Comorbidity and Ventricular and Vascular Structure and Function in Heart Failure with Preserved Ejection Fraction: A Community Based Study

Abstract

Background

Patients with heart failure and preserved ejection fraction (HFpEF) display increased adiposity and multiple comorbidities, factors that in themselves may influence cardiovascular (CV) structure and function. This has sparked debate as to whether HFpEF represents a distinct disease or an amalgamation of comorbidities. We hypothesized that fundamental CV structural and functional alterations are characteristic of HFpEF, even after accounting for body size and comorbidities.

Methods and Results

Comorbidity adjusted CV structural and functional parameters scaled to independently generated and age appropriate allometric powers were compared in community-based cohorts of HFpEF patients (n=386) and age/gender-matched healthy (CON, n=193) and hypertensive (HTN, n=386) controls. Within HFpEF patients, body size and concomitant comorbidity adjusted CV structural and functional parameters and survival were compared in those with and without individual comorbidities. Among HFpEF patients, comorbidities (obesity, anemia, diabetes and renal dysfunction) were each associated with unique clinical, structural, functional and prognostic profiles. However, after accounting for age, gender, body size and comorbidities, greater concentric hypertrophy, atrial enlargement and systolic, diastolic and vascular dysfunction were consistently observed in HFpEF compared to CON and HTN.

Conclusions

Comorbidities influence ventricular-vascular properties and outcomes in HFpEF, yet fundamental disease-specific changes in cardiovascular structure and function underlie this disorder. These data support the search for mechanistically-targeted therapies in this disease.

Patients with heart failure and preserved ejection fraction (HFpEF) are more frequently female, usually elderly with a history of hypertension and commonly have multiple comorbidities including obesity, anemia, diabetes and renal dysfunction. Each of these comorbidities may influence ventricular and vascular structure and function, provoking debate as to whether HFpEF is a distinct disease requiring specific therapy or simply an amalgamation of age-related comorbidities^1–3. We hypothesized that HFpEF, as it presents in the community, is associated with unique alterations in ventricular-vascular properties after accounting for the confounding influences of age, gender, body size and comorbidities.

Accordingly, we compared ventricular-vascular properties in a community-based cohort of HFpEF patients and control populations without HF. To account for the effects of age, gender and hypertension, we compared HFpEF patients to carefully age- and gender-matched healthy (CON) and hypertensive (HTN) comparator populations from the same community. To account for body size differences, ventricular vascular properties were scaled to allometric coefficients generated in an age-appropriate, disease-free, normal body size cohort. To account for the effects of comorbidities, comparison between HFpEF and control populations adjusted for comorbidties and the impact of comorbidities on ventricular-vascular properties and survival among HFpEF patients was defined.

Methods

The study was approved by Mayo Clinic institutional review board and all subjects gave prospective consent and/or consent for use of medical records for research.

Study subjects

Consecutive adult patients with HF (Framingham criteria) were prospectively enrolled between September, 2003, and August, 2009 by real-time interrogation of electronic medical records using natural language processing techniques as previously described⁴. This cohort is derived from an ongoing prospective study and includes subjects included in earlier analyses of smaller subsets addressing other hypotheses. Subjects with significant left-sided valve disease, infiltrative, inflammatory or hypertrophic cardiomyopathy, congenital heart disease, pericardial disease, EF<50% or without assessment of vascular structure (ascending aortic diameter) were excluded. Vital status was determined from Rochester Epidemiology Project procedures as previously described⁴.

To identify unique age- and gender-matched CON (no hypertension, diabetes, vascular or valve disease or HF) and HTN (hypertension but no HF) comparator groups, we used stratified random sampling of Olmsted County residents in the Mayo Clinic Echocardiographic Laboratory database (February, 1998 to June, 2010) and manual medical record review (SFM) with matching in a 1:2:2 ratio of CON: HTN: HFpEF.

Physician’s diagnoses with documentation of clinical features, supportive laboratory or imaging data and/or medication use were used to define clinical characteristics (see Supplemental Methods) including medication use at the time of the index echocardiogram. Body surface area was calculated by the Gehan method (BSA=0.0235*height (cm)^0.42246*weight(kg)^0.51456) and body mass index as weight/height² (kg/m²).

Doppler Echocardiography

Echocardiograms were performed by registered cardiac diagnostic sonographers and interpreted by echocardiologists.

Cardiac structure and function

LV and atrial geometry were measured with 2D or M-mode echocardiography and used to calculate EF, LV mass, relative wall thickness (RWT), left atrial (LA) volume (area-length method), stroke volume (SV) and pulmonary artery systolic pressure (PASP) according to American Society of Echocardiography (ASE) conventions^5,6. LV end-diastolic volume (EDV) was calculated using stroke volume and EF to avoid geometric assumptions (EDV = SV/EF)^{4, 7}. Cardiac power output (CPO) was computed (CO*MAP*2.22*10⁻³) in watts⁸. The early diastolic septal annular tissue velocity (e′), transmitral flow velocity (E) and deceleration time (DT) were used to quantify relaxation, filling pressure (E/e′) and operant diastolic stiffness. To estimate LV operating compliance (EDV at a given end diastolic pressure), natural log of EDV was compared between groups adjusting for age, gender, body size (natural log of height and natural log of weight), comorbidities and quartiles of filling pressure (E/e′) (see Supplemental Methods).

