Geometrical prediction model for late rectal bleeding in radiotherapy of prostate cancer
Abstract
Introduction
Effective prostate radiotherapy requires integrating radiobiological parameters like the α/β ratio with planning factors such as dose gradients, margins, and adaptive strategies. Predicting late rectal bleeding (LRB) remains challenging but essential for optimizing safety. This study presents a preliminary model to predict LRB using only patient imaging data, offering a practical tool to assess toxicity risk daily and support personalized treatment decisions.
Material and methods
We retrospectively analyzed a cohort of 116 prostate cancer patients treated with radiotherapy, divided into standard (2.0 Gy/fraction) and m-Hypo (2.5 Gy/fraction) groups. The toxicity endpoint was late rectal bleeding (LRB); its incidence was characterized and the rectal α/β ratio estimated and used in subsequent analyses. To assess relationships between toxicity and patient/treatment/planning factors, six metrics were defined: one geometric parameter, P₀, representing the rectum–PTV intersection volume, and five DVH parameters. We combined these findings in a sector analysis using linear and logistic models between selected parameters.
Results
Late rectal bleeding (LRB) occurred in 11% of the standard group and 26% of the hypofractionated group. The estimated α/β for the LRB endpoint was 1.5 Gy (95% CI: 0–12 Gy). Dose–volume metrics P₂ and P₄ were significantly associated with binary toxicity in logistic regression. Sector analysis using P₀ as the independent variable identified regions with high toxicity rates (>40%), predominantly within the hypofractionated cohort, as well as regions with negligible toxicity.
Conclusions
A linear relationship between the P₀ parameter and toxicity predictors (P₃, P₄) suggests that volumetric parameters can be integrated into patient-specific risk assessments and treatment planning. Our findings indicate that regions with P₀ < 10 cc (L1, H1) may tolerate variations in fractionation and dose, while regions with higher P₀ (L2, H2) require cautious strategies such as dose reduction, margin optimization, and advanced imaging to minimize toxicity. This framework provides a foundation for personalized radiotherapy optimization, pending further validation.
© 2026 Michał Posiewnik, Paweł Czajkowsk, Dominik Małecki, published by Polish Society of Medical Physics
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.