Hospitalization for heart failure in the presence of a normal left ventricular ejection fraction

Results of the New York heart failure registry

Study population

Patients hospitalized with a primary diagnosis of HF and found to have a normal LVEF (≥50%) within seven days of admission were eligible. Both clinical and radiographic evidence were required to establish the diagnosis of HF. Clinical evidence included pulmonary rales and peripheral edema. Radiographic evidence of pulmonary congestion was documented at each institution by an attending radiologist before patient enrollment. Patients hospitalized primarily for acute myocardial infarction or acute coronary syndrome or with a secondary diagnosis of HF were excluded from the registry. However, we retained those patients in whom an acute coronary syndrome was not clinically recognized upon presentation but diagnosed subsequently on the basis of positive cardiac enzymes or new electrocardiographic changes. Patients satisfying eligibility criteria were consecutively enrolled at each medical center. Data logs within each institution and across all institutions were repeatedly screened to ensure that patients were entered into the database only once. Institutional review board approval was obtained at all sites.

Registry design and data collection

The registry database consisted of 153 variables that captured data on demographics, signs and symptoms of HF, presence of coexisting cardiovascular and medical conditions, medications at presentation and at discharge, laboratory tests, procedures, and hospital course. Clinical history and duration of conditions, such as hypertension and diabetes, were based on patient and physician reporting and review of the medical record. Patients were characterized as having coronary artery disease if they had a history of chronic angina, a stress test positive for ischemia, a prior myocardial infarction, coronary stenoses ≥50% by cardiac catheterization, or prior coronary angioplasty or coronary bypass surgery. Cardiovascular or medical conditions that could precipitate hospitalization were prospectively defined. These included: severe hypertension upon presentation (systolic blood pressure [BP] >200 mm Hg); severe valvular disease (regurgitation ≥3+ or stenosis ≤1.0 cm²); uncontrolled supraventricular arrhythmia (atrial fibrillation, atrial flutter, or other supraventricular tachycardia of <1 week duration or associated with a heart rate ≥130 beats/min); acute coronary syndrome secondarily diagnosed by positive cardiac enzymes or new electrocardiographic changes; restrictive, constrictive, or obstructive cardiomyopathies; stroke; severe chronic obstructive pulmonary disease or asthma; pneumonia; sepsis; or serum creatinine >3.0 mg/dl. Severe chronic obstructive pulmonary disease was determined by pulmonary function tests or by the need for systemic steroid therapy.

Data collection and management were carried out on a secure web-based system called the Advanced Data Entry and Protocol Tracking (New England Research Institutes, Watertown, Massachusetts). This system has multiple tools to immediately assess the validity and completeness of data being entered and to provide real-time feedback to individual sites on incomplete, missing, or incorrect data.

Determination of LVEF

The LVEF was obtained by two-dimensional echocardiography or gated radionuclide ventriculography. Only patients with a normal (≥50%) LVEF were eligible. Echocardiograms or nuclear scans were interpreted by experienced observers without knowledge of a patient's inclusion into the registry. The EF by two-dimensional echocardiography was determined by a visual estimate based on an assessment of LV contractile function in multiple echocardiographic views. The accuracy and reproducibility of visual estimates of LVEF have been previously established (13). The EF by gated radionuclide ventriculography was calculated from the LV time-activity curve (14).

Echocardiographic analysis

The LV internal dimensions and wall thicknesses were measured according to the recommendations of the American Society of Echocardiography (15). Wall motion was assessed in the parasternal long- and short-axis views and apical two-, three-, and four-chamber views. Presence and severity of mitral and aortic regurgitation were assessed with color Doppler imaging and image-guided pulsed Doppler studies with semi-quantitative grading (16). Aortic valve area was calculated from the continuity equation and mitral valve area from pressure half-time analysis. Right ventricular systolic pressure was calculated from the measurement of tricuspid regurgitation blood flow velocity using the modified Bernoulli equation and an assumed right atrial pressure of 10 mm Hg. Cardiac mass was calculated using the Penn convention (17). Mass was indexed by height^2.7(m), with partition values of 46.7 g/m^2.7in women and 49.2 g/m^2.7in men to define increased LV mass (18).

Renal function

Glomerular filtration rate (GFR) (ml/min) was calculated using a modified Cockgroft-Gault formula: GFR = (140 − age/creatinine) × (calculated weight/72) × (0.85 for female), where calculated weight (kg) = ideal weight (kg) + 0.4 × (actual weight − ideal weight) and ideal weight = 51.65 + (1.85 × [height (inches)] − 60]) for men and 48.67 + (1.65 × [height (inches) − 60]) for women (19).

