Association of plasma biomarkers and interstitial myocardial fibrosis in fontan population: A machine learning approach

Abstract

Background: Myocardial fibrosis is a common final pathway in many myocardial diseases. Circulating biomarkers are recognized in heart failure management as potential treatment targets. The current study sought to evaluate the association between biomarkers of myocardial fibrosis and myocardial fibrosis using machine learning techniques. Methods and Results: The permutation-based feature importance and the cooperation network among important features in random forest algorithm were applied to evaluate the relationship of each biomarker to extracellular volume fraction (ECV). A total of 25 Fontan patients were included in the study with multiple serum biomarkers obtained at the time of CMR. The mean age at the time of assessment was 16.3 ± 6.8 years with the mean duration of Fontan circulation of 13.0 ± 6.3 years. Lateral tunnel was the most common Fontan type (52%), and the majority (72%) had a systemic left ventricle. The mean ECV was 28 ± 5%. Among different types of matrix metalloproteinase (MMP) and tissue inhibitor of metalloproteinase (TIMP), our study found that MMP-10, and TIMP-1 were the most strongly associated with cardiac fibrosis measured by ECV. Conclusions: TIMP-1 and MMP-10 are circulating biomarkers that may represent promising targets for the prevention and treatment of myocardial fibrosis in Fontan circulation.

Keywords

• Biomarker
• Myocardial fibrosis
• Fontan
• Machine learning