Abstract 3399: Prediction of cardiac adverse events (AE) in trastuzumab treated breast cancer patients (pts) via a comprehensive genomic and DNA methylation blood based assay

Abstract

Abstract Background: ERBB2 (HER2) targeted therapy is an important therapeutic class used to treat HER2-positive breast cancer (BC) but has been associated with a risk of cardiotoxicity. Identifying pts likely to experience this AE could allow for early intervention and/or a change in treatment plan. Previous studies have demonstrated the ability to detect changes in methylation patterns of pts who had cardiotoxicity from anti-HER2 therapy. Here, we report a retrospective study assessing the feasibility of using the Guardant Infinity platform (genomic and epigenomic NGS analysis) to predict severe cardiac AE in pts treated with trastuzumab. Methods: Samples (n=46) from female pts with BC undergoing treatment with trastuzumab without prior record of heart failure and with a blood sampling <300 days prior to a documented heart failure event defined a positive case cohort. 70 blood samples were matched to the positive case cohort in all above attributes but without record of patient heart failure to comprise a negative cohort. The >20K differentially methylated regions on Guardant Infinity platform may be useful for detection of specific disease attributes such as treatment-associated AE. To explore this, mixture models were first trained using 880 BC samples unrelated to heart failure to distinguish methylation signals coming from tumor versus non-tumor cfDNA. The resulting normalized methylation signals represent the posterior probabilities of the region molecules to be derived from tumor cells. A subset of methylation regions were selected based on significance of association between heart failure status and normalized methylation signals in the positive and negative AE cohorts. A penalized logistic regression model was then trained on the normalized signals of the selected methylation regions to predict whether a patient would have future cardiac AE. Results: A 10-fold cross validation of the penalized logistic regression models trained on normalized methylation signals produced an area under the curve of 0.78 for predicting heart failure status, suggesting feasibility of using blood-based methylation signals to predict cardiac AE on this platform. The differentially methylated regions selected as model predictors show a clear separation in hypermethylation status between positive versus negative cohorts. Conclusion: Our results show the prediction of cardiac AE in BC pts undergoing trastuzumab treatment in advance of documented heart failure using a blood-based assay designed for simultaneous comprehensive genomic and epigenomic profiling. Given the relatively small sample size within this study, further refinement and validation of the classification models is warranted and ongoing. Citation Format: Sean Gordon, Catalin Barbacioru, Daniel Gaile, Denis Tolkunov, Nicole Zhang, Carin Espenschied, Jing Wang, Tara Dinman, Kimberly Banks, Amar Das, Han-Yu Chuang, Helmy Eltoukhy. Prediction of cardiac adverse events (AE) in trastuzumab treated breast cancer patients (pts) via a comprehensive genomic and DNA methylation blood based assay [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 3399.