ESTRO 2024 - Abstract Book

S694

Clinical - Breast

ESTRO 2024

Ruysscher 13 , Ana Vega 14 , Liv Veldeman 15 , Adam Webb 9 , Catharine M L West 1 , Eliana Vasquez-Osorio 1 , Marianne C Aznar 1 1 The University of Manchester, Radiotherapy Related Research, Manchester, United Kingdom. 2 Mahidol University, Radiology, Bangkok, Thailand. 3 Montepellier University, Radiation Oncology, Montpellier, France. 4 German Cancer Research Center (DKFZ), Cancer Epidemiology, Heidelburg, Germany. 5 Fondazione IRCCS Isituto Nazionale dei Tumori, fRadiation Oncology, Milan, Italy. 6 Vall d'Hebron Barcelona, Hereditary Cancer Geneticss Group, Barcelona, Spain. 7 Leuvens Kanker Institut, Radiotherapy-oncology, Leuven, Belgium. 8 University of Heidelberg, Radiation Oncology, Mannheim, Germany. 9 University of Leicester, Leicester Cancer Research Centre, Leicester, United Kingdom. 10 Fondazione IRCCS Instituto Nazionale dei Tumori, Prostate Cancer Program, Milan, Italy. 11 Vall d’Hebron Hospital Universitari, Radiation Oncology, Barcelona, Spain. 12 Icahn School of Medicine at Mount Sinai, Radiation Oncology, New York, USA. 13 Maastricht University Medical Center, Radiation Oncology, Maastricht, Netherlands. 14 Universitario de Santiago, Fundacion Publica Galega Medicina Xenomica, Santiago de Compostela, Spain. 15 Ghent University Hospital, Radiation Oncology, Ghent, Belgium

Purpose/Objective:

Breast radiotherapy (RT) after breast-conserving surgery (BCS) is a standard treatment for early-stage breast cancer. However, some patients experience severe toxicity during or following treatment. For example, pain in the breast, arm, and shoulder regions could potentially delay treatment, limit the patients' ability to work and impact their quality of life. This study aims to discover regions where dose deposition is related to acute and persistent breast, arm and shoulder pain after breast RT using a voxel-based analysis technique known as image-based data mining (IBDM) technique.

Material/Methods:

Data from 923 patients with breast cancer who received RT after BCS from the REQUITE study (www.requite.eu) were included. All patients were treated supine and RT plan data were available. Pain severity was scored on a four point scale both by clinicians and by patients (CRO and PRO) at four time points: before RT (defined as a baseline), immediately post-RT, at one year post-RT, and at two years post-RT following the Common Terminology Criteria for Adverse Events (CTCAE) v4.0 and questionnaire for quality of life in breast cancer patients developed by European Organisation for Research and Treatment of Cancer (EORTC) BR23 guidelines. To characterise the pain trajectory, the area under the pain curve (pAUC) for each patient was calculated using the Simpson’s rule, starting at baseline, finishing at a specific time point and including all intermediate pain reported outcomes. Combined severity in CRO and PRO was considered for breast pain, whereas severity from PRO was used to investigate the pain in arm and shoulder. IBDM was applied to the study cohort (Figure 1). To increase statistical power, we mirrored the data from patients treated on the left breast, thereby bringing all ipsilateral breasts to the right side for the complete cohort. Internal thoracic anatomy was excluded from the analysis. The dose distributions were then converted to equivalent dose in 2 Gy-fraction (EQD 2 ) using the Linear Quadratic model (α/β =1.7 Gy). Then, the 3D EQD 2 distributions were spatially normalised to a single reference patient by deformable image registration (DIR) using NiftyReg. The normalised correlation coefficient (NCC) was used to flag DIR failures. Next, voxel-wise Spearman correlation between the mapped EQD 2 and pAUC post-RT, at one- and two-year time points were calculated, and permutation testing (n=1000) was used to identify the significant region related to breast, arm and shoulder pain (at p <0.05). Relevant organs at risk (OARs) in the chest region were segmented by an expert clinician on the reference patient, including breast, muscle, and lymph node levels. To confirm the added value of the identified sub-regions, multi-variables ordinal logistic regression model including and excluding mean- and maximum-EQD 2 within the