Clinical predictors of immune-related cardiovascular adverse events: updated analysis from a real-world cohort treated with immune checkpoint inhibitors

Abstract Introduction Immune checkpoint inhibitors (ICIs) have reshaped modern oncology. However, their use is associated with immune-related cardiovascular adverse events (ir-CVEs), ranging from isolated biomarker elevation to myocarditis, pericarditis, arrhythmias, heart failure and acute coronary syndromes. This report provides an updated analysis of our ongoing real-world study, aimed at better defining the incidence and clinical predictors of ir-CVEs. Purpose The primary objective of this study was to identify clinical predictors of ir-CVEs at baseline. Secondary objectives included estimating the incidence of ir-CVEs in a real-world population and assessing the prognostic value of baseline risk stratification according to our definition of high-risk patients. Materials and methods This updated analysis includes 250 patients treated with ICIs from January 2020 to April 2025. Baseline and follow-up cardiovascular assessments included clinical examination, ECG, echocardiography, high-sensitivity troponin I and NT-proBNP. Regularised generalized linear model (glmnet) was implemented as a predictive machine learning approach to avoid overfitting and improve performance. The cohort was split into a training (n=175) and a testing (n=75) group. After hyperparameter optimization (penalty 1-10, mixture 0), the model achieved strong performance (accuracy 84%, precision 89%, F-measure 0.91, sensitivity 94%, specificity 0% due to sample size). Classical logistic regression did not identify significant predictors. Results In this updated cohort, 25 patients (10%) experienced ir-CVEs. Most events occurred at first ICI cycle, although some occurred later (mean 5 cycles). We observed: 10 asymptomatic biomarker elevations, 4 myocarditis, 2 pericarditis, 4 acute heart failure, 1 atrial fibrillation and 4 acute coronary syndromes. Median follow-up was 2 years. No ir-CVE–related deaths occurred. This update confirms pre-existing dyslipidemia (7 of 11 patients were on treatment), previous cardiotoxic chemotherapy, and treatment with anti-PD-L1 agents as independent predictors of ir-CVEs. Baseline biomarkers levels, history of cardiovascular disease and intermediate-to-high baseline cardiovascular risk did not correlate with ir-CVEs. Baseline risk assessment based on literature continued to reliably identify patients at higher risk. Our findings also support the use of Bonaca criterias to identify potential cases of ir-myocarditis. Conclusions This updated analysis reinforces the need for early cardiovascular evaluation and monitoring in patients treated with ICIs, particularly those classified as high-risk. Although the incidence of ir-CVEs remains low, these real-world findings support ongoing refinement of risk-prediction tools. Future updates will integrate a larger cohort and more advanced modelling techniques to enhance preventive strategies in clinical practice.