Risk prediction of death and heart transplantation in adult patients with myocarditis

Area of Science:

• Cardiology
• Machine Learning in Medicine
• Immunology

Background:

• Myocarditis outcomes vary significantly, necessitating early identification of high-risk patients.
• Predicting adverse events like death or heart transplant (HTx) is crucial for clinical management.
• Machine learning (ML) offers a powerful approach to uncover complex predictive patterns in patient data.

Purpose of the Study:

• To identify early predictors of death and heart transplant (HTx) in adult myocarditis patients.
• To develop and validate a predictive model using ML techniques.
• To enhance prognostic accuracy for myocarditis, aiding clinical decision-making.

Main Methods:

• Retrospective analysis of 938 adult patients with suspected or biopsy-proven myocarditis.
• Development of a random forest (RF) model to predict death/HTx, ranking variable importance.
• Collection of clinical, laboratory, and imaging data at diagnosis and during follow-up.

Main Results:

• The RF model achieved good predictive accuracy (89.2%) for death/HTx.
• Key predictors identified include NYHA class (10%), LVEF (3.6%), clinical presentation (2.5%), and giant cell myocarditis on EMB (2.9%).
• Other factors like anti-endothelial cell and anti-nuclear antibodies also showed predictive value.

Conclusions:

• Advanced NYHA class, reduced LVEF, and heart failure at diagnosis are significant predictors of worse prognosis in myocarditis.
• Giant cell myocarditis identified on endomyocardial biopsy (EMB) is associated with poorer outcomes.
• ML models can effectively predict adverse events in myocarditis, supporting risk stratification.