Immune-Checkpoint-Inhibitor-Related Cardiovascular Toxicities in Cancer: A Mechanistic Review of Molecular Pathways with AI-Assisted Literature Clustering

Résumé fourni par la source

Since the first approval of CTLA-4 blockade for melanoma, immune checkpoint inhibitors (ICIs) have expanded into a major class of cancer therapy, with more than 100 FDA-approved oncological indications across metastatic and earlier-stage disease settings, including use as monotherapy and in combination regimens. Preclinical research has largely focused on myocarditis and atherosclerosis, but a wider set of phenotypes, such as non-inflammatory left ventricular dysfunction (NILVD), arrhythmias, and vasculitis, can be observed, and they are rarely connected within a single mechanistic model. We aim to build a systems-oriented, mechanistic framework of the most widely studied biological processes; it will link the main checkpoint pathways to relevant cardiac and vascular cell types, molecular pathways, immune synapses, and candidate biomarkers. We searched PubMed, Scopus, and Web of Science using combinations of terms for immune checkpoint inhibition and cardiovascular-immune-related adverse events that provide mechanistic insight into cardiac-immune-related adverse reactions (irAEs). An AI-assisted semantic clustering approach was used only to organize the included literature. The integrated framework identifies PD-1/PD-L1 as the dominant mechanistic hub linking T-cell activation, endothelial recruitment, myocardial injury, and vascular inflammation. Across phenotypes, a shared immune core involving checkpoint pathways, cytokine signaling, and leukocyte trafficking coexists with phenotype-restricted mediators that may bias injury toward myocarditis, vascular inflammation, conduction-system disease, or NILVD. KEGG analyses support the enrichment of T-cell receptor signaling, Th17 differentiation, JAK-STAT signaling, cytokine-cytokine receptor interaction, and lipid and atherosclerosis pathways. Candidate biomarkers emerging from the reviewed literature include troponin, IL-6, CXCL9/CXCL10/CXCL13, S100A family proteins, ROCK2, HLA-linked susceptibility signals, and T-cell receptor clonality markers. The AI-assisted clustering broadly recapitulated the expert-defined thematic structure while identifying finer semantic neighborhoods within the literature. This framework provides a support map for further hypotheses about toxicity patterns with current and next-generation checkpoint strategies on the cardiac system, while AI-assisted clustering provides a complementary method for organizing the literature rather than an independent source of biological inference.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.