Vascular structure and function

Aortic diameter (D) was used to calculate aortic area (πD²/4). Brachial pulse pressure (PP; systolic-diastolic BP), mean arterial (MAP, diastolic BP+0.4128*PP),⁸ end-systolic pressure (ESP; 0.9 * brachial systolic BP)⁹, effective arterial elastance (Ea; ESP/SV)⁹, total systemic arterial compliance (SAC; SV/PP) and systemic vascular resistance (SVR;(MAP/cardiac output)*80) were calculated as previously described⁴. All comparisons of Ea and SAC between groups were adjusted for heart rate and MAP respectively.

Allometric scaling

As the relationship between body size and physiologic parameters is often complex and non-linear, simple ratiometric scaling to height or BSA may yield erroneous conclusions^{8, 10–12}. We assembled a disease free, community-based cohort with a normal body size (<25 kg/m²; n=345; characteristics in Supplemental Table I) to derive age and gender appropriate allometric scaling coefficients (see Supplemental Methods). Allometric coefficients were obtained by regressing the natural log of the ventricular-vascular properties on log of height or BSA after adjustment for age and gender. This log-log method effectively results in a multiplicative model¹³ that has the same functional form as the standard allometric equations and yields nearly identical results as non-linear regression¹². While gender-body size interaction terms were not significant, the age-body size interaction terms were significant for most parameters. Thus, scaling coefficients were derived in a subset (age > median) of the normal cohort with an age distribution similar to the HFpEF and comparator populations (see Supplemental Methods and Supplemental Table I and II). This age-appropriate, normal body size cohort was also used to derive upper normal values (mean+2SD) for allometrically scaled LV mass measurements.

Sensitivity analyses

To further evaluate differences in ventricular-vascular properties adjusting for body size, a sensitivity analysis was performed using multivariable least squares linear regression to compare log transformed ventricular-vascular properties between groups (dummy variables) adjusting for log transformed height and weight as well as age, gender and comorbidities rather than using the allometric scaling indices derived in the normal populations.

Laboratory data

Glomerular filtrationrate (GFR) was estimated using the modification of diet in renal disease (MDRD) formula.

Missingness

The missingness rate for all variables with less than 100% availability is shown in Supplemental Table III.

Statistical Analysis

Data are presented as mean ± standard deviation or % frequency. Unadjusted comparisons variables across groups were performed by one-way analysis of variance (ANOVA) followed by Dunnet’s test for group comparisons to HFpEF for continuous variables. For categorical variables, comparisons across and between groups were performed by the Chi Square test. Multivariable least squares linear regression was used to test multiple covariates where groups were entered as dummy variables appropriately constructed for the comparison of interest.

Comparisons of ventricular-vascular properties between HFpEF patients with or without an individual comorbidity (anemia, diabetes or renal dysfunction) was performed in log-log models where the log transformed variables were compared between groups adjusting for natural log of height, natural log of weight and other concomitant comorbidities. Comparisons of ventricular-vascular properties between HFpEF patients with or without obesity was similarly performed adjusting for natural log of height and other concomitant comorbidities. For parameters not related to body size, comparisons adjusted for other comorbidities. The adjusted geometric means within groups and the p value for the group effect are presented.

Survival up to 5 years after HF diagnosis was assessed. The Kaplan-Meier method tested for differences in survival between groups by the log-rank test. Cox proportional-hazards regression was used to adjust for the effect of differences in baseline characteristics on survival.

All analyses were 2 tailed, and a p<0.05 was considered statistically significant. Analysis was performed using the JMP^® analysis software.

Results

Clinical characteristics

HFpEF subjects were elderly and predominately female with a high prevalence of obesity, anemia, diabetes, renal dysfunction, cardiovascular conditions and medication use (Table 1). Systolic and mean blood pressures were higher in HFpEF than CON but lower in HFpEF than HTN, while PP was higher in HFpEF than CON but similar to HTN.

Table 1.

Characteristics of normal (CON) and hypertensive (HTN) controls and HFpEF patients.

