Phenotyping in heart failure with reduced ejection fraction - A cluster analysis

Abstract Introduction Heart failure with reduced ejection fraction (HFrEF) is a heterogeneous syndrome with high morbidity and mortality. Identifying clinical phenotypes can enhance personalized care, particularly in subgroups with distinct risk profiles. Purpose This study aimed to identify and validate distinct phenotypes of HFrEF using machine learning and cluster analysis and evaluate their association with clinical outcomes, including mortality, hospitalizations, and emergency department visits. Methods A retrospective cohort study analyzed 10,376 ambulatory patients with HFrEF from the InCor-CHF registry, a tertiary cardiology center, between January 2013 and December 2020. Patients were divided into derivation (80%) and validation (20%) cohorts for phenotypic clustering. Clinical phenotypes were derived through cluster analysis based on demographic, clinical, and laboratory variables. The primary outcome was all-cause mortality, while secondary outcomes included total hospitalizations and emergency department visits. Survival analyses were conducted using Kaplan-Meier and Cox proportional hazards models. Results Five phenotypes were identified with distinct clinical characteristics and outcomes. The Chagasic Phenotype, strongly associated with Chagas disease, had the highest mortality and hospitalization rates (HR for mortality: 1.34; 95% CI: 1.17–1.55; p 0.001). The Hypertensive Phenotype, characterized by hypertension and relatively preserved renal function, exhibited lower mortality risk (HR: 0.45; 95% CI: 0.36–0.57; p 0.001). The Cardiorenal Phenotype, with significant ischemic cardiomyopathy and renal impairment, had an increased risk of adverse outcomes (HR: 1.34; 95% CI: 1.17–1.55; p 0.001). The Young Phenotype demonstrated the best prognosis, with the lowest mortality (HR: 0.21; 95% CI: 0.18–0.25; p 0.001) and hospitalization rates (HR: 0.30; 95% CI: 0.26–0.35; p 0.001). The Multimorbid Phenotype, marked by a high burden of comorbidities, showed intermediate outcomes but lower adherence to guideline-directed medical therapy (HR: 0.90; 95% CI: 0.77–1.04; p = 0.2). Conclusion Machine learning-based phenotyping identified distinct HFrEF subgroups with varying prognoses. These findings emphasize the importance of personalized therapeutic strategies, particularly for high-risk phenotypes such as Chagas disease and cardiorenal dysfunction.Characteristics distribution by cluster The all-cause mortality among the phenot