S242
Clinical - Breast
ESTRO 2026
increase in arm circumference of the treated side from baseline, relative to the contralateral side). PRE-ACT-01 also evaluates the influence of the explainable AI- based communication on radiotherapy regimen choices, loco-regional events, quality of life, and the performance of the prediction model. We hereby report the design of the PRE-ACT-01 study. Material/Methods: PRE-ACT-01 leverages an innovative explainable AI model, developed within the European PRE-ACT consortium. The model was trained on a diverse dataset of 4,177 patients from three major European cohorts: REQUITE (ISRCTN98496463), HypoG-01 (NCT03127995) and CANTO (NCT01993498). The model classified as a non-CE-marked class IIa medical device generates a personalized probability (risk) of lymphedema occurrence based on 28 features collected before randomization. Patients with breast cancer cT1-4, cN0-N3, M0 indicated for loco-regional radiotherapy are randomized 1:1. In the experimental arm, physicians utilize a dedicated web application to communicate the AI-predicted risk, empowered by personalized recommendations for lymphedema prevention based on AI prediction explainability, thus fostering shared decision-making with the patient. In the control arm, the predicted risk is blinded to both the patient and physician.
Digital Poster Highlight 1525
PRE-ACT-01: an international, multicenter, randomized study on communicating a personalized AI-based lymphedema risk to breast cancer patients Marie Bergeaud 1 , Tim Rattay 2 , Karolien Verhoeven 3 , Assia Lamrani-Ghaouti 1 , Catherine Gaudin 4 , Guido Bologna 5 , Gabriella Cortellessa 6 , Andre Dekker 3 , Francesca Fracasso 6 , Manuela Joore 7 , Iordanis Koutsopoulos 8 , Yuqin Liang 3 , André Panisson 9 , Alan Perotti 9 , Bram Ramaekers 7 , Cheryl Roumen 10 , Johan van Soest 11 , Hilary Stobart 12 , Christopher J Talbot 13 , Adam Webb 13 , Alexis Bombezin-Domino 14 , Robabeh Ghodssighassemabadi 15 , Stefan Michiels 15 , Sofia Rivera 16 1 UNITRAD, UNICANCER, Paris, France. 2 Leicester Cancer Research Centre, University of Leicester, Leicester, United Kingdom. 3 Department of Radiation Oncology (Maastro), GROW School for Oncology and Reproduction, Maastricht University Medical Centre+, Maastricht, Netherlands. 4 Data and Partnerships Department, UNICANCER, Paris, France. 5 University of Applied Sciences and Arts of Western Switzerland, HES-SO Geneva, Geneva, Switzerland. 6 Consiglio Nazionale delle Ricerche (CNR), ISTC Roma, Rome, Italy. 7 Department of Clinical Epidemiology and Medical Technology Assessment (KEMTA), Maastricht University Medical Centre MUMC+/Care and Public Health Research Institute (CAPHRI), Maastricht University, Maastricht, Netherlands. 8 Department of Informatics, Athens University of Economics and Business, Athens, Greece. 9 CENTAI, CENTAI, Turin, Italy. 10 Department of Applied and Engineering Sciences, Dutch Research Council (NWO), Utrecht, Netherlands. 11 Medical Data Works, Medical Data Works, Maastricht, Netherlands. 12 Independent Cancer Patients’ Voice, Independent Cancer Patients’ Voice, London, United Kingdom. 13 Department of Genetics, Genomics and Cancer Sciences, University of Leicester, Leicester, United Kingdom. 14 TheraPanacea, TheraPanacea, Paris, France. 15 Clinical Research Division, Gustave Roussy, Villejuif, France. 16 Breast Radiotherapy Department, Gustave Roussy, Villejuif, France
Purpose/Objective: PRE-ACT-01 (NCT05701085) is an innovative,
Results: PRE-ACT-01 is currently enrolling 724 patients across 35 centers in three European countries (29 in France, 3 in the Netherlands, and 3 in the UK), using a model with an estimated ROC-AUC of 0.83. The sample size is calculated to provide 80% power to detect a non- inferiority accepting up to a 6% increase in the 2-year lymphedema incidence, with a 5% significance level. The first patient was successfully enrolled in France on October 7, 2025, with recruitment planned over 14 months. The final analysis will be stratified by key
international, multicenter, randomized-controlled pivotal study designed to assess the impact of communicating an individualized, explainable AI- driven risk prediction for arm lymphedema to breast cancer patients indicated for nodal radiotherapy. The primary objective is to demonstrate that sharing this personalized risk information is non-inferior to the current standard of care in terms of the 2-year incidence of arm lymphedema (defined as a ≥ 5%
Made with FlippingBook - Share PDF online