Sep 28

QP 05 - Predicting Outcomes in Breast Cancer: From Multi-Omics to AI-Driven Models

1025 - Association of the Radiosensitivity Index with Overall Survival In Radiotherapy-Treated HER2+ and Triple-Negative Breast

08:05am - 08:10am ET

Room 258

Presenter(s)

Shivani Nellore, BS - Cleveland Clinic, Cleveland, OH

S. Nellore¹, N. Joshi¹, D. T. Bergman², P. Chen³, S. A. Eschrich⁴, J. F. Torres-Roca⁵, and J. G. Scott⁶; ¹Cleveland Clinic Lerner College of Medicine, Cleveland, OH, ²Geisel School of Medicine at Dartmouth, Hanover, NH, ³Case Western University, Cleveland, OH, ⁴Department of Biostatistics and Bioinformatics, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, ⁵Department of Radiation Oncology, H. Lee Moffitt Cancer Center and Research Institute, Tampa, FL, ⁶Department of Radiation Oncology, Taussig Cancer Institute, Cleveland Clinic Foundation, Cleveland, OH

Purpose/Objective(s):

While breast cancer therapies have become more individualized based on molecular subtypes and genome assays, standardized radiation therapy (RT) for patients is driven by anatomical factors, tumor size, nodal status, and margins. Few studies have directly compared mortality benefit of RT across different the different subtypes. The radiosensitivity index (RSI) is a clinically validated radiogenomic biomarker with prior multicohort validation in breast cancer. Comprehensive subtype-level evaluation across large, pooled datasets has not been performed. We evaluated if radiation sensitivity index (RSI) can be used to differentiate radiation sensitivity profiles and predict overall survival (OS) for breast cancer with different molecular subtypes.

Materials/Methods:

We conducted a retrospective, pooled analysis for four breast cancer cohorts treated with RT, with 380 TNBC and 163 HER2+/HR- patients. Molecular subtypes were determined using standard immunofluorescence techniques. Transcriptomic profiles were generated using Affymetrix arrays (MCC n=53; CC n=25), RNA-sequencing (TCGA n=173), and Illumina arrays (Metabric n=292), from which RSI was calculated using 10-gene signature. Non-parametric Wilcoxon rank-sum and Kruskal-Wallis was used to analyze RSI distribution across molecular subtype and TNM stage. Cox proportional hazards analysis was performed with RSI as a continuous variable and TNM stage as categorical variable for OS.

Results:

Radiosensitivity showed considerable variation within each molecular subtype, with distinct RSI distributions observed for HER2+ (mean 0.54, SD 0.10) and TNBC (mean 0.50, SD 0.10). This demonstrates substantial intra-subtype biological heterogeneity under uniform RT treatment. There was a significant difference in RSI distribution between the HER2+ and TNBC patient (p=2e-4). Cox regression analysis demonstrated that RSI is significantly correlated with OS in the TNBC (HR=22.2, p=6e-4) subgroup, but not the HER2+ subgroup (HR=2.01, p=0.57). TNM stage was correlated with both molecular subtypes in univariable analysis as well. In multivariable analysis considering TNM, RSI remained significantly associated with outcome in TNBC patients (TNBC: HR=17.7, p=2e-3).

Conclusion:

RSI is a continuous predictor of mortality in TNBC, but not HER2+ cancers. Across four independent cohorts and multiple transcriptomic platforms, RSI demonstrates broad validation and subtype-specific prognostic relevance in RT-treated breast cancer. These results suggest that incorporating RSI into patient selection criteria could optimize RT for individual TNBC patients and potentially meaningfully improve survival outcomes.