	CON	HTN	HFpEF	p‡
N	193	386	386
Age	77.1±8.5	76.9±11	77.5±11.4
Male, %	44%	44%	44%
Weight, kg	72±18	78±19	83±24*†	<0.001
Height, m	1.67±0.11	1.66±0.11	1.66±0.1	0.48
Body surface area, m2	1.83±0.27	1.91±0.27	1.97±0.31*†	<0.001
Systolic blood pressure, mm Hg	121±12	137±20	132±23*†	<0.001
Pulse pressure, mm Hg	52±12	65±18*	64±19*	<0.001
Heart rate, bpm	71±14	69±15	72±15	0.053
Comorbidities
Hypertension	0%	100%	86%*†	<0.001
% Obese (BMI≥30 Kg/m2)	14%	33%	42%	<0.001
Anemia (<12 g/dl female; <13 g/dl male)	29%	25%	56%*†	<0.001
Diabetes mellitus	0%	18%	35%*†	<0.001
GFR, mL/min/1.73 m2	70±18	65±20	56±23*†	<0.001
Atrial fibrillation	14%	18%	47%*†	<0.001
Coronary artery disease	0%	24%	51%*†	<0.001
Peripheral vascular disease	0%	10%	21%*†	<0.001
Cerebrovascular disease	0%	19%	31%*†	<0.001
Dyslipidemia	30%	51%	65%*†	<0.001
Cigarette smoking (ever)	39%	42%	51%*†	0.013
Medications
Betablockers	8%	44%	63%*†	<0.001
ACE-I/ARBs	0%	43%	52%*†	<0.001
Calcium channel blockers	3%	22%	36%*†	<0.001
Diuretics	3%	48%	63%*†	<0.001
Statins	11%	35%	48%*†	<0.001

^*p<0.05 HFpEF vs CON;

^†p<0.05 HFpEF vs HTN

^‡ANOVA for continuous variables, Chi Square for discrete variables

Differences across groups were similar in men and women (Supplemental Table IV) although men with HFpEF were younger and had more vascular disease than female HFpEF patients.

Cardiac structure

HFpEF patients had higher height- or BSA-scaled LV mass than CON or HTN (Table 2). In the sensitivity analysis, natural log of LV mass was greater in HFpEF than CON or HTN (Table 3). HFpEF patients had more abnormal LV geometry (Figure 1) whether the presence of LV hypertrophy was ascertained using the allometrically scaled or published (sex-specific LV mass/BSA) hypertrophy criteria.

Table 2.

Ventricular structure and function in normal (CON) and hypertensive (HTN) controls and HFpEF patients.

	CON	HTN	HFpEF	ANOVA p	HFpEF vs CON‡		HFpEF vs HTN‡
					PE (HFpEF)	p‡	PE (HFpEF)	p‡
Cardiac structure
LV mass/Height1.6	66 ± 17	77 ± 20	89 ± 27*†	<0.001	9.9	<0.001	4.8	<0.001
LV mass/BSA1.19	74 ± 16	81 ± 19	91 ± 25*†	<0.001	7.9	<0.001	4.1	<0.001
Relative wall thickness	0.42 ± 0.07	0.44 ± 0.07	0.46 ± 0.10*†	<0.001	0.02	<0.001	0.01	<0.001
End diastolic volume/Height 1.4	62.9 ± 13.4	67 ± 15.5	66.1 ± 17.7	0.029	0.8	0.32	−1.0	0.12
End diastolic volume/BSA0.98	72.1 ± 15.2	72.9 ± 17.3	69.9 ± 18.0	0.063	−1.1	0.22	−1.7	0.013
Left atrial volume/Height 2.26	18.9 ± 7.2	21.8 ± 7.7	28.7 ± 10.7*†	<0.001	4.7	<0.001	3.2	<0.001
Left atial volume/BSA1.48	24.7 ± 9.5	26.8 ± 9.9	34.1 ± 13.2*†	<0.001	4.6	<0.001	3.5	<0.001
Systolic performance
Ejection fraction, %	64.1 ± 5.9	64.3 ± 5.7	61.8 ± 6.9*†	<0.001	−1.4	<0.001	−1.5	<0.001
Stroke volume/Height1.29	42.7 ± 8.4	45.6 ± 10.2	42.9 ± 10.9†	0.001	−0.5	0.301	−1.8	<0.001
Stroke volume/BSA0.92	48.0 ± 9.5	48.7 ± 11.3	44.6 ± 10.8*†	<0.001	−1.8	0.001	−2.3	<0.001
Cardiac power output/Height 0.97	0.70 ± 0.16	0.79 ± 0.21	0.74±0.24†	<0.001	0.01	0.23	−0.03	<0.001
Cardiac power output/BSA0.88	0.68 ± 0.15	0.74 ± 0.19	0.67 ± 0.21†	<0.001	0.00	0.73	−0.03	<0.001
Diastolic function
Mitral E/e′	10.8 ± 4.1	13.0 ± 5.8	17.2 ± 8.4*†	<0.001	2.7	<0.001	1.7	<0.001
e′, m/sec	0.070 ± 0.021	0.063 ± 0.019	0.062 ± 0.022*	<0.001	−0.004	<0.001	−0.001	0.44
Deceleration time, ms	230 ± 46	227 ± 55	200 ± 52*†	<0.001	−15	<0.001	−12	<0.001
PASP, mmHg	33 ± 11	36 ± 11	48 ± 15*†	<0.001	6.2	<0.001	4.8	<0.001

^*p<0.05 HFpEF vs CON;

^†p<0.05 HFpEF vs HTN

^‡Parameter estimate (PE) is increment in variable associated with the HFpEF state as compared to CON or HTN when adjusted for age, gender, hemoglobin, glomerular filtration rate, and diabetes.