Body mass index

Body mass index was calculated as weight (kg)/height²(m). Obesity was defined by a body mass index >30 (20).

Statistical methods

Summary statistics are presented as mean ± standard deviation for continuous data or median and interquartile range where indicated and percentages for categorical data. For comparisons by gender and race, Student's ttest was used to compare the distributions of normally distributed variables and the Wilcoxon rank sum test was used to compare the distributions of non-normally distributed variables determined by the examination of Q-Q plots and the Shapiro-Wilk test for normality p value. The Fisher exact test was used to compare group proportions. Differences between presentation and discharge medical regimen were assessed using McNemar's test for agreement. The signed rank test was used to assess the difference between presentation and discharge systolic and diastolic BPs. Linear regression was used to examine differences in hemodynamic and cardiac measures by race and gender adjusted for covariates that differed by race and gender. The skewed outcome measure GFR was log-transformed before linear regression analysis. Least-squares means are reported to express covariate-adjusted mean values by race and gender. Logistic regression analysis was used to examine differences by race and gender in the use of medications adjusted for covariates such as blood pressure and age.

Only p values of ≤0.01 were considered statistically significant. Analyses were conducted with the Statistical Analysis System Version 8.1 (SAS Institute, Cary, North Carolina) and S-Plus 2000 (S-Plus, Insightful Corp., Seattle, Washington) software.

Clinical characteristics and co-morbid conditions

A total of 619 patients (mean age 71.7 ± 14.1 years, range 22 to 106 years) were enrolled. Four hundred forty-nine patients (72.5%) were women and 170 (27.5%) were men. Women were on average four years older than men: 72.8 ± 14.1 years versus 68.6 ± 13.8 years (p < 0.001). Before hospitalization, 75% of patients had chronic symptoms compatible with New York Heart Association functional class II or III, 14% of patients had chronic symptoms compatible with functional class I, and 11% had chronic symptoms with functional class IV.

The LVEF was measured by echocardiography in 96% of patients and by nuclear imaging in 4% of patients. Measurement of LVEF was obtained within three days of admission in 83% of patients and within seven days in all patients. Women and men had identical mean LVEF: 59.9 ± 7.2% and 59.7 ± 7.4%, respectively.

Patients had multiple co-morbid conditions (Table 1). Hypertension was the most common with a mean duration of 14 years (median 10.5 years; interquartile range 7 to 20 years). The BP upon presentation and prevalence of hypertension were similar in women and men. The second most common co-morbid conditions were diabetes and obesity, with a similar prevalence in the total population (46%) and between women and men. Eighty-eight percent of diabetics were treated with oral hypoglycemic agents and/or insulin and 10% were treated with diet alone. Their mean hemoglobin A1C was 8.2 ± 2.4%. Forty-three percent of patients had coronary artery disease. Atrial fibrillation was present in nearly one-quarter of patients.

Two-dimensional echocardiographic Doppler findings

Echocardiographic and Doppler parameters are summarized in Table 2. Chamber dimensions were normal in women and men. Eighty-three percent of women and 81% of men had increased LV mass with a similar median LV mass index: 66.9 versus 65.7 g/m^2.7(p = 0.92). Moderate pulmonary hypertension, as evidenced by a mean right ventricular systolic BP of 47 ± 17 mm Hg, was observed in 272 patients whose clinical characteristics, including the presence of pulmonary disease, were similar to those of the entire cohort.

Factors precipitating hospitalization

Symptomatic deterioration precipitating hospitalization was related to exacerbation or poor control of coexisting conditions in 51% of patients and to new events in 2%. Except for severe valvular heart disease (combined prevalence of severe regurgitation and stenosis, 13.9%), severe hypertension (prevalence 13%), and acknowledged non-compliance to medication (prevalence 12.8%), the prevalence of all other precipitating factors was below 10% (Fig. 1). Two or more precipitating factors were present in 19.3% of patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cardiovascular or medical conditions and new events precipitating hospitalization. Ranked by prevalence. *Syst. BP = systolic blood pressure upon presentation (emergency department); ^†non-compliance = non-compliance with medication; ^‡MR/AR = mitral regurgitation/aortic regurgitation (severity scale from 1+ to 4+); ^§ACS = acute coronary syndrome; ^∥baseline dialysis or creatinine >3 mg/dl; ^¶Afib/Flutter/SVT= atrial fibrillation or atrial flutter/supraventricular tachycardia—<1 week duration or heart rate ≥130 beats/min; ^#COPD = chronic obstructive pulmonary disease—severe was defined by pulmonary function tests or by the need for systemic steroid therapy; **AS/MS = aortic stenosis/mitral stenosis.

Diagnostic procedures and hospital course

Sixty-three patients (10%) underwent cardiac stress testing, which was positive for ischemia in 28 patients. Fifty-six patients (9%) underwent coronary angiography that revealed one or more stenoses ≥70% in 34 patients. Thirty-one patients (5%) underwent right heart catheterization.

Mean BPs at discharge were within normal values for both women and men: 131/71 and 134/73 mm Hg, respectively. Mean and median lengths of stay were 8.8 and 6.0 days, respectively. In-hospital mortality was 4.2%. Patients who died were significantly older than patients who survived: 78.9 ± 14.3 years versus 71.3 ± 14.1 years (p = 0.01). Gender, the presence of co-morbid conditions, or precipitating factors for hospitalization did not affect length of stay or mortality.

Medications

Medications at presentation and at discharge are shown in Figure 2. Therapeutic adjustments during hospitalization included an absolute increase in the percentage of patients receiving diuretics by 12.8% (p < 0.001), angiotensin-converting enzyme inhibitors by 12.6% (p < 0.001), aldactone by 3.5% (p < 0.001), and beta-adrenergic blockers by 6.1% (p = 0.002). The percentage of patients receiving digoxin, calcium channel blockers, and angiotensin receptor blockers remained unchanged.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Percentage of patients at presentation (open bars), and at discharge (solid bars)receiving diuretics, angiotensin-converting enzyme inhibitors (ACEI), beta-blockers (BB), calcium channel blockers (CCB), digoxin (DIG), angiotensin II receptor blockers (ARB), and aldactone (ALDACT).

Patient profile by race

One hundred eighty-four patients (30%) were black non-Hispanic (B-NH). As observed in the entire cohort, 72% of the B-NH patients were women. The B-NH patients were on average eight years younger than patients of other racial background: 67.1 ± 14.3 versus years 75.1 ± 13.4 years for women (p < 0.001), and 62.9 ± 13.5 years versus 71.1 ± 13.3 years for men (p < 0.001). Both LVEF and the percentage of patients with New York Heart Association functional class II to IV were similar in B-NH patients and in others: 59.0 ± 7.7% versus 60.3 ± 7.0% (p = 0.04) and 80% versus 88% (p = 0.03), respectively. A greater percentage of B-NH patients had a history of hypertension: 86% versus 74% (p < 0.001). Mean systolic and diastolic BPs upon presentation and at discharge were significantly greater in B-NH patients than in others (Fig. 3). However, the percentage of patients with increased LV mass was comparable: 84% versus 82% (p = 0.65). Obesity (body mass index >30) was more prevalent in B-NH women than in other women: 59% versus 41% (p = 0.002).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Mean systolic (hatched bars)and diastolic (solid bars)blood pressures upon presentation (emergency department) and at discharge in the entire cohort of patients (Total) and according to racial background (Black non-Hispanic [B-NH] and Others). Blood pressure (mm Hg) upon presentation: B-NH 172 of 92, Other 155 of 81, Total 160 of 84. Blood pressure (mm Hg) on discharge: B-NH 137 of 75, Other 130 of 71, Total 132 of 72.

Age-corrected GFR, after excluding patients on dialysis, was significantly lower in B-NH patients than in others: 47.7 ± 23.3 ml/min versus 52.3 ± 22.9 ml/min (p = 0.008). The use of long-term hemodialysis was 7.1% in B-NH patients and 3.3% in others (p = 0.05). After adjusting for higher BPs in B-NH patients, the use of angiotensin-converting enzyme inhibitors did not differ significantly in B-NH patients compared with others: 48% versus 35.8% upon presentation (unadjusted p = 0.006, BP adjusted p = 0.045) and 61.2% versus 48.7% at discharge (unadjusted p = 0.005, BP adjusted p = 0.110). Racial background did not affect in-hospital mortality or length of stay.

Diagnosis of diastolic HF

When LVEF is normal, HF is attributed to diastolic dysfunction and labeled diastolic HF. However, because it is difficult to noninvasively assess LV diastolic function, the criteria required to diagnose diastolic HF remain uncertain (21–23). Vasan and Levy (24)sought to strengthen the diagnosis of diastolic HF by requiring a shorter interval between the HF event and documentation of a normal LVEF. Accordingly, the diagnosis of diastolic HF was probable in 84% of our patients. Previous studies dealt with only possible diastolic HF as the interval between presentation and measurement of LV function was either not recorded or could be as long as one year (6–10,25,26). The criteria required to diagnose HF in the present registry were also more stringent than the criteria in prior reports dealing with hospitalized patients, as we required both clinical and radiographic findings to document the diagnosis of HF (6–11,25,26).

The presence of diastolic abnormalities in patients with presumed diastolic HF has recently been questioned (27). In particular, the contribution of arterial stiffening and renal dysfunction, common to the elderly, is increasingly entertained as a mechanism that might help precipitate HFNEF (28,29).

Clinical characteristics

Nearly three-quarters of patients hospitalized for HFNEF in the New York metropolitan area are elderly women with increased LV mass and a history of hypertension. Between the ages of 65 and 74 years, women represent 56% of the population in New York State (30). The modest predominance of women in the elderly population cannot account for the high prevalence of women in the present registry. The mechanisms responsible for the prevalence of HFNEF in women are unclear. Gender affects cardiac remodeling. When confronted with pressure overload, the LV hypertrophies more and dilates less in women than men (31). A reduced rate of myocyte loss in women and transcriptional regulation by estrogens of genes implicated in cardiac hypertrophy may contribute to persistent gender related differences in cardiac remodeling (11,24,31–33).

The prevalence of coronary artery disease in our patients was similar to that of previous reports (7,11,25). In contrast, the prevalence of diabetes in our patients (46%) was greater than the range previously reported in this patient population (23% to 33%) (8,9,11,25,34). The greater prevalence of diabetes is concordant with the high prevalence of obesity noted in our patients, an observation that was also made in the Strong Heart Study (10). The association of diabetes, hypertension, and obesity in 24% of our patients suggests the possibility of a high prevalence of the metabolic syndrome in our population. (10,34–36).

Chronicity of symptoms

Eighty-five percent of our patients had chronic overt symptoms of HF that antedated their hospitalization. Functional intolerance and moderate pulmonary hypertension most likely result from chronic elevation of LV filling pressures (37). The marginal status of our patients at baseline renders the identification of precipitating factors for hospitalization difficult, as these factors are likely to be modest in nature. The present experience is similar to that in patients with HF due to LV systolic dysfunction, where precipitating factors are often not identified but generally thought to be related to medical and dietary non-compliance (38).

Therapy

Despite conflicting views regarding the use of angiotensin-converting enzyme inhibitors in patients with diastolic HF, they were, after diuretics, the medications most often prescribed to our patients (8,39). Angiotensin II receptor blockade with losartan has been shown to substantially reduce LV hypertrophy in hypertensive patients with increased LV mass (40). Definite evidence supporting modulation of the renin-angiotensin system in patients with HF and preserved systolic function (EF >40%) awaits the results of ongoing double-blind, randomized placebo-controlled trials (41,42).

Racial differences

Racial profiling in medical research needs to be interpreted with caution. Despite their significantly younger age, B-NH patients presented with more severe hypertension and worse renal function than others. Our data are consistent with the higher prevalence of end organ damage in B-NH hypertensive patients. Obesity was also particularly prevalent in B-NH women compared with other women. Despite their reported reduced efficacy in African Americans, angiotensin-converting enzyme inhibitors tended to be prescribed at least as often in B-NH patients as in others (43).

Study limitations

The present study has several limitations. There was no echocardiographic core laboratory to ensure uniform quality control of all imaging studies. To address the issue of validity of echocardiographic measurements, we performed a “by center analysis” of LV mass index that incorporates measures of wall thickness and chamber dimensions and is corrected for height. There were some differences in mean height-corrected LV mass by site (p = 0.04), but when examined as a threshold measure (defined as elevated LV mass >46.7 g/m^2.7in women and >49.2 g/m^2.7in men), the prevalence of this condition was found to be similar among all sites (p = 0.38). The measurement of B-type natriuretic peptide was not commercially available when the registry was initiated. Standard load-dependent echocardiographic parameters of LV diastolic dysfunction were not collected and the technical ability to collect load-independent (tissue Doppler) parameters was not available at all sites. Last, we did not mandate that all patients undergo cardiac stress testing to unmask occult coronary disease or require serum in all patients for hemoglobin A1C1C.> determinations; we relied on physician clinical judgment. Thus, the prevalence of coronary artery disease or diabetes mellitus may have been underestimated. We also recognize that when defined according to recent guidelines, the duration and prevalence of hypertension and diabetes in our patient population may be underestimated as the physiologic thresholds for establishing these diagnoses continue to be lowered.

In summary, patients hospitalized in the New York metropolitan area for HFNEF are preponderantly elderly women with a history of hypertension and increased LV mass. The chronic disability of these patients and the frequent inability to identify a factor precipitating hospitalization points to an absence of functional reserve and the need to develop aggressive treatment strategies.