Abstract 6076: Mathematical oncology in the context of early breast cancer detection

Abstract

Abstract Breast cancer (BC) is the most common cancer in women globally, with approximately 94% of patients initially diagnosed with early-stage disease. However, despite the initial lack of detectable metastases and administration of subsequent treatments, 40% of these patients will develop recurrence over their lifetime. The standard screening method for BC is mammography, with a tissue biopsy to confirm diagnosis. However, only ~60% of all cases are currently diagnosed via this pathway, primarily due to the limited (86.9%) sensitivity. To demonstrate the feasibility of a liquid biopsy (LBx) test for early BC detection and assessment, we employed the third-generation, High Definition Single Cell Assay (HDSCA3.0) to detect and characterize rare cells (e.g., circulating tumor cells [CTCs]) and acellular structures (e.g., oncosomes) in patient peripheral blood samples using immunofluorescent (IF) imaging. We have previously investigated this signal in a cohort of early-stage BC patients (n = 74), late-stage BC patients (n = 26), and age- and gender-matched controls (n = 30) and observed statistically significant differences in levels of total rare cells, CTCs, and oncosomes via a manual approach. For this study, we employed a hybrid methodology, that utilizes both automatic and manual techniques, to investigate the reproducibility of the signal in an expanded validation set. Within this modified approach, we employ an outlier detection algorithm to accelerate the rare event identification and curation process (minimizing the manual portion), a separate machine learning model to classify the events based on their IF expressions, and a suite of applications to enumerate and quantify the result. After implementation, we quantified the results of the fully manual and hybrid approaches to determine the signal loss associated with scale up. Utilizing the hybrid approach, we observed similar results as with the fully manual and were able to stratify the three cohorts with high accuracy. Namely, there was an increase in the total rare cell and CTC populations in late-stage BC patients, and an increase in the oncosome population for the early-stage BC patients. To note, these patients included a subset with invasive lobular carcinoma (ILC; n = 19), which manifests in ~5-15% of BC cases and is considered “mammogram silent” due to the biological development of the disease. Additionally, employing the hybrid approach allowed for analyzing patient samples in less time while minimizing overall signal loss. In summary, we have been able to detect reproducible patterns in the enumeration of rare cells and oncosomes with high accuracy. When utilized at the patient-level, these analytes can power prediction models that are capable of stratifying disease from non-disease. The results presented here demonstrate the utility of a LBx test as a companion to screening mammography in the detection of early-stage BC across various histologies. Citation Format: Jeremy Mason, Stephanie N. Shishido, George Courcoubetis, Dean Tessone, Emmett Liljegren, Amin Naghdloo, James Hicks, Jorge J. Nieva, Peter Kuhn. Mathematical oncology in the context of early breast cancer detection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6076.