Heart failure with preserved ejection fraction (HFpEF) is a prevalent condition with no established prevention or treatment strategies. Furthermore, the pathophysiology and predisposing risk factors for HFpEF are incompletely understood. Therefore, we sought to characterize the incidence and determinants of HFpEF in the Multi-Ethnic Study of Atherosclerosis (MESA). Our study included 6,781 MESA participants (White, Black, Chinese, and Hispanic men and women age 45 to 84 years, free of baseline cardiovascular disease). The primary end point was time to diagnosis of HFpEF (left ventricular ejection fraction ≥45%). Multivariable adjusted hazard ratios (HRs) with 95% confidence intervals were calculated to identify predictors of HFpEF. Over median follow-up of 11.2 years (10.6 to 11.7), 111 subjects developed HFpEF (cumulative incidence 1.7%). Incidence rates were similar across all races/ethnicities. Age (HR 2.3 [1.7 to 3.0]), hypertension (HR 1.8 [1.1 to 2.9]), diabetes (HR 2.3 [1.5 to 3.7]), body mass index (HR 1.4 [1.1 to 1.7]), left ventricular hypertrophy by electrocardiography (HR 4.3 [1.7 to 11.0]), interim myocardial infarction (HR 4.8 [2.7 to 8.6]), elevated N-terminal of the prohormone brain natriuretic peptide (HR 2.4 [1.5 to 4.0]), detectable troponin T (HR 4.5 [1.9 to 10.9]), and left ventricular mass index by magnetic resonance imaging (MRI; 1.3 [1.0 to 1.6]) were significant predictors of incident HFpEF. Worsening renal function, inflammatory markers, and coronary artery calcium were significant univariate but not multivariate predictors of HFpEF. Gender was neither a univariate nor multivariate predictor of HFpEF. In conclusion, we demonstrate several risk factors and biomarkers associated with incident HFpEF that were consistent across different racial/ethnic groups and may represent potential therapeutic targets for the prevention and treatment of HFpEF.
Heart failure with preserved ejection fraction (HFpEF) is an increasingly prevalent condition associated with significant morbidity and mortality. In contrast to heart failure with reduced ejection fraction (HFrEF), there are currently no evidence-based therapies approved for the treatment of HFpEF, and several recent trials have all had negative outcomes. Because of the lack of effective therapies for such a prevalent condition, a multidisciplinary group from the Food and Drug Administration, academia, and industry released a document underscoring the need for additional research to better understand the pathophysiology of new-onset HFpEF. The identification of risk factors has the potential to elucidate mechanisms of disease and to highlight possible targets for disease prevention. Although previous epidemiologic studies have described the prevalence, co-morbidities, and outcomes associated with HFpEF, little is known about the risk factors associated with incident HFpEF. Furthermore, the few studies that have described risk factors for HFpEF have been from ethnically homogeneous Caucasian populations. Thus, even less is known about the incidence and risk factors for new-onset HFpEF in those from other racial/ethnic backgrounds. Therefore, we sought to characterize the incidence and risk factors associated with new-onset HFpEF among subjects from various racial/ethnic groups in the Multi-Ethnic Study of Atherosclerosis (MESA).
Methods
MESA is a prospective observational cohort of 6,814 men and women aged 45 to 84 years who were free of known cardiovascular disease at the time of enrollment. Subjects from different racial/ethnic backgrounds (White, Black, Hispanic, and Chinese) were enrolled from July 2000 to September 2002 at 6 different field centers in the United States (Baltimore; Chicago; Forsyth County, North Carolina; Los Angeles; New York City; and St. Paul, Minnesota). Full details of the MESA study design have been published previously. The study was approved by the institutional review board of each site, and all participants gave written informed consent. Of the total population, 33 subjects were missing necessary covariates and were, therefore, excluded from the overall analysis, resulting in a final study population of 6,781 participants.
At the initial examination, staff at each center collected baseline information on demographics, medical history, and cigarette use. Blood pressure, anthropometric measurements, electrocardiograms, and laboratory data were obtained as previously described. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Hypertension was defined as systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg or medical treatment for hypertension. Diabetes was defined as fasting plasma glucose >126 mg/dl or a history of medical treatment for diabetes. Interim myocardial infarction (MI) was defined as an MI (diagnosed by combination of symptoms, electrocardiographic findings, and levels of cardiac biomarkers) that occurred during follow-up before the diagnosis of heart failure (HF).
