ESTRO 2025 - Abstract Book

S3736

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

1215

Mini-Oral Estimating radiosensitivity and volume effect for acute proctitis after hypofractionated prostate cancer radiotherapy Christian AM Jongen, Ben JM Heijmen, Luca Incrocci, Wilma D Heemsbergen Department of Radiotherapy, Erasmus MC Cancer Institute, University Medical Center Rotterdam, Rotterdam, Netherlands Purpose/Objective: Acute radiation proctitis clinically presents with blood loss, mucus loss, pain, cramps, tenesmus, increased frequency or diarrhea symptoms. Acute proctitis risks unexpectedly increased with the introduction of hypofractionation for prostate cancer [1]. Currently, little is known about NTCP model parameters for acute proctitis. The objective of this study is to estimate radiosensitivity ( α/β ) and the volume effect ( n ) for EUD-based NTCP modelling of acute proctitis. Material/Methods: We evaluated N=321 prostate cancer patients treated within our centre: 105 with 39 x 2 Gy over 8 weeks, 97 with 19 x 3.4 Gy over 6.5 weeks (both groups from a previous randomized trial), and 119 with 20 x 3 Gy over 4 weeks from a recent cohort study. Seven patient-reported symptoms related to proctitis were scored no (0), any (1), or moderate/severe (2) at the end of the fourth treatment week. A summed increase in symptom score of ≥4 over baseline was defined moderate to severe acute proctitis. We converted rectal dose distributions delivered in the first 4 weeks to Biological Effective Dose (BED) for α/β ratios from 2 to 20 Gy in steps of 1 Gy. EUDs were calculated with n values between 0.05 and 1.00 in steps of 0.05. A grid search was performed to find the combination of n and α/β with the highest log-likelihood in logistic regression models. 95% confidence intervals (CI) were based on profile likelihood. Results: The combination of n = 0.30 (CI 0.15-0.45) and α/β = 17 Gy (CI 6->20) resulted in the best data fit (Fig. 1). The differences in log-likelihood for α/β ratios of >10 Gy appeared small compared to the differences for varying n (Fig. 1). Using the optimal n and α/β , the dose-response curves obtained by fitting the data from each fractionation scheme separately were similar to the one fitted to their combined data (Fig. 2a). For in particular a suboptimal n , these curves do not overlap (Fig. 2b-d).