Coronary Microvascular Dysfunction and Diffuse Myocardial Fibrosis Measured by Cardiovascular Magnetic Resonance are Characteristic of HFpEF

Abstract

Heart failure with preserved ejection fraction (HFpEF) is frequently accompanied by comorbidities and a systemic proinflammatory state, resulting in coronary microvascular dysfunction (CMD), as well as myocardial fibrosis. Recent studies have shown a high prevalence of CMD in HFpEF. The purpose of this study is to examine the relationship between myocardial perfusion reserve (MPR) and diffuse myocardial fibrosis in patients with HFpEF using CMR. A single center study was performed in 19 patients with clinical HFpEF and 15 healthy control subjects who underwent quantitative first-pass perfusion imaging to calculate global MPR. T1 mapping was used to assess fibrosis and to calculate extracellular volume (ECV). Spiral cine displacement encoded stimulated echo (DENSE) was used to calculate myocardial strain. Comprehensive 2D echocardiograms with speckle tracking, cardiopulmonary exercise testing, and BNP levels were also obtained. In patients with HFpEF, mean left ventricular (LV) EF was 61 ± 9% and LV mass index 45 ± 12 g/m². Compared to controls, HFpEF patients had reduced global MPR (2.29 ± 0.64 vs 3.38 ± 0.76, p = 0.002) and VO₂ max (16.5 ± 6.8 vs 30.9 ± 7.7 ml/kg*min, p < 0.001) while ECV (0.29 ± 0.04 vs 0.25 ± 0.04, p = 0.02), pulmonary artery systolic pressure (35.4 ± 13.7 vs 22.3 ± 5.4 mmHg, p = 0.004), and average E/e’ (15.0 ± 7.6 vs 8.6 ± 2.0, p = 0.005) were increased. DENSE peak systolic circumferential strain (p = 0.60) as well as echocardiographic derived global longitudinal strain (p = 0.07) were similar between both groups. The prevalence of CMD, defined as global MPR < 2.5, among the HFpEF group was 69%. In conclusion, HFpEF patients have a high prevalence of CMD and diffuse fibrosis. These parameters may be useful clinical endpoints for future therapeutic trials.

Background

Heart failure (HF) with preserved ejection (HFpEF) accounts for >50% of HF cases with a mortality similar to that of HF with reduced EF (HFrEF). HFpEF is frequently accompanied by systemic comorbidities such as diabetes, hypertension, and metabolic syndrome.^{1, 2} A systemic proinflammatory state may be involved in the pathophysiology of HFpEF.² Inflammation affects the coronary microvascular endothelium which results in vasoconstriction of the coronary microvascular bed in HFpEF patients.² Coronary microvascular dysfunction (CMD), assessed invasively,^{3, 4} non-invasively,^{5, 6} and on autopsy⁷ may be involved in the pathogenesis of HFpEF. A recent prospective, multi-center study of coronary flow reserve (CFR) in HFpEF found the prevalence of CMD was high (75%).¹

CMD assessment by quantitative stress perfusion cardiovascular magnetic resonance (CMR) has not been previously studied in HFpEF. Myocardial perfusion reserve (MPR) by CMR has been validated against invasive⁹ and noninvasive modalities (PET).¹⁰ It has also been shown to provide prognostic utility and can improve diagnostic performance over qualitative assessment of ischemia in CAD.^{11, 12}

CMR has the advantage over other modalities in that it can also quantify diffuse myocardial fibrosis which is a common endpoint of many cellular processes in HFpEF patients^{7, 14}. T1 mapping post-gadolinium and quantification of extracellular volume fraction (ECV) by CMR has been shown to be markers of diffuse myocardial fibrosis.^15-19 Thus, myocardial fibrosis and microvascular dysfunction measured by CMR may provide promising therapeutic targets in HFpEF.

The aim of our study was to quantify MPR and diffuse myocardial fibrosis by using CMR in patients with a clinical diagnosis of HFpEF. We hypothesized that these patients will have reduced MPR and increased myocardial fibrosis compared to age-matched controls. Furthermore, we hypothesize that MPR and fibrosis indices would be associated with echocardiographic evidence of diastolic dysfunction, exercise capacity, and peak VO2.