Table 3.

Comorbidity adjusted analysis showing effect of body size and group on ventricular vascular parameters from log log models

	Natural log of Height (m)		Natural log of Weight (kg)		HFpEF vs CON‡		HFpEF vs HTN‡
log transformed variable	PE	p	PE	p	PE (HFpEF)	p	PE (HFpEF)	p
LV mass, g	0.304	0.15	0.523	<0.001	0.096	<0.001	0.045	<0.001
End diastolic volume, ml	0.812	<0.001	0.238	<0.001	−0.009	0.49	−0.024	0.012
Left atrial volume, ml	0.276	0.40	0.558	<0.001	0.160	<0.001	0.116	<0.001
stroke volume, ml	0.691	<0.001	0.229	<0.001	−0.035	0.004	−0.050	<0.001
cardiac output, l/min	0.604	0.003	0.266	<0.001	−0.035	0.004	−0.034	<0.001
cardiac power output, watts	0.366	0.13	0.319	<0.001	−0.012	0.39	−0.057	<0.001
aortic area, mm2	0.470	0.009	0.199	<0.001	0.017	0.11	0.002	0.78
arterial elastance, mmHg/ml	−0.906	<0.001	−0.179	<0.001	0.066	<0.001	0.019	0.064
vascular resistance, dyn•s•cm−5	−0.749	0.002	−0.225	<0.001	0.058	<0.001	0.010	0.38
arterial compliance, ml/mmHg	0.887	0.003	0.176	0.004	−0.068	<0.001	−0.033	0.013

Figure 1.

LV geometry in healthy (CON) and hypertensive (HTN) controls and HFpEF. The prevalence of normal (Nl), concentric remodeling (CR) and concentric (CH) or eccentric (EH) hypertrophy based on relative wall thickness (≤ or > 0.42) and the presence or absence of LV hypertrophy (LVH) are shown using alternate methods of defining LVH.

Height- or BSA-scaled EDV was not different in HFpEF than CON but tended to be (height-scaled) or was (BSA-scaled) smaller in HFpEF than HTN (Table 2). In the sensitivity analysis, natural log of EDV was similar in HFpEF and CON but smaller in HFpEF than HTN (Table 3).

HFpEF patients had larger height- or BSA-scaled LA volume than CON or HTN (Table 2). The difference in LA volume between HFpEF and CON (p<0.001) or HTN (p<0.001) persisted after also adjusting for the presence of atrial fibrillation. Similarly, in the sensitivity analysis, natural log of LA volume was greater in HFpEF than CON or HTN (Table 3) and when also adjusting for atrial fibrillation (p<0.001).

LV Systolic Function

EF was slightly but significantly lower in HFpEF than CON or HTN (Table 2). The lower EF in HFpEF as compared to CON (p=0.002) or HTN (p<0.001) persisted after also adjusting for the presence of coronary disease.

Height-scaled SV was similar in HFpEF and CON but BSA-scaled SV was lower in HFpEF than CON (Table 2). Height- or BSA-scaled SV was lower in HFpEF than HTN. In sensitivity analysis, natural log of SV was smaller in HFpEF than CON or HTN (Table 3). Differences in CO were similar to SV (and significant) across groups(data not shown).

Height- or BSA-scaled CPO in HFpEF was similar to CON but lower than HTN (Table 2). In sensitivity analysis, natural log of CPO in HFpEF was similar to CON but lower than HTN (Table 3).

LV Diastolic Function

E/e′, PASP and DT were more abnormal in HFpEF than CON or HTN (Table 2). Adjusting for age, gender, natural log of height, natural log of weight, comorbidities and filling pressures (E/e′), natural log of EDV was smaller in HFpEF than HTN (Table 4) consistent with lower diastolic operating compliance. Given the limited range of E/e′ in CON, this analysis was restricted to HFpEF and HTN subjects.

Table 4.

Left ventricular operating compliance

Model for natural log of end diastolic volume (R2 = 0.29) in HTN and HFpEF subjects
Term	Parameter Estimate	95% CI	p value
Intercept	3.676	3.179; 4.173	<0.001
Gender (Female)	−0.066	−0.094; −0.039	<0.001
Natural log Weight (kg)	0.241	0.146; 0.337	<0.001
Natural log Height (m)	0.833	0.375; 1.291	<0.001
Hemoglobin (g/dl)	−0.017	−0.028; −0.006	0.002
E/e′ quartile	0.031	0.011; 0.050	0.002
Age (yr)	−0.002	−0.004; 0.0002	0.079
Glomerular filtration rate (ml/min/1.73 m2)	0.000	−0.0006; 0.0013	0.48
Diabetes	0.007	−0.017; 0.030	0.58
HFpEF vs HTN (HFpEF)	−0.033	−0.054; −0.012	0.002