Estimated glomerular filtration rate was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation. Serum inflammatory markers, high-sensitivity C-reactive protein (hsCRP), and interleukin (IL)-6 were measured as part of the baseline examination. HsCRP was measured using the BNII nephelometer (Dade-Behring Inc., Deerfield, Illinois), and IL-6 was measured by ELISA (Quantikine HS Human IL-6 Immunoassay; R&D Systems, Minneapolis, Minnesota). A subgroup of 5,569 subjects had baseline measurements of N-terminal of the prohormone brain natriuretic peptide (NT-proBNP) and troponin T measured by Elecsys immunoassay (Roche Diagnostics Corporation, Indianapolis, Indiana).
All participants underwent coronary artery calcium (CAC) scoring as part of the baseline examination. Details regarding the methods for scanning and interpretation have been reported previously. Each of the 6 centers measured CAC with either a cardiac-gated electron beam CT scanner (Chicago, Los Angeles, New York) or a multidetector computed tomography (Baltimore, Forsyth County, St. Paul). Individuals were scanned twice, and mean CAC (Agatston) score was used for all analyses. Images were interpreted at the MESA computed tomography reading center (Los Angeles Biomedical Research Institute, Torrance, California).
A subgroup of 4,980 subjects underwent baseline cardiac MRI using 1.5 T magnetic resonance scanners for the determination of left ventricular (LV) mass as has been previously described. Briefly, the endocardial and epicardial borders were contoured using a semi-automated method. The difference between the epicardial and endocardial areas for all slices was multiplied by slice thickness and section cap and then multiplied by the specific myocardial density (1.04 g/ml) to determine LV mass. The LV mass index was defined as LV mass divided by body surface area.
Subjects were followed for a median of 11.2 years (interquartile range [IQR] 10.6 to 11.7 years). An interviewer contacted each participant or family member by telephone at 9- to 12-month intervals to inquire about interim hospital admissions, outpatient cardiovascular diagnoses, and deaths. MESA obtained medical records for approximately 98% of hospital events and 95% of outpatient diagnoses. Two physicians from the MESA mortality and morbidity review committee independently classified events. In the event of disagreement, the full committee made the final classification.
The primary outcome of this study was time to new diagnosis of HF, defined as a first HF event. HF was an adjudicated event in MESA requiring symptoms such as shortness of breath or edema, a physician diagnosis of HF, and documented medical treatment for HF. Subjects with an adjudicated diagnosis of HF were included in this analysis if they had an evaluation of left ventricular ejection fraction (LVEF) by echocardiography at the time of HF diagnosis that could be obtained from review of medical records. Each new HF diagnosis was categorized as either HFpEF (LVEF ≥45%) or HFrEF (LVEF <45%) as has been previously described.
Baseline characteristics were reported according to 3 categories: subjects who developed HFpEF, those who developed HFrEF, and those who did not develop HF. Continuous variables were reported as mean (SD) for normally distributed data and as median (IQR) for right-skewed data including hsCRP and IL-6. For the subgroup of subjects who had baseline measures of NT-proBNP, they were categorized as being above or below the 75th percentile (112 pg/ml), which has previously been shown within MESA to be a robust cut point for predicting incident HF. The subgroup of subjects with baseline measurements of troponin T was categorized as having detectable or undetectable levels (<0.01).
The cumulative incidence of HFpEF was described using Kaplan-Meier estimates. Proportional hazards regression models were used to evaluate the association between baseline characteristics and incident HFpEF. The Fine-Gray model was used to account for competing risk of death or developing HFrEF. The association between interim MI and new-onset HFpEF was analyzed by including interim MI as a time-varying covariate into the existing model. Variables that reached significance in univariate analysis (p <0.05) were carried forward into multivariable-adjusted models (adjusted for all covariates that reached significance in the univariate analysis plus the following: gender, race/ethnicity, socioeconomic status [based on level of education], and MESA site). Hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated per increase in SD for continuous variables and for change in classification of binary variables. To calculate the HR associated with each racial/ethnic group, the risk of HFpEF in White subjects was used as reference. HRs for hsCRP and IL-6 were calculated using log transformation.
Because NT-proBNP and troponin T were not available in the entire population, they were not entered into the main multivariate model. Instead, univariate and multivariate-adjusted HRs for NT-proBNP and troponin T were calculated in a separate but similar model (adjusting for age, gender, race/ethnicity, socioeconomic status, MESA site, and all covariates that reached significance in the univariate analysis) for the 5,569 subjects who had these baseline measures.
