Heart failure with preserved ejection fraction (HFpEF) is a heterogeneous syndrome. We aimed to derive HFpEF phenotype-based groups based on clinical features using machine learning, and to compare clinical characteristics, outcomes and treatment response across the phenogroups.
We applied model-based clustering to 11 clinical and laboratory variables collected in 970 HFpEF patients. An additional 290 HFpEF patients was enrolled as a validation cohort. During 5-year follow-up, all-cause mortality was used as the primary endpoints, and composite endpoints (all-cause mortality or HF hospitalization) were set as the secondary endpoint.
We identified three phenogroups, for which significant differences in the age and gender, the prevalence of concomitant ischaemic heart disease, atrial fibrillation and type 2 diabetes mellitus, the burden of B-type natriuretic peptide level and HF symptoms. Patients with phenogroup 3 had higher all-cause mortality or composite endpoints, whereas patients in phenogroup 1 had less adverse events after 5-year follow-up. Moreover, it was indicated that beta-blockers or angiotensin-converting enzyme inhibitor/angiotensin II receptor blocker (ACEI/ARB) use was associated with a lower risk of all-cause mortality or composite endpoints in phenogroup 3, instead of the other phenogroups. This HFpEF phenogroup classification, including its ability to stratify risk, was successfully replicated in a prospective validation cohort.
Machine-learning based clustering strategy is used to identify three distinct phenogroups of HFpEF that are characterized by significant differences in comorbidity burden, underlying cardiac abnormalities, and long-term prognosis. Beta-blockers or ACEI/ARB therapy is associated with a lower risk of adverse events in specific phenogroup.

Copyright © 2020. Published by Elsevier B.V.

Author