Profiling heart failure with preserved or mildly reduced ejection fraction by cluster analysis

Area of Science:

• Cardiology
• Biostatistics
• Internal Medicine

Background:

• Heart failure (HF) patient heterogeneity, especially in preserved or mildly reduced ejection fraction (HFp/mrEF) groups, remains poorly understood.
• Existing knowledge gaps hinder tailored treatment strategies for diverse HF populations.

Purpose of the Study:

• To identify distinct heart failure (HF) subtypes within the HFp/mrEF population.
• To analyze the clinical characteristics and prognostic implications of identified HF phenotypes.

Main Methods:

• Utilized the k-prototypes clustering algorithm on a cohort of 2570 HFp/mrEF patients.
• Applied k-means for quantitative and k-modes for qualitative variables to identify HF phenotypes.

Main Results:

• Identified three distinct HF phenotypes: Cluster A (low comorbidity women), Cluster B (diabetic women with high comorbidity), and Cluster C (men with respiratory comorbidities).
• Observed significant variations in 1-year mortality rates (18% A, 33% B, 26.4% C) and time to death across clusters.
• Found that age, cancer, respiratory failure, and lab parameters predicted mortality.

Conclusions:

• Cluster analysis revealed three unique HFp/mrEF phenotypes with differing clinical profiles and prognoses.
• Women presented two distinct phenotypes: low-risk and high-mortality diabetic groups.
• Men exhibited a more uniform phenotype characterized by a higher prevalence of respiratory disease.