Methods

Between May 2015 and November 2018, patients with a confirmed clinical diagnosis of HFpEF were prospectively enrolled in this single center study. This study was performed under a locally approved IRB protocol (HSR# 17997), and all patients gave written informed consent before study enrollment. The following strict criteria were required for study inclusion: age 18–85, New York Heart Association (NYHA) functional class ≥ II or BNP ≥ 150 pg/mL, EF> 45%, and at least grade 1 diastolic dysfunction by echo or elevated pulmonary capillary wedge pressure on right heart catheterization. Exclusion criteria were history of known myocardial infarction, severe valvular heart disease, secondary causes of hypertension, infiltrative (i.e. amyloidosis, sarcoidosis) or hypertrophic cardiomyopathy, pericardial disease, primary pulmonary arterial hypertension, or previously reduced left ventricular (LV) ejection fraction (EF) with recovery of function. Subjects with contraindications to MRI such as metallic implants, severe claustrophobia, pacemakers/defibrillators, and estimated GFR < 45 mL/min/1.73 m² were also excluded. History of persistent atrial fibrillation with HR > 100 bpm were excluded due to concern regarding poor image quality. Patients with a contraindication to adenosine or who refused adenosine were excluded from stress myocardial perfusion imaging but were eligible for the remainder of the study, including native and contrast-enhanced T1 mapping. One patient screened for the study was excluded due to presence of previously unknown constrictive pericarditis.

Patients were compared to 15 age matched normal controls. Age-matched healthy controls, free of known cardiovascular disease were recruited. All subjects underwent a history and physical examination, electrocardiogram (ECG), cardiopulmonary exercise testing, comprehensive transthoracic echocardiogram (including strain), blood tests, and CMR.

CMR studies were performed on a 1.5 Tesla scanner (Avanto or Aera, Siemens Healthineers, Erlangen, Germany). Protocol included scout imaging followed by balanced steady-state free precession cine sequences to determine LV volumes, mass, and function. A short-axis stack of 8 mm thick short axis images with 2 mm gap covered the entire LV. Three long-axis images were obtained (2, 3, and 4-chamber views). A spiral cine displacement encoding with stimulated echoes (DENSE) pulse sequence was used to quantify peak systolic circumferential strain and early diastolic strain rate. ^{20, 24, 25} Base- and mid-ventricular short-axis modified Look-Locker inversion recovery (MOLLI) T1 mapping images, were then obtained prior to contrast administration, at 5- and 10 minutes after rest perfusion, and at 5-, 10-, and 15 minutes after stress perfusion imaging.

For stress perfusion assessment, 3 short axis slice locations were imaged per heart beat over a 60 heart beat acquisition during an IV bolus of 0.075 mmol/kg of gadolinium contrast (Magnevist, Bayer Healthcare, Whippany, New Jersey, USA) injected via power injector at 4 mL/s. Adenosine was infused at 140 μg/kg per minute through a peripheral IV line for 3–4 minutes during the stress imaging. Rest perfusion imaging was performed 15 minutes following stress imaging. Quantitative first-pass perfusion imaging was performed using a dual-sequence approach using a vendor-supplied works in progress (WIP) package. The saturation-recovery gradient echo pulse sequence acquires a low-resolution arterial input function (AIF) image, and 3 myocardial tissue function (TF) images every RR interval. Two proton density weighted images were acquired for the AIF and TF for correcting surface-coil related intensity inhomogeneity and Bloch simulation modeling was used to convert the signal intensity into gadolinium concentration units as previously described.^{26, 27}

Late gadolinium enhancement (LGE) using phase-sensitive inversion recovery pulse sequences were also obtained 5 minutes after resting perfusion using standard SCMR guideline protocols.²⁸

Cine CMR images were analyzed by an experienced investigator using QMASS (Medis Medical Imaging Systems, Leiden, the Netherlands). Short axis cine images were contoured for each short-axis slice, with total LV mass, end-diastolic volume, end-systolic volume, stroke volume and EF measured.

Perfusion quantification was performed on a pixel-wise basis using the constrained Fermi function deconvolution method implemented in MATLAB (Mathworks, Natick, Massachusetts, USA).¹³ MPR was calculated as a ratio of the stress perfusion divided by rest perfusion for each subject. Figure 1 illustrates the perfusion maps for a patient with reduced global MPR.

Figure 1.

Pixel based perfusion maps and bulls-eye plots at rest (top) and stress (bottom) for 3 short axis slices. Global MPR for this patient was: 2.19 ± 0.31.

