Abstract 4138701: Phenotypic clustering of HFpEF using cardiac magnetic resonance imaging

Background: Improved understanding of the high-risk features associated with HFpEF may facilitate risk stratification. Cardiac magnetic resonance (CMR)-based strain and structural data can define detailed myocardial structure, which can be leveraged to define heterogenous HFpEF phenotypes. This study used phenotypic clustering integrating CMR-based strain and structural data to identify a high-risk HFpEF phenotype. Methods: Retrospective analyses of 48 HFpEF participants, who underwent cardiac MRI and invasive cardiopulmonary exercise testing (median time between the two studies=40 days), was performed. We conducted unsupervised and unbiased K-means clustering analyses using CMR-based strain and myocardial structural variables. We associated two identified clusters with mortality using Cox proportional hazards modeling and constructed Kaplan-Meier survival curves. A p-value was considered significant if < 0.05. Results: We identified a low (n=32) and high-risk (n=16) cluster. Patients in the identified clusters had similar ages and comorbidities. The high-risk cluster had more male participants and hemodynamically had a higher exercise mPAP/CO and PAWP/CO slope, which are metrics of pulmonary vascular disease. The high-risk cluster displayed greater LV mass, worse biventricular longitudinal strain, decreased right ventricular ejection fraction, and greater interventricular septal angle (all P<0.05) as characterized by CMR as compared to the low-risk cluster (Table 1, Figure 1A). Mortality for the high-risk cluster was 5.43 times that for the low-risk cluster at 12 months (95% CI 1.69-17.42, Figure 1B). Conclusions: Using clustering analyses with CMR strain and structural variables, we identified a high-risk HFpEF phenotype with increased LV mass, reduced biventricular strain, and reduced RV ejection fraction.