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

Abstract Background Significant knowledge gaps remain regarding the heterogeneity of heart failure (HF) phenotypes, particularly among patients with preserved or mildly reduced left ventricular ejection fraction (HFp/mrEF). Our aim was to identify HF subtypes within the HFp/mrEF population. Methods K-prototypes clustering algorithm was used to identify different HF phenotypes in a cohort of 2570 patients diagnosed with heart failure with mildly reduced ejection fraction or heart failure with preserved left ventricular ejection fraction. This algorithm employs the k-means algorithm for quantitative variables and k-modes for qualitative variables. Results We identified three distinct phenotypic clusters: Cluster A (n = 850, 33.1%), characterized by a predominance of women with low comorbidity burden; Cluster B (n = 830, 32.3%), mainly women with diabetes mellitus and high comorbidity; and Cluster C (n = 890, 34.5%), primarily men with a history of active smoking and respiratory comorbidities. Significant differences were observed in baseline characteristics and 1-year mortality rates across the clusters: 18% for Cluster A, 33% for Cluster B, and 26.4% for Cluster C (P 0.001). Cluster B had the shortest median time to death (90 days), followed by Clusters C (99 days) and A (144 days) (P 0.001). Stratified Cox regression analysis identified age, cancer, respiratory failure, and laboratory parameters as predictors of mortality. Conclusion Cluster analysis identified three distinct phenotypes within the HFp/mrEF population, highlighting significant heterogeneity in clinical profiles and prognostic implications. Women were classified into two distinct phenotypes: low-risk women and diabetic women with high mortality rates, while men had a more uniform profile with a higher prevalence of respiratory disease.