Circulating miRNAs as Predictive Biomarkers of Cardiotoxicity in HER2-Positive Breast Cancer: A Prospective Clinical Study

Individuals with HER2-positive breast cancer treated with chemotherapy and targeted agents, such as anthracyclines and trastuzumab, are at heightened risk of developing cardiotoxic effects that may progress to long-term cardiovascular complications. Early identification of predictive biomarkers is therefore critical for timely clinical intervention. Circulating microRNAs (miRNAs), which regulate gene expression and play important roles in cardiovascular biology, have recently been proposed as potential indicators of cardiotoxicity. This study investigates alterations in circulating miRNA expression in HER2-positive breast cancer patients receiving chemotherapy and assesses their ability to predict treatment-related cardiotoxicity using machine learning. A total of forty-seven patients were followed for cardiac toxicity assessment at baseline and every 3 months for a period of up to 15 months. Baseline peripheral blood samples were obtained. Expression profiling of 84 microRNAs was conducted using the miRCURY LNA miRNA PCR Panel. Relative expression differences were calculated using the 2−∆∆Ct method. The five most significantly upregulated and five most downregulated miRNAs were subsequently evaluated using univariate logistic regression and receiver operating characteristic (ROC) curve analysis. Five machine learning algorithms (Decision Tree, Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Machine (GBM), and k-Nearest Neighbors (KNN)) were developed to classify cardiotoxicity based on miRNA expression patterns. A total of forty-five miRNAs demonstrated significant differential expression between patients with and without cardiotoxicity. ROC analysis identified hsa-miR-155-5p (AUC 0.76, p = 0.006) and hsa-miR-124-3p (AUC 0.75, p = 0.007) as the most powerful individual predictors. The kNN, SVM, and RF models achieved high predictive performance. The decision tree approach highlighted hsa-miR-17-5p and hsa-miR-185-5p as key discriminative features. Both SVM and RF models revealed additional cardiotoxicity-associated miRNAs (SVM: hsa-miR-143-3p, hsa-miR-133b, hsa-miR-145-5p, hsa-miR-185-5p, hsa-miR-199a-5p; RF: hsa-miR-185-5p, hsa-miR-145-5p, hsa-miR-17-5p, hsa-miR-144-3p, and hsa-miR-133a-3p). Overall evaluation metrics showed that SVM, kNN, and RF outperformed the decision tree model in predictive accuracy. Pathway enrichment analysis of top-ranked miRNAs indicated significant enrichment in apoptosis, p53, MAPK, and focal adhesion signaling pathways, all of which are implicated in chemotherapy-associated cardiac injury and remodeling. Circulating miRNAs may serve as valuable biomarkers for predicting cardiotoxicity in patients with breast cancer. The integration of machine learning approaches can improve miRNA-based risk stratification, supporting personalized surveillance and early cardioprotective interventions.