Baseline proteomic biomarkers for predicting chemotherapy-induced cardiotoxicity in breast cancer patients

Abstract Background Currently, there are no reliable biomarkers to identify patients at high risk of treatment-related cardiotoxicity before initiation of chemotherapy. The aim of this study was to identify biomarkers for predicting asymptomatic decrease in left ventricular ejection fraction (LVEF) after doxorubicin (DOX)-based chemotherapy. Methods A total of 83 breast cancer patients were treated with DOX-based chemotherapy. LVEF was measured before initiation and after completion of therapy. Thirty-nine patients were randomly assigned to the biomarker discovery cohort in which 9 had asymptomatic LVEF reduction (> 10% absolute reduction) after completing the therapy. The remaining 44 patients were grouped into the biomarker validation cohort in which 10 experienced asymptomatic LVEF reduction. Baseline plasma samples from the discovery cohort were analyzed by the SOMAscan proteomic assays for biomarker discovery. Biomarkers were verified using Olink proteomic assays in the validation cohort. Predictive models were generated using partial least squares discriminant analysis. Results SOMAscan analysis for the biomarker discovery cohort revealed that baseline plasma levels of biglycan, carbonic anhydrase 6, cadherin-5, CD109, and thrombospondin-4 were higher, and cystatin-F was lower in the patients with asymptomatic LVEF reduction compared to those without. The levels of these proteins were verified in the validation cohort and associated with increased odds of asymptomatic LVEF reduction. Among the individual protein biomarkers, thrombospondin-4 achieved the highest accuracy of prediction for asymptomatic LVEF reduction (79.5% [95% CI, 64.7%-90.2%]). A predictive model using these 6 proteins, in combination with baseline LVEF, was able to predict asymptomatic LVEF reduction with a sensitivity of 80% (95% CI, 44.4%-97.5%), a specificity of 88% (95% CI, 72.5%-96.7%), and overall accuracy of 86% (95% CI, 72.6%-94.8%). Conclusions These biomarkers and the model are potential new tools for the prediction of DOX-based treatment-related cardiotoxicity and the stratification of breast cancer patients before the initiation of chemotherapy. However, large-scale multi-center validation studies are required.