Transthyretin amyloid cardiomyopathy (ATTR-CM) progressively stiffens the myocardium, increasing the risk of major adverse cardiovascular events (MACE). Accurate prediction of these events is critical for personalized patient management. This study was presented at the 2025 European Society of Cardiology Conference.

Data from 372 patients across four Swiss centers were analyzed. This included clinical, demographic, laboratory, medication, electrocardiography, and echocardiographic imaging information. Three machine learning models—CoxNet, Random Survival Forest (RSF), and Gradient Boosting Survival (GBS)—were trained and externally validated. Feature selection and preprocessing ensured high-quality inputs for modeling. RSF and GBS achieved a c-index of 0.73 in external validation, while CoxNet reached 0.72. One-year cumulative AUCs ranged from 0.69 to 0.74.

Integration of multimodal data sources improved risk stratification and enabled precise identification of patients at high risk for MACE. These AI-driven models support clinicians in designing personalized monitoring and treatment strategies for ATTR-CM, with potential to enhance outcomes by targeting interventions to those most at risk.