T1 quantification was performed using manual segmentation of the myocardium in MATLAB. The partition coefficient of gadolinium, λ, was calculated from the slope of the linear fit of the plot of 1/T1 myocardium versus 1/T1 blood. ECV was calculated as λ*(1-hematrocrit).^{29, 30} Hematocrit was measured on the day of the CMR study. Peak systolic circumferential strain and early diastolic strain rates were computed offline from DENSE images with a custom MATLAB script using previously described methods.^{20, 31, 32}

All study participants underwent comprehensive 2-D with Doppler, tissue Doppler imaging, and speckle tracking using commercially available ultra sound systems (GE Healthcare, General Electric Corp., Waukesha, WI, USA, Philips Medical Systems). LV function, diastolic function, and global longitudinal strain (GLS) were quantified as recommended by the American Society of Echocardiography/European Association of Cardiovascular Imaging.^{33, 34} GLS was analyzed using vendor-independent speckle-tracking echocardiographic software (Tomtec 2D Cardiac Performance Analysis, Tom Tec Imaging Systems, Munich, Germany) on optimized (maximized frame rate, minimizing foreshortening) two-, three-, and four-chamber apical view images. If regional tracking was suboptimal in more than two myocardial segments in a single view then GLS calculation was not performed.

All exercise tests were performed at our center’s cardiopulmonary exercise laboratory. Metabolic measures were determined during a continuous progressive treadmill protocol (Cardiac Science treadmill, Waukesha, Wisconsin). VO₂ max was chosen as the highest one minute value obtained at volitional exhaustion. Respiratory exchange ratio (RER) ≥ 1.1, Borg Rating of Perceived Exertion (RPE) > 18, and age predicted maximal HR were also used as indicators of attaining VO₂ max.

Analysis was performed using SAS, version 9.4 (SAS Institute, Inc. Cary, North Carolina, USA). Continuous normally distributed data are presented as mean ± SD, and were compared using two-sided t-tests. Categorical data are presented as N and percentages and compared using chi-square analysis and the Fisher’s exact test where appropriate. For all statistics a p <0.05 was considered statistically significant. Pearson test was used to determine the correlation of different covariates with ECV and MPR.

Results

A total of 34 subjects were included in the study, 19 patients with HFpEF (63 ± 11 years, 42% female) and 15 healthy age-matched controls (59 ± 9 years, 60% female). A comparison of demographic and clinical characteristics between HFpEF and control group is summarized in Table 1. There was no significant difference in age, gender, systolic blood pressure, or history of tobacco use. Average heart rate and body mass index were significantly higher in the HFpEF group. There are notable differences in cardiovascular risk factors between the two groups. The HFpEF group had a significantly higher prevalence of hypertension (84%), hyperlipidemia (74%), and diabetes (58%). In addition, there were more African Americans in the HFpEF group (26%). As expected, there were more patients in the HFpEF group taking cardiovascular medications including diuretics, beta blockers, ACE-inhibitor or ARB, and statins. 95% of the HFpEF patients were on a loop-diuretic at the time of enrollment. BNP was elevated in HFpEF patients compared to normal controls.

Table 1.

Baseline Characteristic of HFpEF Patients and Controls

	HFpEF (n = 19)	Controls (n =15)	p-value
Age (years)	63 ± 11	59 ± 9	p = 0.22
Body Mass Index (kg/m2)	35 ± 7	27 ± 5	p < 0.001
Female (n,%)	8 (42)	9 (60)	p = 0.49
Heart Rate (bpm)	74 ± 14	64 ± 12	p = 0.03
SBP (mmHg)	128 ± 22	135 ± 11	p = 0.33
DBP (mmHg)	74 ± 14	71 ± 10	p = 0.41
African American	5 (26)	0 (0)	p = 0.05
Smoking (n,%)	2 (11)	0 (0)	p = 0.50
Hypertension (n,%)	16 (84)	3 (20)	p < 0.001
Hyperlipidemia (n,%)	14 (74)	1 (7)	p < 0.001
Diabetes (n,%)	11 (58)	0 (0)	p < 0.001
Loop diuretic (n,%)	18 (95)	1 (7)	p < 0.001
Beta Blocker (n,%)	14 (74)	0 (0)	p < 0.001
ACE-I or ARB (n,%)	14 (74)	3 (20)	p = 0.005
Statin (n,%)	14 (74)	1 (7)	p < 0.001
BNP (pg/ml)	135 ± 153	26 ± 37	p = 0.01
Hematocrit	39 ± 5	44 ± 4	p = 0.007

Values are presented as mean ± SD or n (%). ACEI = angiotensin converting enzyme inhibitors, ARB = angiotensin II receptor blocker, BNP = brain natriuretic peptide, DBP = diastolic blood pressure, and SBP = systolic blood pressure.

