMAT), 11/400 vs 3/283 events were seen(OR=2.64,CI95%=0.73-9.55,p=0.12). Models demonstrated good agreement between predicted and observed toxicity, both for the dosimetric model (calibration slope=0.53,R ² =0.829,Hosmer–Lemeshow p=0.99) and for the complete model (calibration slope=0.49,R ² =0.844,Hosmer–Lemeshow p=0.99). In Figure1 the resulting rates of FATP against the predicted ones are shown for the dosimetry-only model. The mean skin DVH of patients with and without FATP are shown in Figure2.The resulting AUC improved from 0.62±0.09 for the dosimetric model to 0.84±0.08 for the complete model.
Conclusion: The introduction of arc techniques is associated to a reduction of FATP, although not yet statistically significant. An NTCP model for predicting moderate-to- severe FATP after moderately hypofractionated breast radiotherapy was successfully validated over time, including patients treated with advanced delivery
techniques. References:
[1] Cicchetti et al. Doi: 10.1016/j.radonc.2024.110183 Keywords: model validation, late toxicity, breast cancer
Mini-Oral 2737 NTCP models for severe radiation-induced lymphopenia in lung cancer radiotherapy: a blood dose-based analysis of IMRT and IMPT Shuting Wang 1,2 , Xiaoying Fan 1,2 , Yong Yin 2 , Tianyuan Dai 2 1 Department of Graduate, Shandong First Medical University, Shandong Academy of Medical Sciences, Jinan, China. 2 Department of Radiation Oncology Physics and Technology, Shandong Cancer Hospital and Institute, Shandong First Medical University and Shandong Academy of Medical Sciences, Jinan, China Purpose/Objective: Severe radiation-induced lymphopenia (SRIL) is a poor prognostic factor in lung cancer[1]. This study aimed to develop and validate normal tissue complication probability (NTCP) models for SRIL based on hematologic dose in patients receiving intensity- modulated radiotherapy (IMRT) and intensity- modulated proton therapy (IMPT). Material/Methods: We retrospectively analyzed 131 lung cancer patients treated with curative-intent radiotherapy (94 IMRT, 37 IMPT) between 2022 and 2025. Whole-body blood dose-volume histograms (DVHs) were calculated using the HEDOS framework [2]. The Lyman-Kutcher- Burman (LKB) NTCP model was adopted, with parameters optimized via maximum likelihood estimation [3]. Model robustness was assessed using bootstrap resampling with 1000 iterations. Statistical
Made with FlippingBook - Share PDF online