https://doi.org/10.1007/s00392-025-02625-4
1IKDT - Institut Kardiale Diagnostik und Therapie GmbH Berlin, Deutschland; 2Howto Health GmbH Berlin, Deutschland; 3Deutsches Herzzentrum der Charite (DHZC) Klinik für Kardiologie, Angiologie und Intensivmedizin | CBF Berlin, Deutschland; 4Universitätsmedizin der Johannes Gutenberg-Universität Mainz Zentrum für Kardiologie Mainz, Deutschland; 5Universitätsklinikum Düsseldorf Klinik für Kardiologie, Pneumologie und Angiologie Düsseldorf, Deutschland
Background: The analysis of endomyocardial biopsies (EMB) is a prerequisite for a definitive diagnosis in patients with unexplained heart failure (HF). The use of machine learning (ML) methods is of great importance to identify high-risk patients and initiate therapy. In this study, we present a ML model for risk stratification of parvovirus B19 (B19V) positive patients based on key features from multiparametric EMB analyses.
Methods: We retrospectively enrolled 263 B19V positive patients with unexplained HF (62% men, age 51±15 years) and followed them over a median period of 22 months. Survival and the development of left ventricular ejection fraction were considered as combined endpoints. EMB were analyzed for inflammatory and infectious markers at baseline, and patients' prognosis was assessed using ML methods based on biopsy characteristics.
Results: Detection of intramyocardial inflammation and active B19V infection was associated with a poor clinical outcome (hazard ratio (95% CI) = 3.52 (1.97-5.87), P<0.001). The linear combination of demographic and clinical data with multiparametric EMB markers increased the prognostic accuracy (AUC=0.724) compared to the use of single features (AUC=0.667). The use of gradient boosting ML methods again significantly improved the accuracy of risk prediction (AUC=0.926).
Conclusion: Intramyocardial inflammation and B19V viral activity are risk factors for death and left ventricular dysfunction. Using multiparametric EMB data, we present a ML-based prognostic tool that can determine the clinical outcome of patients with a high degree of accuracy. This is of great clinical relevance and helpful to make appropriate treatment decisions.