Clinical Characteristics and Outcomes in Patients With Heart Failure: Are There Thresholds and Inflection Points in Left Ventricular Ejection Fraction and Thresholds Justifying a Clinical Classification?

BACKGROUND: Recent guidelines proposed a classification for heart failure (HF) on the basis of left ventricular ejection fraction (LVEF), although it remains unclear whether the divisions chosen were biologically rational. Using patients spanning the full range of LVEF, we examined whether there was evidence of LVEF thresholds in patient characteristics or inflection points in clinical outcomes. METHODS: Using patient-level information, we created a merged dataset of 33 699 participants who had been enrolled in 6 randomized controlled HF trials including patients with reduced and preserved ejection fraction. The relationship between the incidence of all-cause death (and specific causes of death) and HF hospitalization, and LVEF, was evaluated using Poisson regression models. RESULTS: As LVEF increased, age, the proportion of women, body mass index, systolic blood pressure, and prevalence of atrial fibrillation and diabetes increased, whereas ischemic pathogenesis, estimated glomerular filtration rate, and NT-proBNP (N-terminal pro-B-type natriuretic peptide) decreased. As LVEF increased >50%, age and the proportion of women continued to increase, and ischemic pathogenesis and NT-proBNP decreased, but other characteristics did not change meaningfully. The incidence of most clinical outcomes (except noncardiovascular death) decreased as LVEF increased, with a LVEF inflection point of around 50% for all-cause death and cardiovascular death, around 40% for pump failure death, and around 35% for HF hospitalization. Higher than those thresholds, there was little further decline in the incidence rate. There was no evidence of a J-shaped relationship between LVEF and death; no evidence of worse outcomes in patients with high-normal ("supranormal") LVEF. Similarly, in a subset of patients with echocardiographic data, there were no structural differences in patients with a high-normal LVEF suggestive of amyloidosis, and NT-proBNP levels were consistent with this conclusion. CONCLUSIONS: In patients with HF, there was a LVEF threshold of around 40% to 50% where the pattern of patient characteristics changed, and event rates began to increase compared with higher LVEF values. Our findings provide evidence to support current upper LVEF thresholds defining HF with mildly reduced ejection fraction on the basis of prognosis. REGISTRATION: URL: https://www. CLINICALTRIALS: gov; Unique identifiers: NCT00634309, NCT00634400, NCT00634712, NCT00095238, NCT01035255, NCT00094302, NCT00853658, and NCT01920711.