Similarly, cardiac MRI was not performed in the entire population, and only 4,980 subjects had LV mass index and the necessary covariates for our study. Therefore, LV mass index was not included in the main multivariate model. The association between LV mass index and incident HFpEF was evaluated in a separate univariate and multivariate-adjusted model (adjusted for age, gender, race/ethnicity, socioeconomic status, MESA site, and all covariates that reached significance in the univariate analysis) for the subgroup of subjects who underwent baseline MRI.
To further characterize the relation between race/ethnicity and incident HFpEF, baseline characteristics for subjects who developed HFpEF were presented according to race/ethnicity. To evaluate for potential effect modification by race/ethnicity for each of the significant multivariate predictors, we included a corresponding interaction term for each individual race/ethnicity with the significant multivariate predictors. A p value <0.05 was considered statistically significant. All analyses were performed using Stata, version 12 (StataCorp LP, College Station, Texas).
Results
In total, 257 subjects developed incident HF, of whom 39 did not have a recorded LVEF at the time of diagnosis. Individuals with and without documented LVEF at the time of new HF diagnosis had similar baseline characteristics. Of the 218 subjects with known LVEF at the time of new HF diagnosis, 111 had HFpEF and 107 had HFrEF ( Supplementary Figure 1 ). Median LVEF in the HFpEF group was 57% (52 to 63), whereas median LVEF in the HFrEF group was 33% (26 to 37). Figure 1 demonstrates the distribution of LVEF for the overall cohort at the time of HF diagnosis. The distribution of LVEF in subjects with HFpEF is shown in Supplementary Figure 2 . Baseline characteristics according to HFpEF, HFrEF, or no HF are presented in Table 1 . The median time to diagnosis of HFpEF was 6.6 (3.3 to 8.8) years with a cumulative incidence of 1.7%. Figure 2 demonstrates the cumulative incidence of new-onset HFpEF. When stratified by race/ethnicity, the cumulative incidence of HFpEF was similar across groups (White = 2.0%, Chinese = 1.5%, Black = 1.3%, Hispanic = 1.7%, p = 0.369), and there was no difference in time to new-onset HFpEF (p = 0.8779), as shown in the Kaplan-Meier plot ( Figure 3 ).
| HFpEF N = 111 | HFrEF N = 107 | No HF N = 6524 | |
|---|---|---|---|
| Demographics | |||
| Age (years) | 70 (±9) | 67 (±9) | 62 (±10) |
| Female | 54 (48.7%) | 30 (28%) | 3,480 (53.3%) |
| Race/Ethnicity | |||
| White | 51 (46%) | 42 (39.3%) | 2,507 (38.4%) |
| Chinese | 12 (10.8%) | 1 (0.9)% | 784 (12%) |
| Black | 24 (21.6%) | 44 (41.1)% | 1,799 (27.6%) |
| Hispanic | 24 (21.6)% | 20 (18.7%) | 1,434 (22%) |
| Clinical Characteristics | |||
| Heart Rate (beats per minute) | 65 (±10) | 65 (±11) | 63 (±10) |
| Systolic BP (mm Hg) | 140 (±24) | 136 (±22) | 126 (±21) |
| Hypertension | 81 (73%) | 77 (72%) | 2,860 (43.8%) |
| Antihypertensive medication use | 68 (61.3%) | 64 (59.8%) | 2,370 (36.3%) |
| Body mass index (kg/m 2 ) | 29.8 (±5.8) | 29.2 (±5.4) | 28.3 (±5.5) |
| Diabetes | 35 (31.5%) | 30 (28%) | 772 (11.9)% |
| Current Smoking | 13 (11.7%) | 21 (19.6%) | 847 (13%) |
| Total Cholesterol (mg/dL) | 186 (±34) | 188 (±35) | 194 (±36) |
| HDL-Cholesterol (mg/dL) | 49 (±15) | 48 (±13) | 51 (±15) |
| eGFR (ml/min) | 71 (±18) | 70 (±20) | 78 (±16) |
| CRP (mg/L) | 2.7 (1.1-6.0) | 3.2 (1.2-5.9) | 1.9 (0.8-4.2) |
| IL-6 (pg/mL) | 1.6 (1.1-2.8) | 1.4 (0.9-2.3) | 1.2 (0.8-1.9) |
| LVH by ECG | 5 (4.6%) | 5 (4.7%) | 55 (0.85%) |
| CAC > 0 | 77 (69.4)% | 79 (73.8%) | 3,193 (48.9%) |
| Interim MI | 16 (14.4%) | 12 (11.2%) | 140 (2.2%) |
| NT-proBNP > 75th percentile ∗ | 55 (59.8%) | 51 (62.2%) | 1,269 (23.7%) |
| Detectable Troponin T ∗ | 12 (13%) | 9 (11%) | 58 (1.1%) |
| LV mass index (g/m 2 ) † | 111.0 (±23.2) | 129.7 (±31.3) | 103.5 (±18.0) |
∗ Available in a subset of the population (5,569 individuals).
