Machine Learning--Enhanced Pharmacovigilance Assessment of Cardiac Adverse Events Associated With Minoxidil

Oral minoxidil is increasingly used off-label for the management of androgenetic alopecia; however, concerns regarding its cardiovascular safety persist, particularly with expanding use in real-world settings. This study evaluated cardiac adverse event reporting associated with oral minoxidil by integrating conventional pharmacovigilance methods with machine learning–based risk assessment using the U.S. Food and Drug Administration Adverse Event Reporting System (FAERS). A retrospective analysis was conducted on FAERS reports submitted between January 2014 and the latest available release in 2025. Cardiac adverse events were identified using standardized MedDRA Preferred Terms. Disproportionality analysis was performed using the Reporting Odds Ratio (ROR) and Proportional Reporting Ratio (PRR) to detect population-level safety signals. Subsequently, supervised machine learning models, including Random Forest and Gradient Boosting Machine algorithms, were developed to assess patient-level cardiac risk. Class imbalance was addressed using the Synthetic Minority Over-sampling Technique applied exclusively to the training dataset, with model performance evaluated on an independent test set. Among 61,223 screened reports, 614 met inclusion criteria, of which 276 (45.0%) involved cardiac adverse events. Disproportionality analysis revealed a strong reporting association between oral minoxidil and cardiac adverse events (ROR 314.7; 95% CI 272.8–363.0; PRR 198.6), consistent with substantial disproportionate reporting. In machine learning analyses, the Random Forest model demonstrated superior discriminative performance compared with the Gradient Boosting Machine (AUC-ROC 0.798 vs 0.686), achieving a more favorable balance between sensitivity and specificity. Feature importance analysis identified patient age as the most influential contributor to cardiac adverse event classification. Overall, integrating machine learning with traditional disproportionality analysis enhanced safety signal characterization for oral minoxidil. These findings are hypothesis-generating and highlight the value of patient-level modeling in pharmacovigilance, while underscoring the need for confirmation using controlled pharmacoepidemiologic study designs.