Phenomapping of patients with heart failure with preserved ejection fraction using machine learning-based...

Area of Science:

• Cardiology
• Biostatistics
• Machine Learning in Medicine

Background:

• Heart failure with preserved ejection fraction (HFpEF) is a heterogeneous condition.
• Identifying distinct HFpEF subgroups is crucial for targeted therapies and improved patient outcomes.
• Previous attempts to stratify HFpEF patients have faced challenges due to data complexity and heterogeneity.

Purpose of the Study:

• To apply unsupervised clustering analysis to a high-dimensional, mixed-data cohort of HFpEF patients.
• To identify distinct phenotypic subgroups within the HFpEF population.
• To evaluate the differential clinical characteristics and long-term outcomes associated with identified HFpEF phenogroups.

Main Methods:

• Utilized penalized finite mixture model-based clustering on 61 phenotypic variables from 654 Treatment of Preserved Cardiac Function Heart Failure with an Aldosterone Antagonist (TOPCAT) participants.
• Characterized three distinct HFpEF phenogroups based on co-morbidities, natriuretic peptides, and left ventricular structure/function.
• Validated findings in an internal TOPCAT cohort (n=1113) and an external RELAX trial cohort (n=198) using adjusted Cox models.

Main Results:

• Phenogroup 1 exhibited a high burden of co-morbidities and cardiac abnormalities, associated with significantly higher risks of adverse events, mortality, and HF hospitalization.
• Phenogroup 2 showed lower co-morbidities but higher diastolic dysfunction, linked to increased HF hospitalization risk but lower atherosclerotic event risk.
• Phenogroup 3 presented with lower natriuretic peptides, intermediate co-morbidities, and favorable diastolic function, serving as the reference group for lower adverse outcomes.

Conclusions:

• Machine learning-based cluster analysis effectively identifies HFpEF phenogroups with distinct clinical profiles.
• These identified phenogroups demonstrate significantly different long-term clinical outcomes, highlighting HFpEF heterogeneity.
• The findings support the potential for personalized medicine approaches in managing HFpEF based on identified patient subgroups.