† Available in a subset of the population (4,980 individuals).
Unadjusted and multivariable-adjusted HRs for demographic, clinical, and biomarker predictors of incident HFpEF are listed in Table 2 . After multivariable adjustment, the following demographic and clinical characteristics remained significant predictors of HFpEF: age (HR 2.3 [1.7 to 3.0]), hypertension (HR 1.8 [1.1 to 2.9]), diabetes (HR 2.3 [1.5 to 3.7]), BMI (HR 1.4 [1.1 to1.7]), and interim MI (HR 4.8 [2.7 to 8.6]). The following markers were also significant predictors of HFpEF after multivariate adjustment: left ventricular hypertrophy (LVH) by electrocardiography (ECG) (HR 4.3 [1.7 to 11.0]), elevated NT-proBNP (HR 2.4 [1.5 to 4.0]), detectable troponin T (HR 4.5 [1.9 to 10.9]), and LV mass index by MRI (1.3 [1.0 to 1.6]). Increased heart rate, worsening renal function (glomerular filtration rate), elevated inflammatory markers (CRP and IL-6), and presence of CAC (CAC >0) were all significant univariate predictors that were no longer significant after multivariable adjustment. Female gender was not associated with an increased risk of HFpEF in either univariate or multivariate-adjusted models.
| Unadjusted HR (95% CI) | p-value | Multivariable adjusted HR (95% CI) | p-value | |
|---|---|---|---|---|
| Demographics | ||||
| Age (per SD) | 2.33 (1.91-2.86) | <0.001 | 2.27 (1.72-3.01) | <0.001 |
| Female | 0.84 (0.58-1.22) | 0.369 | 0.89 (0.54-1.46) | 0.638 |
| Race/Ethnicity | ||||
| White | Ref (–) | — | Ref (–) | — |
| Chinese | 0.79 (0.42-1.47) | 0.454 | 1.53 (0.64-3. 67) | 0.337 |
| Black | 0.69 (0.42-1.12) | 0.132 | 0.46(0.26-0.82) | 0.009 |
| Hispanic | 0.87 (0.54-1.41) | 0.576 | 0.66 (0.34-1.30) | 0.231 |
| Clinical Characteristics | ||||
| Heart Rate (per SD) | 1.26 (1.05-1.50) | 0.012 | 1.11 (0.92-1.36) | 0.279 |
| Hypertension | 3.44 (2.26-5.23) | <0.001 | 1.81 (1.14-2.90) | 0.013 |
| Body mass index (per SD) | 1.27 (1.09-1.49) | 0.002 | 1.35 (1.08-1.68) | 0.009 |
| Diabetes | 3.42 (2.29-5.11) | <0.001 | 2.33 (1.47-3.71) | <0.001 |
| Current Smoking | 0.91 (0.51-1.62) | 0.743 | — | — |
| Total Cholesterol (per SD) | 0.79 (0.65-0.97) | 0.023 | 0.94 (0.76-1.16) | 0.555 |
| HDL-Cholesterol (per SD) | 0.86 (0.69-1.06) | 0.151 | — | — |
| eGFR (per SD) | 1.51 (1.25-1.84) | <0.001 | 0.92 (0.72-1.17) | 0.508 |
| CRP (per log SD) | 1.27 (1.09-1.49) | 0.003 | 1.17 (0.93-1.46) | 0.177 |
| IL-6 (per log SD) | 2.15 (1.64-2.80) | <0.001 | 1.32 (0.91-1.93) | 0.147 |
| LVH by ECG | 5.00 (2.01-12.44) | 0.001 | 4.33 (1.70-11.04) | 0.002 |
| CAC > 0 | 2.35 (1.57-3.51) | <0.001 | 0.91 (0.54-1.51) | 0.702 |
| Interim MI | 6.66 (3.91-11.34) | <0.001 | 4.80 (2.67-8.62) | <0.001 |
| NT-proBNP > 75th percentile ∗ | 4.65 (3.07-7.03) | <0.001 | 2.41 (1.45-4.00) | 0.001 |
| Detectable Troponin T ∗ | 11.55 (6.24-21.39) | <0.001 | 4.52 (1.88-10.87) | 0.001 |
| LV mass index (per SD) † | 1.36 (1.13-1.64) | 0.001 | 1.29 (1.04-1. 60) | 0.018 |
∗ Available in a subset of the population (5,569 individuals).
