Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

In International journal of cardiology ; h5-index 68.0

OBJECTIVE : Heart failure with preserved ejection (HFpEF) represents nearly half of all patients with heart failure (HF). The objective of this study was to determine whether patient characteristics identify discrete kinds of HFpEF.

METHODS : Data were collected on 196 patients with HFpEF in a non-hospitalized setting. Clinical and laboratory variables were collected, and 47 candidate variables were examined by the unsupervised clustering strategy partitioning around medoids. The Meta-analysis Global Group in Chronic Heart Failure (MAGGIC) risk score was calculated. Follow-up data on all-cause mortality, cardiovascular mortality, and HF exacerbation, were collected and were not part of the data used to identify subgroups.

RESULTS : Six significantly different groups or clusters were found. There were three groups of women (i) individuals with a low proportion of vascular risk factors (HFpEF1) (ii) individuals with a high proportion of hypertension and diabetes, but lower proportion of kidney disease and diastolic dysfunction (HFpEF3) (iii) older individuals with high rates of atrial fibrillation (AF), chronic kidney disease. They had the worst long-term outcomes (HFpEF4). There were three groups of men (i) individuals with a high proportion of coronary artery disease (CAD), dyslipidemia, higher serum creatinine, and diastolic dysfunction (HFpEF2) (ii) individuals with highest BMI, and high proportion of CAD, obstructive sleep apnea, and poorly controlled diabetes (HFpEF5) (iii) individuals with high rates of AF, elevated BNP, biventricular remodeling (HFpEF6). They had a high cardiovascular mortality.

CONCLUSIONS : HFpEF consists of a heterogenous group of individuals with six distinct clinical subsets that have different long-term outcomes.

Nouraei Hirmand, Rabkin Simon W


Cluster analysis, Heart failure, Preserved ejection fraction, Unsupervised machine learning