Key result
AI models show promise for predicting cancer therapy-related cardiotoxicity but still require external validation.
Why the study?
Improved cancer survival has been accompanied by a growing burden of cancer therapy-related cardiovascular toxicity, making early prediction important for preventing adverse events. Although artificial intelligence-based models have been developed to predict cardiotoxicity, a critical review of their performance, limitations, and integration into cardio-oncology pathways was needed.
Does the integration of artificial intelligence-based predictive models and imaging/biomarker surveillance improve the prediction and detection of cancer therapy-related cardiovascular toxicity in patients with cancer receiving anthracyclines and HER2-targeted therapies?
Does the integration of artificial intelligence-based predictive models and imaging/biomarker surveillance improve the prediction and detection of cancer therapy-related cardiovascular toxicity in patients with cancer receiving anthracyclines and HER2-targeted therapies?
This narrative review synthesizes evidence on cardiotoxicity from anthracyclines and HER2-targeted therapies, concluding that while AI-based predictive models may aid risk stratification, they remain limited by retrospective designs and lack of external validation and should complement rather than replace standard clinical surveillance.
Answers are grounded in this paper and cite their sources.
AI models should complement, not replace, standard cardio-oncology surveillance; leaves open the need for prospective external validation.
BACKGROUND: Advances in hematology-oncology have substantially improved survival among patients with cancer. However, longer survival has been accompanied by an increasing burden of cancer therapy-related cardiovascular toxicity (CTR-CVT), which contributes to overall morbidity and mortality among cancer survivors. Cancer therapy-related cardiac dysfunction (CTRCD), one component of CTR-CVT, may manifest as asymptomatic reductions in left ventricular ejection fraction (LVEF) or as severe, symptomatic heart failure, whereas myocarditis, arrhythmias, and pericardial disease represent distinct manifestations of CTR-CVT. Outcomes may be worsened by patient-related risk factors, such as age, obesity, diabetes mellitus, and hypertension. Predicting cardiotoxicity is therefore important for preventing adverse events, and artificial intelligence (AI)-based models have been developed for this purpose. This narrative review focuses primarily on cardiotoxicity related to anthracyclines and human epidermal growth factor receptor 2 (HER2)-targeted therapies, together with the risk factors specific to these two drug classes. Immune checkpoint inhibitors are discussed as a clinically relevant comparison, and fluoropyrimidines, vascular endothelial growth factor inhibitors, anaplastic lymphoma kinase inhibitors, and radiotherapy are summarized more briefly for context. RECENT FINDINGS: This review examines diagnostic and monitoring tools such as echocardiography, cardiac biomarkers, and advanced imaging techniques, highlighting their strengths and limitations. Furthermore, this review critically examines AI-based models developed to predict cardiotoxicity, with particular attention to their performance, limitations, and potential clinical applications. Although these models may improve risk stratification and support earlier identification of high-risk patients, the available evidence remains predominantly retrospective and heterogeneous, with limited external and prospective validation. CONCLUSION: By summarizing recent studies and selected preventive strategies, this review proposes a clinically oriented framework for future evaluation of AI-assisted cardio-oncology pathways while emphasizing that current models should complement, rather than replace, established clinical assessment and surveillance.
No takes yet. Share an insight, caveat, or question.
Attieh et al. (2026) conducted a review in Cancer therapy-related cardiovascular toxicity (CTR-CVT) and cardiac dysfunction (CTRCD). Anthracyclines and HER2-targeted therapies (with AI-based predictive models for cardiotoxicity) was evaluated on Cancer therapy-related cardiac dysfunction (CTRCD) and cardiovascular toxicity. AI-based models show promise for predicting anthracycline and HER2-targeted cardiotoxicity, but require external validation and must complement standard clinical surveillance.
Synapse has enriched 10 related papers. Consider them for comparative context: