|Home | About | Journals | Submit | Contact Us | Français|
The contributions of risk factors and disease etiology to heart failure with preserved ejection fraction (HFPEF) vs. heart failure with reduced ejection fraction (HFREF) have not been fully explored.
We examined clinical characteristics and risk factors at time of heart failure (HF) onset, and long-term survival in Framingham Heart Study participants according to left ventricular ejection fraction ≤45% (n=314, 59%) vs. >45% (n=220, 41%) and hierarchical etiologic classification. HF was attributed to coronary heart disease (CHD) in 278 participants (52%), valvular heart disease in 42 (8%), hypertension in 140 (26%), or other/unknown etiologies in 74 (14%). Multivariable predictors of HFPEF (vs. HFREF) included elevated systolic blood pressure (odds ratio [OR]=1.13 per 10 mmHg, 95% CI; 1.04–1.22), atrial fibrillation (OR=4.23, 95% CI; 2.38–7.52), and female sex (OR=2.29, 95% CI; 1.35–3.90). Conversely, prior myocardial infarction (OR=0.32, 95% CI; 0.19–0.53) or left bundle-branch block QRS morphology (OR=0.21, 95% CI; 0.10–0.46) reduced the odds of HFPEF. Long-term prognosis was grim with a median survival of 2.1 years (5-year mortality rate: 74%), and was equally poor in men and women with HFREF or HFPEF.
Among community patients with new-onset HF, there are differences in etiology and time-of-onset clinical characteristics between those with HFPEF versus HFREF. In people with HFREF, mortality is increased when CHD is the underlying etiology. These findings suggest that HF with reduced vs. preserved left ventricular systolic function are partially distinct entities, with potentially different approaches to early detection and prevention.
Individuals with heart failure (HF) are at increased risk for recurrent symptomatic exacerbations resulting in hospitalization or death.1 The traditional categorization of HF has been according to underlying left ventricular (LV) systolic function based on LV ejection fraction (LVEF).2, 3 Although a substantial proportion of HF cases have preserved LV systolic function, few population-based studies have provided a clinical picture of patients before, during, and after the onset of symptomatic HF with preserved (HFPEF) versus reduced ejection fraction (HFREF).4 Thus, the risk factors and underlying cardiac conditions antedating HF onset and their relations to LV systolic function have not been fully elucidated in the community.
Characterization of HF by LVEF and underlying disease etiology may enhance our knowledge of pathogenesis and prognosis, and thereby promote earlier identification of individuals at increased risk for HF. Although potentially important, the implications of etiologic factors on disease presentation and outcomes have not been clearly delineated. Previous work from the Framingham Heart Study demonstrated that HF can be feasibly categorized etiologically as due to: a) coronary heart disease, b) valvular heart disease, c) hypertensive heart disease, and d) other/unknown etiologies using a hierarchical approach.5 However the inter-relations of pre-onset HF risk factors, etiologic factors, LV systolic function, and outcomes have not been explored in the community.
In this study, we examined risk factors, clinical onset characteristics, and outcomes of patients with HF. Our objectives were to examine clinical features before and at the time of HF onset, and to determine their associations with LV systolic function at time of heart failure presentation. We also compared outcomes according to LV functional status and HF etiology. We hypothesized that risk factor profiles and survival differ substantially in those with HFPEF versus HFREF, suggesting that these subtypes of HF have distinct pathophysiologic mechanisms and outcomes.
The design and description of the Framingham Heart Study original cohort and offspring cohort have been reported previously.6, 7 All participants have been under continuous surveillance for HF since the inception of the original and offspring cohorts and have undergone periodic examinations. In the present study, we included participants with incident HF occurring between 1981–2004 with an evaluation of LV systolic function near the time of their initial HF episode. The study protocol was approved by the Institutional Review Board of the Boston University Medical Center, and all participants provided written informed consent.
Each original and offspring cohort participant underwent Framingham Heart Study clinic examinations approximately every 2 and 4 years, respectively, during which health history updates, physical examinations, and blood tests were performed routinely. Onset of HF was confirmed by a three-physician panel using published criteria after examination of all outpatient and hospital records.8 Pre-onset clinical data were obtained from the most recent Heart Study clinic visit prior to HF onset.
Myocardial infarction and acute coronary insufficiency (defined as prolonged ischemic symptoms with new electrocardiographic abnormalities in the absence of biomarker elevations indicative of infarction) were adjudicated by three physicians after examination of medical records. Occurrence of any atrial fibrillation was determined by examining all available electrocardiograms (ECGs) from clinic, hospital and outpatient physician visits prior to (pre-onset) or on the day of HF onset.
Antecedent clinical variables were obtained from the most recent Framingham Heart Study clinic examination prior to HF onset. At each clinic visit, the cardiovascular history, cigarette smoking, and medications were recorded. Blood pressure was physician-measured in duplicate with a mercury column sphygmomanometer, and the values averaged. Participants were examined by a physician for the presence of systolic and/or diastolic heart murmurs. Total and HDL cholesterol were determined at most visits, and diabetes was defined by a fasting glucose ≥126 mg/dL, nonfasting blood glucose ≥200 mg/dL, or use of insulin or oral hypoglycemic medications.
The characteristics at HF onset were abstracted primarily from emergency department records at the time of hospitalization. The first-measured blood pressure, heart rate, respiratory rate, and blood tests were ascertained. All ECGs from the date of HF onset were examined. We determined LVEF from the echocardiogram or radionuclide ventriculogram performed at or near the HF onset date. Assessments of LVEF were eligible if performed after HF onset (e.g., during hospital admission), or within one year before HF onset provided no intervening myocardial infarction occurred. Intervening ST-elevation and non-ST-elevation myocardial infarctions were determined by detailed review of hospital, physician, and laboratory records. The presence of severe aortic or mitral valve disease (stenosis or regurgitation) at the time of HF onset were determined from the hospital echocardiogram.
Each incident HF was classified sequentially by: (a) LVEF, and (b) hierarchical etiological classification. Heart failure was categorized as HFREF (LVEF ≤45%) or HFPEF (LVEF >45%), based on an a priori cutoff value derived from a prior evaluation, which demonstrated a linear increase in mortality risk for ejection fractions ≤45%.9 Heart failure etiology was classified as coronary heart disease (CHD) if prior myocardial infarction or coronary insufficiency were present. Those without CHD were classified as having valvular heart disease (VHD) if there was severe aortic or mitral valve disease on echocardiography using standard criteria.10 The majority of participants with a systolic murmur grade ≥3/6 or any diastolic murmur by physician auscultation were also found to have significant valvular disease on echocardiography and were therefore also classified to have VHD. Heart failure was attributed to hypertension (HTN) if the participant did not have CHD or VHD, but had JNC stage II hypertension (blood pressure ≥160/100 mmHg) at a clinic visit or was taking antihypertensive therapy.11 Those without CHD, VHD, or HTN were classified as “other or unknown.” Therefore, after the classification process described above, HF patients were categorized into four mutually exclusive etiologic groups.
We performed descriptive analyses stratified by: (a) LVEF, and (b) HF etiologic classification. Age/sex-adjusted comparisons across etiologic strata were performed using ANCOVA for continuous covariates, and logistic regression for categorical variables. We examined pre-onset and time-of-onset characteristics after age- and sex-adjustment as predictors of HFPEF vs. HFREF using logistic regression, and multivariable associations using stepwise multiple logistic regression analysis with p<0.10 for selection and <0.05 for covariate inclusion.12 Model calibration was assessed using the Hosmer-and-Lemeshow statistic (p>0.05) and discrimination using the C-statistic.
Survival after HF onset was assessed using Cox proportional hazards regression adjusting for age and sex, stratified by HFREF vs. HFPEF. We further examined survival compared to control participants from the Framingham Heart Study, who were matched on age, sex, Offspring vs. Original cohort, exam cycle, and year of study entry, using Kaplan-Meier analysis. Similarly, we compared the effect of etiology on survival time with the CHD group as the referent category. The assumption of proportional hazards was confirmed. We examined survival of HF patients stratified by HFREF or HFPEF and the four hierarchical disease etiologies, using Kaplan-Meier analysis. Survival time was defined as the time of heart failure onset until the occurrence of death, and censored on the last visit date for those still alive at the end of follow-up. In the analysis of matched controls, survival time began at the time that the participant attained the same age as the corresponding case with heart failure. A p<0.05 was considered statistically significant. All analyses were performed using SAS version 8.2 (Cary, NC).
Among 655 participants with new-onset HF, those without LVEF (n=76) and laboratory test information (n=45) were excluded, leaving 534 participants for the analysis. Baseline characteristics between excluded individuals and the study cohort did not differ substantially. Participants were followed for 3.2±3.6 years, with a total of 1727 person-years of follow-up accrued. HFREF was present in 314 individuals (59%) and 220 had HFPEF (41%). Individuals with HFPEF were more likely to be women (65% vs. 35%, p<0.001); those with HFREF were more likely to be men (60% vs. 40%, p<0.001). The hierarchical approach classified 278 (52%) cases as due to CHD (215 with prior myocardial infarction), 42 (8%) as due to VHD, 140 (26%) as due to HTN, and 74 (14%) as other or unknown etiologies.
Pre-heart failure onset characteristics at the previous Heart Study clinic visit and clinical features at the time of acute HF onset, stratified by LVEF, are shown in Table 1 and Table 2, respectively. Those with HFPEF had significantly higher systolic blood pressure (p=0.04) whereas resting heart rate and serum potassium concentration were higher among those with HFREF at HF onset. Sex and etiological classification were the only pre-onset features that differed between HFPEF and HFREF (Table 1). Those with HFREF were more likely to have CHD (p<0.001), whereas patients with HFPEF were more likely to have VHD or HTN (p=0.05 and p<0.001, respectively). Participants with sinus rhythm and left-bundle branch block (LBBB) QRS morphology at HF onset more frequently had HFREF, whereas atrial fibrillation was more often presented with HFPEF (Table 2, all p<0.001).
Demographic features, pre-onset characteristics, and tests of homogeneity across etiologic classification strata are shown in Table 3. Those with HTN were older than those with CHD (HTN vs. CHD, p=0.03). Patients with VHD or HTN had a greater proportion of women than those with CHD (VHD vs. CHD, p=0.001; HTN vs. CHD, p<0.001). HF cases with CHD were less likely to be women (CHD vs. all others, p<0.001) but more likely to have diabetes (CHD vs. all others, p=0.003) than other etiologic categories.
Characteristics at time of HF onset are shown by etiologic class in Table 4. Systolic blood pressure at HF onset was significantly increased in those with HTN etiology (HTN vs. others, p<0.001). Electrocardiographic features were significantly different across etiologic groups, with sinus rhythm more frequent in those with CHD (CHD vs. others, p<0.001) and atrial fibrillation present more commonly in those with non-CHD etiologies (CHD vs. others, p<0.001). Left bundle branch block (LBBB) morphology was more frequent in those with HTN or VHD etiologies (HTN vs. CHD, p<0.001; VHD vs. CHD, p=0.06).
Women were more likely to present with HFPEF, with an age-adjusted odds ratio (OR) of 2.55 (95% CI, 1.77–3.68; p<0.001). Age at HF onset (sex-adjusted) was not associated with HFREF. Other pre-HF onset characteristics at Heart Study clinic visits relating to HFPEF (vs. HFREF) adjusted for age and sex are shown in Figure 1. The hierarchical etiologic classification was associated with HFPEF vs. HFREF. Specifically, CHD was associated with reduced odds of HFPEF with an age/sex-adjusted OR of 0.38 (95% CI, 0.27–0.55; p<0.001). In contrast, those with HTN were at substantially higher odds of HFPEF with OR 2.13 (95% CI, 1.43–3.23; p<0.001). VHD conferred a trend to HFPEF with OR 1.92 (95% CI, 0.99–3.70; p=0.05).
Characteristics at HF onset that were associated with HFPEF (vs. HFREF) are presented in Figure 2. Atrial fibrillation on the day of HF onset was associated with HFPEF with OR 2.51 (95% CI, 1.66–3.79; p<0.001). Wider QRS duration was associated with reduced odds of HFPEF with an OR of 0.92 per 10 msec increment (95% CI, 0.85–0.99; p=0.025). However, QRS complex morphology was also important; presence of LBBB conferred lower odds of HFPEF (OR 0.29, p<0.001), whereas increased odds of HFPEF was present with right bundle branch block (RBBB: OR 1.80, p=0.05). Higher blood pressure at HF onset was associated with greater odds of HFPEF (p=0.04) whereas higher heart rate (p=0.01), serum potassium (p<0.001), and blood urea nitrogen (p=0.06) were associated with HFREF (Figure 2).
Although female sex conferred a 2.3-fold increased odds of HFPEF, age was not associated with HFPEF (vs. HFREF) after multivariable adjustment (Table 5). Higher SBP and lower heart rates at HF onset were associated with HFPEF, and prior myocardial infarction predicted a lower probability of HFPEF. The presence of LBBB on the electrocardiogram strongly predicted reduced odds of HFPEF (OR 0.21, 95% CI, 0.10–0.46; p<0.001) after multivariable adjustment. The c-statistic for the multivariable model was 0.80, suggesting good discrimination of HFPEF vs. HFREF. There was no evidence of lack of model calibration (Hosmer-and-Lemeshow χ2 statistic 4.37, p=0.82). A model adjusted only for age and sex had considerably lower discriminative ability (c-statistic=0.64) than the multivariable model (Table 5).
Long-term survival after HF onset was grim, with a median survival of 2.1 years (interquartile range: 0.3 to 5.1 years) and mortality rates of 74% and 95% at 5 and 10 years, respectively. Survival after HF onset did not differ between HFREF and HFPEF with age- and sex-adjusted HR 1.14 (95% CI, 0.94–1.37; p=0.18). In sex-specific analysis, survival did not differ between HFREF and HFPEF in men (p=0.14) or women (p=0.59). However, participants with HFREF or HFPEF had significantly worsened survival compared to age-, sex-, and cohort-matched controls (both p<0.001; Figure 3).
Among those with HFPEF, etiology did not differentiate age- and sex-adjusted survival (Figure 4). However, in those with HFREF, CHD etiology conferred greater risk of death than VHD or HTN etiologies (Figure 5), with an age- and sex-adjusted HR of 1.36 (95% CI, 1.02–1.80; p=0.037). Comparing all four etiologies in those with HFREF, there was a tendency for enhanced survival in those with HTN (age- and sex-adjusted HR 0.74; 95% CI, 0.54–1.01; p=0.06) or VHD (adjusted HR 0.72; 95% CI, 0.41–1.28; p=0.27), but no differences for those with other/unknown etiologies (adjusted HR 1.03; 95% CI, 0.71–1.49; p=0.88) compared with the CHD group.
We found that modifiable cardiovascular disease risk factors, including diabetes, smoking, and hypertension commonly preceded the onset of both HFREF and HFPEF, but pre-onset risk factors did not distinguish between HF with preserved or reduced LV systolic function. By examining factors present before and at time of HF onset simultaneously, we found that the major factors conferring increased odds of HFREF were male sex, etiologic classification (i.e. prior myocardial infarction), presence of LBBB, and higher potassium at HF onset. In contrast, HFPEF was more likely in those without CHD etiology, women, those with higher time-of-onset systolic blood pressure, and atrial fibrillation. Thus, the differing features of HFREF and HFPEF were most apparent at the time of acute onset, and were less discernible by risk factors antedating HF. This was exemplified by blood pressure, where ambulatory pre-HF-onset blood pressure was not elevated among those with HTN, but was increased and perhaps “unmasked” at the onset of acute heart failure.
In addition, we found worsened survival in those with HF with LV systolic dysfunction or preserved ejection fraction compared to controls without heart failure. However, survival was comparable in those with HF and preserved or reduced ejection fractions. Our findings seemingly contrast with an earlier smaller report from Framingham of worse prognosis with HFREF.3 The different findings in our current report may be due to: (a) larger sample size in this study (e.g., 534 vs. 73 HF cases)3, (b) recent improvement in relative survival of HFREF13, or (c) LV function assessment that occurred approximately 2 years after HF onset in the prior study, which may have introduced conditional survival bias.3 In contrast, in the present analysis LV function was assessed during the acute HF presentation. For HFREF, those with HTN or VHD fared better than CHD or other/unknown etiologies. However, outcomes for HFPEF were uniform across etiologic categories.
In prior studies of HF prognosis, the presence of LV systolic dysfunction had no effect on survival14 or conferred marginally worse prognosis15 than preserved ejection fractions. Some recommend that HF should be classified pathophysiologically or by underlying disease etiology, rather than LVEF, since the latter is continuously distributed with heart failure symptoms possible at any ejection fraction.5, 16 The effect of ischemic or nonischemic HF etiologies on outcome have not been well studied in the community. In tertiary care referral samples with cardiomyopathy, angiographically-identified coronary disease was associated with higher mortality, however, such patients were highly selected.17, 18 Randomized trial data have reported conflicting results. In SOLVD, an ischemic etiology did not influence survival,19 whereas other trials suggest worse outcome in those with ischemic etiology.20 A Swedish of patients aged ≤65 years, found higher mortality in those with ischemic heart disease,21 but in a population-based study of HF patients of all ages, prior myocardial infarction was not a multivariate predictor of mortality.1
Our study suggests that both LV systolic function and disease etiology are important given the heightened mortality risk in those with HFREF and CHD (Figure 5). This study extends the existing literature,22 by demonstrating that etiologic classification and onset-related clinical factors are predictors of HFPEF vs. HFREF, and by examining long-term survival according to a hierarchical HF classification. Among the predictors, atrial fibrillation at HF onset was associated with HFPEF. Atrial fibrillation may trigger diastolic HF via loss of the atrial contribution to LV filling and shortening of diastolic filling time.23 In addition, we found that QRS duration24, 25 and the morphology of the QRS complex were of importance. In particular, the presence of LBBB was associated with HFREF, whereas RBBB was associated with HFPEF.
The interrelationship between disease etiology and systolic versus diastolic HF, and their joint effects on outcome demonstrated in our study suggest that disease etiology is an important facet of HF providing further understanding of risk factors and outcomes in HF. The association of etiology with systolic or diastolic HF lends further support to the premise that HFREF and HFPEF are pathophysiologically disparate entities. Although we found factors strongly associated with HFREF or HFPEF, we do not propose that our statistical model should replace imaging to determine presence of preserved LV function or systolic dysfunction. However, many HF patients hospitalized in the community do not undergo LV function evaluation,2, 22 and the model may provide early clues to diagnosis of HFPEF vs. HFREF. The hierarchical classification of HF proposed in this study emulates the clinical practice of identifying the potential contributions of ischemic, valvular, and hypertensive heart disease in HF patients, and is potentially useful because of its simplicity, clinical relevance, and applicability in other settings. Thus, classification of HF based on ejection fraction and etiology provides a foundation for future studies on risk and prevention of subtypes of HF.
A limitation of our study is that although we classified HF patients into common etiologic categories, we did not further stratify the “other/unknown” group, which may have been comprised of rarer disease etiologies including familial and idiopathic cardiomyopathies. Alternatively, there may have been unrecognized coronary, valvular or hypertensive disease in these individuals, which were not captured. Evaluation of LV function in some cases was not performed at the time of HF onset; therefore, we did not utilize LV function data if interim myocardial infarction occurred between HF hospitalization and performance of echocardiography. We excluded participants who did not have echocardiography performed; however, the rate of imaging was high compared to other community-based studies and additionally, patients excluded on this basis were comparable to the overall cohort. To enhance specificity, we defined valvular and hypertensive HF etiologies using stringent criteria (e.g., VHD did not include moderate valve disease and HTN did not include stage I hypertension). However, this would not have impacted the finding of worsened survival in those with CHD compared with non-CHD etiologies in HFREF. Finally, our study was comprised of participants who were primarily white with access to medical care – the generalizability of our findings to non-white patients and those with limited health care access could not be determined.
In conclusion, etiologic classification may help reduce syndromic heterogeneity, providing a disease-oriented approach to patients with HF. Blood pressure, prior coronary disease, and clinical factors including electrocardiographic features may provide insights into pathogenesis and underlying cardiac function at time of HF onset. Furthermore, differences in prognosis emerge when HF is stratified by LV ejection fraction and HF etiology. These findings advance understanding of HF with reduced vs. preserved LV systolic function and may provide insight into approaches to earlier detection and prevention.
Supported by NIH contract N01-HL-25195 to The Framingham Heart Study, by a Canadian Institutes of Health Research Team Grant in Cardiovascular Outcomes Research, a Canadian Institutes of Health Research clinician-scientist award (DSL), and a Canada Research Chair in Health Services Research (JVT). Dr. Vasan was supported in part by 2K24 HL04334 (National Heart, Lung, and Blood Institute).
No conflicts of interest to disclose.