Abstract PS11-07: Predicting the response of locally advanced breast cancer to neoadjuvant therapy using MRI-based mathematical modeling of the I-SPY 2 dataset

Le résumé fourni par la source

Abstract Introduction: Neoadjuvant therapy (NAT) is the standard of care for patients with locally advanced breast cancer; unfortunately, 30-65% of patients have residual disease after completion of NAT.1-2 Accurate and early prediction of individual patient response to NAT is essential for clinicians to make changes to improve patient outcomes. Unlike current population-based methods, mechanism-based mathematical models make patient-specific predictions that capture tumor heterogeneity and longitudinal changes.3. We previously developed a model3 that achieved a concordance correlation coefficient (CCC) of 0.95 for total tumor cellularity (TTC) and a CCC of 0.94 for tumor volume (TV) between the observed and predicted changes in a dataset of 56 triple negative (TN) breast cancer patients.4 Here, we aim to show the generalizability of this approach by applying it to the multi-site, multi-subtype I-SPY 2 trial dataset of patients imaged with standard-of-care magnetic resonance imaging (MRI).5. Methods: I-SPY 2 is a clinical trial for locally advanced breast cancer that acquired dynamic contrast-enhanced (DCE) and diffusion-weighted (DW) MRI scans before (V1), three weeks into (V2), and after completion of (V3) the first course of NAT.5 Our subset of 93 patients includes 42 hormone (estrogen and/or progesterone) receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2-), 22 HER2+, and 29 TN breast cancer patients. Our mathematical model is a partial differential equation characterizing the rate of change in voxel-wise tumor cellularity, NTC(x̄,t), as a function of cell movement constrained by tissue mechanical properties, logistic proliferation, and death due to NAT.3-4 We calculated NTC(x̄,t) at each visit from the apparent diffusion coefficient maps obtained from DW-MRI scans.3 The breast and tumor tissue was segmented by clustering and defined the modeling geometry. Drug concentration was initialized as proportional to contrast agent accumulation obtained from DCE-MRI. We calibrated global drug efficacy and spatially-resolved proliferation rates to the V1 and V2 NTC(x̄,t) data and ran the calibrated model forward to make patient-specific predictions of tumor status at V3.3. Results: At the time of submission, we have completed our analysis on 77 patients, achieving CCC values of 0.94 and 0.91 between the observed and predicted TTC and TV (from V1 to V3), respectively. Fisher’s z-tests between these CCC values and those achieved in the previous single site study4 indicated our model’s predictive accuracy was statistically equivalent in the two data sets. In general, model predictions for TTC at V3 were highly accurate, but the predictions for TV and local cellularity at V3 were less accurate. The model was limited in capturing tumor tissue compression, which often led to overestimated TV and underestimated local cellularity at V3. However, our model was successful in predicting which voxels will have no observable tumor cells at V3. As a result, the median absolute percent difference across all voxels between the observed and predicted change in NTC(x̄,t) had a cohort mean of 11.3%. Discussion: Our results indicate our mechanism-based mathematical model informed by early changes to NAT can accurately predict tumor status after a course of NAT, supporting the feasibility of personalized treatment using only clinically-available MRI data. For instance, after a patient receives three NAT cycles and imaging, we can calibrate a model and make predictions for different available treatment schedules, and a clinician can use the optimal predicted schedule(s) to guide treatment. References [1]. Shien T and Iwata H. Jpn J Clin Oncol. 2020. [2]. Houvenaeghel G et al. Cancer Med. 2024. [3]. Jarrett AM et al. Nat Protoc. 2021. [4]. Wu C et al. Cancer Res. 2022. [5]. Barker AD et al. Clin Pharmacol Ther. 2009. Citation Format: Reshmi Patel, Chengyue Wu, Casey E. Stowers, Rania M. M. Mohammed, Jingfei Ma, Gaiane M. Rauch, Thomas E. Yankeelov. Predicting the response of locally advanced breast cancer to neoadjuvant therapy using MRI-based mathematical modeling of the I-SPY 2 dataset [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr PS11-07.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.