Table 2 summarizes the CMR, echocardiographic, and CPET test results for the HFpEF and control groups. There was no between group difference in LVEF, mass index, or end-diastolic volume index. 16 of 19 HFpEF patients and all 15 controls underwent stress perfusion CMR. 3 patients did not undergo stress, 1 opted not to undergo stress, and 2 had technical difficulties with perfusion. Global myocardial perfusion was lower in the HFpEF group (2.29 ± 0.64 vs 3.38 ± 0.76, p = 0.002). Furthermore, 69% of the HFpEF group had CMD, defined as a global MPR < 2.5. ECV was higher in the HFpEF group (0.29 ± 0.04 vs 0.25 ± 0.04, p = 0.02). However, native T1 was similar between groups 6/19 (32%) HFpEF patients had LGE, 3 patients had subendocardial LGE in a single vessel territory, and 3 had LGE in non-ischemic patterns (mid-wall and RV insertion site). DENSE peak systolic circumferential strain and early diastolic circumferential strain rate were similar between groups. Global longitudinal strain (GLS) by echo trended towards a decrease in GLS in the HFpEF group. All 15 patients in the control group had sufficient echo quality to perform strain analysis. Only 12 (63%) of patients in the HFpEF group met appropriate image quality for performing strain analysis. The HFpEF group had a significantly higher pulmonary artery systolic (PASP) (35.4 ± 13.7 vs 22.3 ± 5.4 mmHg, p = 0.004) and average E/e’ (15.0 ± 7.6 vs 8.6 ± 2.0, p = 0.005) but similar left atrial volume index. HFpEF group had a significantly lower VO₂max (16.5 ± 6.8 vs 30.9 ± 7.7 ml/kg*min, p < 0.001) and METS (4.7 ± 1.9 vs 8.8 ± 2.2 kcal/kg*h, p < 0.001) compared to the control group.

Table 2.

CMR, Echocardiogram, and CPET Parameters comparing HFpEF patients and Controls

	HFpEF (n = 19)	Controls (n =15)	p-value
LVEF (%)	61 ± 9	65 ± 5	p = 0.13
LVMI (g/m2)	45 ± 12	46 ± 11	p = 0.76
LVEDVI (ml/m2)	70 ± 14	70 ± 14	p = 0.96
Peak average circumferential strain	−0.15 ± 0.03	−0.16 ± 0.03	p = 0.60
Average Circumferential e'SR (s−1)	1.37 ± 0.65	1.26 ± 0.48	p = 0.59
Native T1 (ms)	1040 ± 74	1020 ± 43	p = 0.36
λ	0.48 ± 0.05	0.45 ± 0.05	p = 0.14
ECV	0.29 ± 0.04	0.25 ± 0.04	p = 0.02
Global Myocardial Perfusion Reserve*	2.29 ± 0.64	3.38 ± 0.76	p = 0.002
Stress Perfusion*	1.62 ± 0.52	1.78 ± 0.57	p = 0.43
Rest Perfusion*	0.73 ± 0.23	0.54 ± 0.19	p = 0.02
LA Volume Index (ml/m2)	34.5 ± 15.4	26.5 ± 7.2	p = 0.09
PASP (mmHg)	35.4 ± 13.7	22.3 ± 5.4	p = 0.004
Average E/e'	15.0 ± 7.6	8.6 ± 2.0	p = 0.005
GLS (%)	−17.9 ± 3.5	−20.5 ± 3.6	p = 0.07
VO2max (ml/kg*min)	16.5 ± 6.8	30.9 ± 7.7	p < 0.001
METS (kcal/kg*h)	4.7 ± 1.9	8.8 ± 2.2	p < 0.001

^*16/19 HFpEF patients and all control subjects underwent stress perfusion CMR.

Values are presented as mean ± SD. ECV = extracelluar volume, E/e' = early mitral inflow velocity/ mitral annular early diastolic velocity ratio, e'_SR = early diastolic strain rate, GLS = global longitudinal strain, LVEDVI = left ventricular end diastolic volume index, LVEF = left ventricular ejection fraction, LVMI = left ventricular mass index, METS = metabolic equivalents, PASP = pulmonary artery systolic pressure, VO₂max = maximum oxygen consumption, and λ = partition coefficient.

Correlation between MPR or ECV and different covariates are summarized in Table 3. VO₂ max was positively correlated with MPR while average E/e’ and PASP were negative correlated. PASP had the strongest negative correlation. VO2 max had negative correlation with ECV while PASP had a positive correlation. Average E/e’ had no correlation with ECV. In addition, a significant negative correlation between ECV and MPR can be seen in Figure 2.

Table 3.