Model for natural log of end diastolic volume (base model R 2 = 0.27) in HFpEF subjects
Y	Parameter Estimate	95% CI	p value
Intercept	3.286	2.657; 3.915	<0.001
Age (year)	−0.002	−0.004; 0.001	0.22
Female	−0.063	−0.100; −0.025	0.001
Natural log Height (m)	0.464	−0.169; 1.097	0.15
Natural log Weight (kg)	0.321	0.199; 0.443	<0.001
E/e′ quartile	0.032	0.006; 0.058	0.016
addition of:
Obesity	0.014	−0.031; 0.059	0.55
Anemia	0.045	0.018; 0.072	0.001
Diabetes	0.012	−0.019; 0.043	0.45
Renal dysfunction (GFR<60)	0.002	−0.028; 0.032	0.91

Vascular structure and function

Aortic area was larger in HFpEF than CON but similar to HTN (Table 5). In sensitivity analysis, natural log of aortic area was similar in HFpEF, CON and HTN (Table 3).

Table 5.

Vascular structure and function in in normal (CON) and hypertensive (HTN) controls and HFpEF patients.

	CON	HTN	HFpEF	ANOVA p	HFpEF vs CON‡		HFpEF vs HTN‡
					PE (HFpEF)	p‡	PE (HFpEF)	p‡
Aortic area	955 ± 193	991 ± 257	986 ± 262	0.23	32	0.004	6	0.50
Arterial elastance/Height−1.25	2.58 ± 0.53	2.79 ± 0.79	2.89 ± 0.88*	<0.001	0.19	<0.001	0.06	0.059
Arterial elastance/BSA−0.77	2.02 ± 0.43	2.19 ± 0.64	2.27 ± 0.70*	<0.001	0.15	<0.001	0.05	0.066
Vascular resistance/Height−1.06	2241 ± 512	2398 ± 645	2391 ± 766*	0.038	137	<0.001	32	0.24
Vascular resistance/BSA−0.62	1879 ± 431	2080 ± 548	2106 ± 671*	<0.001	151	<0.001	36	0.13
Arterial compliance/Height1.47	0.78± 0.27	0.69 ± 0.25	0.65 ± 0.25*	<0.001	−0.03	0.011	−0.02	0.092
Arterial compliance/BSA0.79	1.04 ± 0.37	0.88 ± 0.32	0.82 ± 0.31*†	<0.001	−0.07	<0.001	−0.03	0.035

^*p<0.05 HFpEF vs CON;

^†p<0.05 HFpEF vs HTN

^‡Parameter estimate (PE) is increment in variable associated with the HFpEF state as compared to CON or HTN when adjusted for age, gender, hemoglobin, glomerular filtration rate, and diabetes.

Height- or BSA-scaled Ea was higher in HFpEF than CON and tended to be higher in HFpEF than HTN (Table 5). In sensitivity analysis, natural log of Ea was higher in HFpEF than CON and tended to be higher than HTN (Table 3).

Height- or BSA-scaled SVR was higher in HFpEF than CON but similar in HFpEF and HTN (Table 5). In sensitivity analysis, natural log of SVR was higher in HFpEF than CON but similar in HFpEF and HTN (Table 3).

Height- or BSA-scaled SAC was lower in HFpEF than CON and tended to be (Height-scaled) or was (BSA-scaled) lower in HFpEF than HTN (Table 5). In sensitivity analysis, natural log of SAC was lower in HFpEF than CON or HTN (Table 3).

Comorbidities and Ventricular and Vascular Structure and Function in HFpEF

Obesity

Obesity was present in 42% of HFpEF patients (Table 6). Obese HFpEF patients were younger, more likely diabetic and had higher hemoglobin and pulse pressures than non-obese HFpEF patients. Obese patients had higher LV mass, EDV, and atrial volume with higher RWT suggesting concentric remodeling. Obese HFpEF patients tended to have higher EF and had higher SV, CO and CPO than non-obese HFpEF patients but diastolic function indices were similar. While LV operating compliance (Table 4) increased with increasing weight and height, there was no leftward shift of this relationship in obese patients to suggest worse diastolic function. Aortic area and SAC were larger, Ea was and SVR tended to be lower in obese HFpEF patients suggesting that the higher PP in obese patients reflects the higher SV rather than greater arterial stiffness. Mean and median follow up among HFpEF survivors was 4.2 ± 1.1 and 4.9 years respectively. Obesity was associated with better outcome adjusting for age, gender and other concomitant comorbidities (Figure 2 A and E). The relationship between obesity and outcomes did not appear U shaped (Supplemental Figure).

Table 6.