† Available in a subset of the population (4,980 individuals).
In the unadjusted model, Black subjects had a nonsignificant trend toward lower risk of HFpEF compared with White subjects (HR 0.7 [0.4 to 1.1]), which was significant after multivariable adjustment (HR 0.5 [0.3 to 0.8]).
To better characterize the risk of HFpEF by race/ethnicity, we compared baseline characteristics of subjects who developed incident HFpEF according to racial/ethnic group. As listed in Table 3 , the different ethnic groups had relatively similar baseline characteristics with a few notable exceptions including BMI, CAC >0, detectable troponin T, and diabetes. The proportion of Black and Hispanic subjects with diabetes (58% and 50%, respectively) was markedly higher than the proportion of White and Chinese subjects with diabetes (12% and 25%, respectively). Although a similar trend was seen in the general population with Black and Hispanic subjects having higher rates of diabetes (18% each vs 6% and 13% for White and Chinese subjects, respectively), the disparities were far greater in subjects who went on to develop HFpEF. Statistical testing revealed no interaction between race/ethnicity and the effect of diabetes on incident HFpEF. The small number of subjects with LVH and detectable troponin T limited our ability to perform interaction testing between race/ethnicity and the impact of these 2 risk factors on incident HFpEF. However, interaction testing between race/ethnicity and the effect of the remaining clinical risk factors on incident HFpEF showed that race/ethnicity did not modify the effect of the other risk factors on incident HFpEF.
| White n = 51 | Chinese n = 12 | Black n = 24 | Hispanic n = 24 | p-value | |
|---|---|---|---|---|---|
| Demographics | |||||
| Age (years) | 71 (±7) | 71 (±10) | 67 (±10) | 69 (±9) | 0.16 |
| Female | 21 (41.2%) | 9 (75%) | 15 (62.5%) | 9 (37.5%) | 0.057 |
| Clinical Characteristics | |||||
| Systolic BP (mm Hg) | 136 (±22) | 141 (±23) | 148 (±25) | 140 (±25) | 0.223 |
| Hypertension | 36 (70.6%) | 8 (66.7%) | 21 (87.5%) | 16 (66.7%) | 0.328 |
| Body mass index (kg/m 2 ) | 29.5 (±5.7) | 25.4 (±2.8) | 30.6 (±4.8) | 31.9 (±7.0) | 0.011 |
| Diabetes | 6 (11.8%) | 3 (25%) | 14 (58.3%) | 12 (50%) | <0.001 |
| Current Smoking | 6 (11.8%) | 0 (0%) | 5 (20.8%) | 2 (8.3%) | 0.285 |
| Total Cholesterol (mg/dL) | 188 (±33) | 180 (±31) | 183 (±36) | 190 (±38) | 0.81 |
| HDL-Cholesterol (mg/dL) | 50 (±16) | 49 (±11) | 50 (±16) | 46 (±12) | 0.68 |
| eGFR (ml/min) | 71 (±15) | 75 (±18) | 74 (±20) | 67 (±22) | 0.469 |
| CRP (mg/L) | 2.4 (1.1-4.6) | 1.9 (1.0-3.2) | 4.8 (1.1-14.1) | 4.0 (1.8-7.0) | 0.0849 |
| IL-6 (pg/mL) | 1.5 (1.1-2.7) | 1.5 (0.9-1.8) | 2.4 (1.3-3.0) | 2.1 (1.1-3.8) | 0.1064 |
| LVH by ECG | 2 (4.1%) | 0 (0%) | 2 (8.3%) | 1 (4.2%) | 0.709 |
| CAC > 0 | 37 (72.6%) | 12 (100%) | 12 (50%) | 16 (66.7%) | 0.02 |
| Interim MI | 6 (11.8%) | 3 (25%) | 3 (8.3%) | 5 (20.8%) | 0.407 |
| NT-pro-BNP > 75th percentile ∗ | 26 (57.8%) | 6 (54.6%) | 10 (71.4%) | 13 (59.1%) | 0.802 |
| Detectable Troponin T ∗ | 3 (6.7%) | 0 (0%) | 2 (14.3)% | 7 (31.8%) | 0.018 |
| LV mass index (g/m 2 ) † | 107.4 (±18.3) | 109.9 (±19.5) | 115.6 (±31.9) | 116.7 (±27.5) | 0.5854 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
