Contribution of Cardiovascular Reserve to Prognostic Categories of Heart Failure With Preserved Ejection Fraction: A Classification Based on Machine Learning

The authors used cluster analysis of data from cardiovascular domains associated with exercise intolerance to help define prognostic phenotypes of patients with heart failure with preserved ejection fraction (HFpEF).Resting and postexercise echocardiography was performed in 177 patients with HFpEF and 51 asymptomatic control subjects sharing a common clinical profile. Patterns of features that determine exercise capacity were sought from automated hierarchical clustering of left ventricular (LV) diastolic and systolic function, left atrial function, right ventricular function, ventricular-arterial coupling, chronotropic reserve and myocardial fibrosis.Automated clustering separated a distinct subgroup characterized by a relatively isolated impairment of LV systolic reserve. The clinical factors identified by this process were used to define two phenotypes of patients with symptomatic HFpEF: those with reduced chronotropic and/or diastolic reserve (abnormal CR/DR; n = 137) and those with preserved heart rate reserve and exertional E/e' ratio 14 (normal CR/DR; n = 40). Change in global LV strain rate from rest to exercise was similar in patients with abnormal CR/DR (0.16 ± 0.18 secDiminished LV systolic reserve may represent the major identifiable cardiac functional abnormality associated with exercise intolerance in some patients with HFpEF. Despite significant functional limitation, these patients are characterized by a better prognosis than subjects with HFpEF with more physiologic abnormalities.