Clinical characteristics and ventricular and vascular structure and function in HFpEF patients according to comorbidities

	Non-obese	Obese	No Anemia	Anemia	No Diabetes	Diabetes	GFR≥56	GFR<56
n	222	164	165	206	251	135	183	188
Clinical characteristics	Unadjusted Mean ± Standard Deviation
Age (years)	80 ± 10	74 ± 12*	76 ± 12	78 ± 11*	79 ± 12	75 ± 11*	74 ± 12	80 ± 11*
Male (%)	43%	45%	41%	47%	42%	46%	52%	37%*
Diabetic (%)	22%	53%*	33%	36%	0%	100%	33%	38%
GFR (ml/min/1.73m2)	56 ± 24	57 ± 21	59 ± 17	54 ± 26*	58 ± 23	53 ± 21*	73 ± 18	40 ± 13*
Hemoglobin(g/dl)	11.9 ± 2.0	12.4 ± 2.2*	13.9 ± 1.4	10.7 ± 1.3*	12.1 ± 2.0	12.1 ± 2.3	12.6 ± 2.0	11.6 ± 2.1*
Body mass index (kg/m2)	25 ± 3	37 ± 7*	31 ± 8	29 ± 7	28 ± 6	34 ± 9*	31 ± 8	30 ± 8
Pulse Pressure (mmHg)	63 ± 19	67 ± 19*	61 ± 19	67 ± 18*	62 ± 19	69 ± 18*	61 ± 19	68 ± 19*
Cardiac structure and function	Adjusted Geometric Means
LV mass, g	182	218 †	192	200	190	201 ‡	191	200
End diastolic volume, ml	103	111 †	124	137 †	128	132	131	130
Relative wall thickness	0.45	0.48 †	0.46	0.46	0.47	0.46	0.45	0.46 ‡
Left atrial volume, ml	79	93 †	83	85	85	83	82	86
Ejection fraction, %	61	62 ‡	62	61	62	61	61	62 †
stroke volume, ml	76	86 †	76	83 †	79	81	79	80
cardiac output, l/min	5.36	5.86 †	5.23	5.90 †	5.58	5.53	5.65	5.46
cardiac power output, watts	1.1	1.24 †	1.1	1.21 †	1.16	1.15	1.16	1.14
Mitral E/e′	17.1	17.4	16.8	17.6	16	18.4 †	16.6	17.8
e′, m/sec	0.061	0.061	0.06	0.062	0.063	0.059	0.063	0.06
Deceleration time, ms	199	207	207	199	200	206	203	203
PASP, mmHg	48	46	49	45 ‡	47	47	45	49 †
Vascular structure and function
aortic area, mm2	908	1005 †	943	952	988	908 †	955	940
arterial elastance, mmHg/ml	1.53	1.41 †	1.55	1.42 †	1.46	1.50	1.46	1.51
vascular resistance, dyn•s•cm−5	1378	1293 ‡	1452	1246 †	1338	1351	1311	1379
arterial compliance, ml/mmHg	1.23	1.36 †	1.30	1.26	1.35	1.22 †	1.31	1.25

^*Unadjusted ANOVA p <0.05;

^†Adjusted p < 0.05,

^‡Adjusted p < 0.10

Figure 2.

Kaplan Meier survival curves in HFpEF according to the presence of obesity (A), anemia (B), diabetes (C) or renal dysfunction (D) with age, gender and concomitant comorbidity adjusted hazard ratios and 95% confidence intervals for obesity, anemia, diabetes or renal dysfunction (E).

Anemia

Anemia was present in 56% of HFpEF patients (Table 6). Anemic HFpEF patients were older, had worse renal function and higher PP than non-anemic HFpEF patients. Adjusting for age, gender, height, weight and other comorbidities, EDV and diastolic operating compliance (Table 4) were higher in anemic patients while other diastolic function parameters were similar to non-anemic HFpEF patients. SV, CO and CPO were higher and SVR and Ea were lower with no difference in SAC suggesting that the increased PP in anemic patients reflects the higher SV. Anemia was associated with poorer outcome adjusting for age and gender and other concomitant comorbidities (Figure 2 B and E).

Diabetes

Diabetes was present in 35% of HFpEF patients (Table 6). Diabetic HFpEF patients were younger, more obese and had worse renal function, and higher PP than non-diabetics. Adjusting for age, gender, height, weight and other comorbidities, diabetic HFpEF patients tended to have higher LV mass but had similar EDV and systolic function to non-diabetics. Diabetics had higher filling pressures (E/e′) but other diastolic parameters, including operating compliance (Table 4) were similar. Diabetic patients had smaller aortas and lower SAC than non-diabetics. While Ea and SVR were similar, PP was higher in diabetics before (p<0.001) and after adjustment for age, gender and other comorbidities (Table 6). The higher PP was also apparent after also adjusting for MAP and MAP² (p<0.001) and given the similar SV, these parameters suggest stiffer vasculature in diabetic HFpEF patients¹⁴. Diabetes was not associated with differences in outcome adjusting for age and gender and other concomitant comorbidities (Figure 2 C and E). As compared to non-diabetics, diabetic HFpEF patients were more frequently treated with beta blockers (58% vs 72%), angiotensin antagonists (40% vs 73%) and statins (37% vs 67%) (p<0.01 for all).

Renal dysfunction

HFpEF patients with renal dysfunction were older, more often female and had lower hemoglobin and pulse pressures than patients without renal dysfunction (Table 6). Adjusting for age, gender, height, weight and other comorbidities, EF and PASP were higher in renal dysfunction patients but all other ventricular-vascular properties were similar to those with better renal function (Tables 5 and 6). Renal dysfunction was associated with poorer outcome adjusting for age and gender and other concomitant comorbidities (Figure 2 D and E).

Discussion

In this large, prospectively identified, rigorously characterized HFpEF cohort, we found that as compared to age- and gender-matched normotensive and hypertensive controls, HFpEF patients consistently displayed abnormalities in ventricular-vascular properties above and beyond that explainable by comorbidity burden and body size. Among HFpEF patients, comorbidities were associated with unique clinical, structural, functional and prognostic profiles. Obese patients were younger, and while they displayed greater concentric remodeling, their LV systolic and vascular function and outcomes were better than non-obese HFpEF patients. Anemic patients were older, displayed ventricular and vascular characteristics consistent with a high(er) output state and had worse outcomes than non-anemic HFpEF patients. Diabetic patients were younger and heavier and had higher filling pressures and stiffer vasculature, yet they were aggressively treated and their outcomes were similar to non-diabetics. Patients with renal dysfunction were older and despite a lack of unique structural or functional features, they had worse outcomes than patients with better renal function. While these data confirm that comorbidities influence ventricular-vascular properties and outcomes among HFpEF patients, they underscore that fundamental disease-specific changes in cardiovascular function underlie this disorder and support the search for specific therapies.

Comorbidities and Ventricular and Vascular Structure and Function in HFpEF

We have previously characterized cardiovascular structure and function in a smaller cohort of HFpEF patients^{4, 7, 15}. The current analysis differs in several important ways. The prospectively-enrolled HFpEF cohort is larger and the comparator groups are matched for age and gender, whereas comparator populations in the previous studies were on average 10 (HTN) to 20 (CON) years younger than the HFpEF population. The current data utilizes more appropriate analyses to account for differences in body size, vascular structure is assessed, analyses are adjusted in the primary and sensitivity analyses for comorbidities, and potential differences among HFpEF patients according to the presence of key comorbidities are evaluated.

The current focus on the association between comorbidities and differences in ventricular-vascular properties between HFpEF and control populations or within HFpEF patients is important as comorbidities have significant effects on cardiovascular structure and function in non-HF populations and in animal models. Obesity has been linked to pathologic LVH, atrial enlargement, systolic and diastolic LV dysfunction, endothelial dysfunction, vasoconstriction and increased vascular stiffness^{11, 16}. Chronic severe anemia decreases SVR and leads to volume expansion, eccentric remodeling and LV systolic and diastolic dysfunction^{17, 18}. Diabetes has been linked to LVH, systolic and diastolic LV dysfunction, endothelial dysfunction and increased vascular stiffness^{19, 20}. Intrinsic renal disease is associated with apoptosis, fibrosis, hypertrophy and LV dysfunction^{21, 22}. Thus, the potential for comorbidities to account for the alterations in LV structure and function observed in HFpEF patients is real.

Here, with strict attention to matching for age, gender and hypertension, appropriate analyses for body size-related differences and adjustment for comorbidities, we demonstrate that HFpEF is independently associated with more severe alterations in ventricular-vascular properties than observed in healthy or hypertensive controls. Specifically, as compared to healthy controls, HFpEF patients had more cardiac remodeling (concentric LVH and larger atria), systolic (lower EF) and diastolic (higher E/e′ and PASP and lower e′ and DT) dysfunction, and more abnormal vascular function (higher Ea and SVR and lower SAC). As compared to hypertensive controls, HFpEF patients had more cardiac remodeling (concentric LVH and larger atria), systolic (lower EF, CO and CPO) and diastolic (higher E/e′ and PASP and lower e′, DT and operating compliance) dysfunction while vascular function was not consistently and only subtly more impaired in HFpEF than HTN.

Notably, adjusting for body size differences (Table 3 and 4) aortic size was similar across the CON, HTN and HFpEF groups. These data suggest that intrinsic stiffening of the aorta, rather than further degenerative expansion, contributes to vascular stiffening in HFpEF vs CON and HTN as previously suggested ²³.

It may be that HFpEF represents a unique synergistic interaction between the effects of aging, hypertension and comorbidities, particularly cardiovascular comorbidities, to promote ventricular and vascular remodeling and dysfunction but in the absence of a large myocardial infarction leading to reduced EF²⁴. If so, measures to ensure “healthy aging”, treatment of risk factors and prevention of other comorbidities would likely prevent or delay the onset of HFpEF. The factors that mediate progression from hypertension to HFpEF are not yet clear. Quantification of the severity, duration and control of hypertension over time is difficult and thus we are unable to determine if a greater lifetime burden of arterial hypertension mediates the more severe remodeling, ventricular dysfunction and vascular dysfunction associated with HFpEF as compared to age/sex matched HTN controls. As essential hypertension is a polygenic disease, there may be greater susceptibility to HFpEF in certain genetic subtypes or unique interactions with behavioral modulators.

