Using Artificial Intelligence to Identify Patterns and Predictors of Adverse Drug Reactions in Cardiovascular Patients: A Comprehensive Narrative Review Publisher Pubmed



Shahabinejad E ; Soflaei Shahrbabak F ; Gholami Shoa M ; Azhdarimoghaddam A ; Mozafar M ; Yektaee Rad P ; Monajemi S ; Ghanouni S ; Qiu S ; Khorsand Askari M
Source: American Journal of Cardiovascular Drugs Published:2026

Abstract

Artificial intelligence (AI) has emerged as a transformative tool for improving the detection, prediction, and prevention of adverse drug reactions (ADRs) in cardiovascular (CV) medicine, a domain characterized by high drug utilization, complex multimorbidity, and substantial polypharmacy. Traditional pharmacovigilance (PV) systems, particularly spontaneous reporting, remain limited by underreporting, variable data quality, and delayed signal recognition, underscoring the need for computationally enhanced approaches. This narrative review synthesizes current evidence on AI applications across the cardiovascular PV continuum, including safety signal detection, patient-level risk prediction, and text-based surveillance. Machine learning (ML) models, applied to structured datasets such as the Food and Drug Administration (FDA) Adverse Event Reporting System (FAERS), VigiBase, and electronic health records (EHRs), and curated drug–target databases have shown improved discriminatory performance compared with conventional approaches in identifying bleeding risk under direct oral anticoagulants, predicting acute kidney injury in patients with heart failure, and estimating susceptibility to drug-induced QT prolongation, primarily in internal validation settings. Natural language processing (NLP) advances, particularly transformer-based models, further facilitate extraction of drug–event relationships from clinical notes and patient-generated content. Despite these advances, major challenges persist, including data heterogeneity, selective reporting biases, mechanistic ambiguity in CV ADRs, and limited external validation. Ethical considerations, including privacy preservation, algorithmic transparency, bias mitigation, and human oversight, remain critical for responsible implementation. Looking forward, the integration of real-time biosignal streams from wearable devices, multimodal multiomics data, and explainable AI frameworks may further refine individualized risk assessment and facilitate earlier detection of cardiotoxic events in appropriately validated settings. Collectively, AI holds substantial promise for enhancing cardiovascular drug safety, provided its deployment is accompanied by rigorous validation, robust governance, and alignment with clinical workflows. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2026.