Pearson Correlation Coefficients

	Pearson r	p
ECV
VO2max	−0.52	0.002
Average E/e'	0.34	0.057
PASP	0.51	0.006
MPR
VO2max	0.40	0.028
Average E/e'	−0.47	0.008
PASP	−0.60	0.001

ECV = extracelluar volume, E/e' = early mitral inflow velocity/ mitral annular early diastolic velocity ratio, METS = metabolic equivalents, PASP = pulmonary artery systolic pressure, and VO₂max = maximum oxygen consumption.

Figure 2.

Scatter plot of correlation between ECV and MPR.

Discussion

The present study demonstrates that patients with a clinical diagnosis of HFpEF have reduced MPR and increased myocardial fibrosis as compared to age matched controls. In addition, global MPR and ECV were inversely correlated. Patients in the HFpEF cohort had significantly reduced VO2 max, increased E/e’ and PASP on echo, and elevated BNP, thus confirming their clinical HFpEF. VO₂ max, average E/e’, and PASP also correlated with MPR with PASP having the highest negative correlation. The absence of LVH and known obstructive CAD underscores that the reduction in global MPR is not due to these factors. This suggests that CMD and diffuse fibrosis may play an important role in the pathophysiology of HFpEF and potentially be key targets for therapeutic interventions.

In this cohort, the prevalence of CMD, defined as global MPR < 2.5,^{1, 35} was high (69%). This is consistent with the high (75%) prevalence of CMD reported in a recent large multi-center study of coronary flow reserve (CFR) assessment using adenosine stress transthoracic Doppler echocardiography in patients with HFpEF.¹ This adds to the emerging literature suggesting that HFpEF is a systemic disorder associated with endothelial dysfunction.^1-7 This is the first study to assess CMD in HFpEF by CMR first pass perfusion imaging. Fully quantitative CMR has recently identified reduced MPR in patients with typical angina and risk factors for CMD who had no obstructive coronary artery disease.¹³

CMR has the advantage over other modalities in that in can also quantify diffuse myocardial fibrosis. ECV has been shown to be elevated in patients with LVH.²⁰ ECV has also recently been studied in HFpEF patients and found to help differentiate them from hypertensive heart disease.²¹ Despite our cohort demonstrating no hypertrophy, ECV was still elevated compared to the control group. Interestingly our study showed that measurements of myocardial strain, both by CMR and echocardiography, were not reduced in the HFpEF group, although there was a trend towards decreased GLS (p = 0.07). A recent study among patients with HFpEF, hypertensive patients, and healthy controls showed that although both GLS and ECV could discriminate between hypertensive heart disease and HFpEF, ECV was a better discriminatory marker of HFpEF (AUC 0.88) than GLS (AUC 0.78)²¹. Their cohort had significantly less diabetics compared to the present study (9.7% vs 58%) and higher LV mass index (70.8 ± 20.2 vs 45 ± 12 g/m²) suggesting that in a HFpEF cohort predominated by diabetic patients without LVH, increase in ECV occurs before impairment in strain. Furthermore, this study illustrates the challenge of acquiring high quality images for speckle tracking assessment of GLS in HF patients which was only achievable in <2/3 of the HFpEF cohort.

Multiple techniques have been used to assess microvascular function.³⁵ The advent of noninvasive techniques increase the feasibility of diagnosing CMD without the associated risk of catheter-based techniques. Doppler echocardiography techniques rely on the quality of acoustic windows and are not often used in the clinical setting. PET perfusion imaging has become a gold standard; however CMR has the added advantage of including fibrosis and strain imaging. The global MPR measured in our cohort (2.29 ± 0.64) was very similar to measurements by PET (2.16 ± 0.69) in a cohort of HFpEF patients without known obstructive epicardial CAD.⁵ Our study is consistent with recent data that less than one third of patients with HFpEF don’t have evidence of CMD.¹ Future studies are needed to assess if MPR can be used to assess the stage of disease progression in HFpEF. Clinical trials targeting fibrosis and perfusion in the HFPEF population are needed.

This study is limited by a relatively small sample size; however, this is the only study to date evaluating quantitative CMR first pass perfusion imaging in HFpEF. Furthermore the study was adequately powered to highlight important differences in CMR, echocardiographic, and CPET parameters in a cohort with pure HFpEF (no LVH, no infiltrative cardiomyopathies and normal EF).

In conclusion, we have demonstrated by quantitative CMR perfusion imaging that HFpEF patients without LVH and with normal myocardial strain, have diffuse fibrosis and a high prevalence of CMD. Comprehensive evaluation of perfusion, fibrosis, and strain by CMR may become important endpoints for therapeutic efficacy in HFpEF.