ESTRO 2024 - Abstract Book

S5074

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

[5] Ravanelli (2018) AJNR Am J Neuroradiol

1677

Poster Discussion

First late rectal bleeding biological NTCP model for multiple fractionation schedules

Christian AM Jongen 1 , Ben JM Heijmen 1 , Wilco Schillemans 1 , Andras Zolnay 1 , Marnix G Witte 2 , Floris J Pos 2 , Ben Vanneste 3,4 , Ludwig J Dubois 5 , David van Klaveren 6 , Luca Incrocci 1 , Wilma D Heemsbergen 1 1 Erasmus MC Cancer Institute, University Medical Center Rotterdam, Department of Radiotherapy, Rotterdam, Netherlands. 2 The Netherlands Cancer Institute, Radiation Oncology, Amsterdam, Netherlands. 3 Ghent University Hospital, Department of Radiation Oncology, Ghent, Belgium. 4 GROW - Research School for Oncology and Reproduction, Maastricht University Medical Center, Department of Radiation Oncology (MAASTRO), Maastricht, Netherlands. 5 GROW - Research School for Oncology and Reproduction, Maastricht University, Department of Precision Medicine, Maastricht, Netherlands. 6 Erasmus MC, University Medical Center Rotterdam, Department of Public Health, Rotterdam, Netherlands

Purpose/Objective:

With the current large variations in radiotherapy fractionation schedules for prostate cancer, normal tissue complication probability (NTCP) models that apply to multiple schedules are needed. The objective of this study is to develop the first biological NTCP model for grade ≥2 late rectal bleeding as a function of dose parameters expressed in biological effective dose (BED).

Material/Methods:

The development dataset consisted of prostate cancer patients (n=656) previously randomized to conventional (CF, 39 x 2 Gy) or hypofractionated (HF, 19 x 3.4 Gy) radiotherapy in the Dutch multicenter phase III HYPRO trial. Grade ≥2 late rectal bleeding was scored based on clinician reports and patient questionnaires. Rectal dose distributions were converted to BED assuming α/β = 3 Gy [1]. Candidate predictors were obtained from literature. Logistic regression models were fitted to the data with backward selection. We developed five separate models, each with one of the following dose parameters as candidate predictor: Equivalent uniform dose (EUD) with n=0.1, EUD with n=0.2, the relative volume receiving ≥111.88 Gy in BED (V111.88, the equivalent of physical V70 for CF), minimum dose to the hottest 0.1 cm 3 (D 0.1cm3 ) or 2 cm 3 (D 2cm3 ). The values for n were chosen from the range of n’s reported in literature [2], D 0.1cm3 and D 2cm3 were chosen because they are reported to be predictive for bleeding after EBRT and brachytherapy for cervical cancer [3]. Previous abdominal surgery was included in every model as clinical candidate predictor. After fitting a model, the model coefficients were shrunk with a bootstrap-based uniform shrinkage factor to reduce overfitting [4]. The measures of performance included Area Under the Curve (AUC), calibration slope and Integrated Calibration Index (ICI) [5]. For internal validation we used 100 repetitions of 5-fold cross validations.