Impact of comorbidities among HFpEF patients

It is notable that male, obese and diabetic HFpEF patients present at a younger age, underscoring the variability in the syndrome beyond the stereotypical profile of frail, elderly, hypertensive females. This may suggest that the interaction of aging and hypertension is sufficient to cause HFpEF in some patients, but that comorbidities may play a role in accelerating progression to HFpEF in others.

Obese HFpEF patients had higher LV mass, EDV and atrial size, better preserved systolic performance and more preserved vascular function as compared to non-obese HFpEF patients. As most of these parameters are related to body size, and particularly weight, determining whether these changes represent a physiologic adaptation to higher metabolic demands or pathologic remodeling is difficult. It is important to note that in non-HF cohorts, body size-independent vascular parameters are more abnormal in obese subjects and increases in height-scaled LV mass is associated with poorer outcomes in general population-based studies^{11, 12}. However, here, as compared to non-obese HFpEF patients, obese HFpEF patients had better survival as has been noted in both HFpEF and HFrEF^25–28. This “obesity paradox” has been ascribed to diagnosis bias (misdiagnosis of HF in obese patients), lead time bias (earlier symptom onset due to additional metabolic demands of obesity), improved nutritional status or an anti-inflammatory effect of obesity in HF^25–28. The characterization of ventricular-vascular parameters in obese and non-obese HFpEF patients in the current study is unique and the better preserved systolic and vascular function may provide insight into the better outcomes observed in obese HFpEF patients. In this community study, the U-shaped relationship between obesity and mortality reported in a large clinical trial HFpEF cohort²⁸ after adjustment for a large number of covariates was not observed, but our data set and sample size may limit the power to detect such a relationship.

Consistent with studies of severe anemia in animal models and humans without HF^{17, 29, 30}, the relatively mild anemia in HFpEF patients was associated with increased LV size and operating compliance, a high(er) output state and lower SVR and Ea, suggesting that anemia uniquely influences ventricular-vascular properties in HFpEF. Whether the well described association between anemia and mortality in HFpEF and HFrEF³¹ reflects the additional load and reduced oxygen carrying capacity related to anemia or the effects of the factors causing anemia etiology (iron deficiency, inflammation, hypervolemia) is not clear³². Of note, a recent study demonstrated improved functional and structural indices in HFpEF with carefully titrated erythropoieitin treatment³³. It is of interest that LV dilatation; higher CO and reduced arterial load (Ea) were seen in both anemic and obese as compared to non-anemic/non-obese HFpEF patients but with divergent associations with outcome. This may suggest relatively more physiologic load in obesity versus pathophysiologic load in anemia in HFpEF but such interpretations are speculative.

Diabetic HFpEF patients had higher filling pressures and evidence of greater arterial stiffness, consistent with invasive studies suggesting increased chamber and myocyte stiffness in HFpEF with diabetes³⁴ and animal and human studies demonstrating vascular dysfunction in diabetic, non-HF subjects¹⁹. The lack of association of diabetes with poorer outcomes is surprising given that diabetes was associated with worse outcomes in observational studies and clinical trials including HFpEF patients^{35,36,37–40}. However, variable strength of the association of diabetes with outcomes and variable associations of diabetes with outcomes according to HF age/etiology/comorbid conditions/treatment exist in the published studies. The older age of this HFpEF cohort, the more aggressive treatment of diabetic HFpEF patients, the un-quantified duration/severity/type of diabetes in the current study, the community setting and the high prevalence of obesity in diabetic subjects may have influenced our findings. Potentially, the lack of significant association of DM with worse outcomes relates to the presence of undiagnosed DM/glucose intolerance/metabolic syndrome in the (still overweight) “non-diabetic” HFpEF group.

HFpEF patients with renal dysfunction were older, had higher systolic, pulse and pulmonary artery systolic pressures, and higher EF but did not otherwise differ strikingly from those with better preserved renal function. The strong association of renal dysfunction, age and diabetes may confound adjustment and obscure structural and functional changes specific to renal dysfunction. Renal function is dynamic in HF patients and a single point assessment may not reflect the chronic state. None the less, renal dysfunction was associated with worse outcomes in HFpEF as previously described⁴¹.

Limitations

Ventricular and vascular function indices were estimated from brachial blood pressure and echocardiographic measurements. All measurements were resting and this may limit the identification of important differences in ventricular or vascular reserve function among groups⁴². Cause-specific mortality data were unavailable. As compared to clinical trial populations, all-cause mortality and rehospitalizations in observational studies such as this one are higher than that observed in clinical trial cohorts. However, the severity of HF relative to comorbidity severity may be less in observational studies. Some ventricular-vascular properties may be best scaled to lean body mass but neither lean body mass nor measures of body fat distribution were available.

Conclusion

While HFpEF occurs in elderly patients with multiple comorbidities, perturbations in cardiovascular structure and function in HFpEF are greater than can be attributed to comorbidities alone, and the search for specific therapies for the more advanced ventricular and vascular dysfunction present in HFpEF should continue. However, among HFpEF patients, comorbidities are associated with important differences in the clinical profile and ventricular-vascular properties which provide insight into their impact on the natural history of HFpEF.