Prognostic Significance of Peripheral Microvascular Endothelial Dysfunction in Heart Failure With Reduced Left Ventricular Ejection Fraction

Abstract

Background: Endothelial dysfunction plays a crucial role in heart failure (HF), but the association between peripheral microvascular endothelial function assessed by reactive hyperemia peripheral arterial tonometry (RH-PAT) and prognosis remains unknown in HF with reduced left ventricular (LV) ejection fraction (HFREF). We prospectively investigated the association between peripheral microvascular endothelial function and HF-related near-future cardiovascular outcomes in HFREF patients.

Methods and Results: The 362 HFREF patients (LVEF <50%) were followed for HF-related events (composite of cardiovascular death and HF hospitalization) up to 3 years. A natural logarithmic-scaled RH-PAT index (Ln-RHI) was obtained for each patient. A total of 82 HF-related events were recorded. The lower-RHI group (Ln-RHI ≤0.49, median) experienced a higher rate of HF-related events compared with the higher-RHI group by Kaplan-Meier analysis (30.9% vs. 14.4%, log-rank test: P<0.001). Multivariable Cox hazard analysis identified Ln-RHI as an independent predictor for HF-related events (per 0.1, hazard ratio: 0.84, 95% confidence interval: 0.75–0.95, P=0.005). Adding Ln-RHI to the Meta-analysis Global Group in Chronic HF risk score (MAGGICs) and Seattle Heart Failure Model (SHFM), powerful prognostic predictors of HF, significantly improved the net reclassification index (MAGGICs: 20.11%, P=0.02, SHFM: 24.88%, P<0.001), and increased the C-statistics for prediction of HF-related events (MAGGICs+Ln-RHI: from 0.612 to 0.670, SHFM+Ln-RHI: from 0.662 to 0.695).

Conclusions: Peripheral microvascular endothelial dysfunction assessed by RH-PAT was associated with future HF-related events in HFREF. (Circ J 2015; 79: 2623–2631)

The prevalence of heart failure (HF), and the incidence of HF-associated deaths and hospitalizations are increasing in aging populations. Many studies have reported evidence for the effect of drugs, cardiac rehabilitation, and therapeutic devices in HF, but they have had limited efficacy.¹ Worldwide, almost 23 million people experience HF and have poor prognosis.² Some patients experience a rapid recurrence of symptoms of HF despite optimal treatment. Several factors (eg, sex, biomarkers including B-type natriuretic peptide (BNP), New York Heart Association (NYHA) class, renal function) have been well investigated for the determination of risk and prognosis of HF.^3–8 However, they are not sufficient for practical risk stratification and we need to identify new pathophysiologic factors involving in prognosis in HF.

Endothelial dysfunction is associated with the pathogenesis and progression of HF.⁹ The endothelium produces various factors, including nitric oxide (NO), and maintains vascular tone, contractility of cardiomyocytes, and perfusion of the tissues. Endothelial dysfunction increases vascular stiffness and resistance, and decreases tissue perfusion, leading to organ dysfunction and an increase in the afterload of the heart.¹⁰ Endothelial dysfunction in conduit arteries and microvasculature could occur under similar pathogenic conditions. Microvascular NO, prostaglandin, and endothelium-derived hyperpolarizing factor collaboratively regulate the tissue circulation of systemic organs, whereas reduced NO in conduit arteries is associated with atherothrombosis followed by endothelial dysfunction caused by systemic and local factors (inflammation or shear stress, etc).^11,12 In HF, we need to focus on the microcirculatory dysfunction related to cardiac pump failure and following multiple organ dysfunction rather than arterial occlusive changes (atherothrombosis) in larger arteries. Several studies using flow-mediated vasodilation (FMD) have reported that endothelial dysfunction is associated with symptom severity and clinical outcomes in patients with HF.^13–15 FMD of the brachial arteries is a widely used technique to noninvasively measure endothelial function. FMD measures the respective diameter changes in response to shear stress of the brachial artery (conduit artery) as reactive vasodilation to hyperemia. However, this test is complex and provides limited quantitative data.¹⁶ The association between peripheral microvascular endothelial function and cardiovascular prognosis still remains unclear in HF with reduced left ventricular (LV) ejection fraction (HFREF). Reactive hyperemia-peripheral arterial tonometry (RH-PAT) is an alternative test that can noninvasively assess peripheral microvascular endothelial function by examining the pulse volume changes in the distal finger as reactive vasodilation to actual hyperemia in the microvasculature.¹¹ The RH-PAT data are digitally analyzed by a computer in an operator-independent manner. Previous studies have demonstrated good reproducibility of RH-PAT measurements.^17,18 In the Framingham Heart Study, RH-PAT was not statistically associated with FMD when fully adjusted by coronary risk factors or cardiovascular disease, and the researchers concluded that each method reflected different pathophysiologic responses to cardiovascular risk factors.¹⁹ The target vasculature is essentially different between RH-PAT and FMD.¹²

In current clinical settings, only RH-PAT examination can noninvasively provide a practical assessment of peripheral microvascular dysfunction. We focused on impairment of peripheral microcirculation, but not large conduit vessel circulation, in the pathogenesis of HFREF.¹¹ In the present study, we prospectively investigated the association between peripheral microvascular endothelial function, volumetrically assessed using RH-PAT, and HF-related near-future cardiovascular outcomes in patients with HFREF.

Methods

Study Population

This study included patients with HFREF admitted to the Kumamoto University Hospital and Yokohama City University Medical Center between September 2006 and December 2012. HFREF was established according to the diagnostic criteria for HF of the Framingham Heart Study,²⁰ with LVEF <50% assessed by transthoracic echocardiography. We excluded patients with severe HF of NYHA class IV and patients with acute coronary syndrome who required emergency coronary angiography, defined as either acute myocardial infarction (MI) or class II or III unstable angina by Braunwald’s classification. We also excluded patients with advanced collagen disease, endocrine disease, active inflammatory disease, severe liver dysfunction, neoplasms, hemiplegia, and those undergoing hemodialysis.

This study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Human Ethics Review Committee of Kumamoto University Graduate School of Medicine and the Yokohama City University Institutional Review Board, and a signed consent form was obtained from each subject.

Study Protocol

When patients were in a compensated and stable condition after guideline-recommended therapy for HF, RH-PAT was performed in a comfortable environment in the early morning before administration of any daily medications before discharge. We determined physical status, including NYHA class, and performed laboratory tests, such as BNP measurement, at the same time.

RH-PAT

The principle of RH-PAT has been described previously.¹⁷ Briefly, it noninvasively measures blood volume changes that accompany pulse waves in the distal finger. Data are automatically analyzed by computer software in an operator-independent manner (Endo-PAT2000; Itamar Medical, Caesarea, Israel, software versions 3.0.4 and 3.4.4). Because the RH-PAT index (RHI) shows a skewed distribution, the natural logarithmic-scaled RHI was calculated and expressed as Ln-RHI (see also Supplementary File 1, Figures S1, S2).

HF Risk Models

The integer risk score for HF of the Meta-analysis Global Group in Chronic Heart Failure (MAGGICs) is a model with multivariable risk assessment based on 30 cohort studies, and provides a strong indication of HF patients’ mortality risk over 3 years.⁵ The detailed method for calculation of this score has been described previously.⁵ The Seattle Heart Failure Model (SHFM) is another HF risk model that predicts mortality of HF patients.²¹ We evaluated whether the addition of Ln-RHI to MAGGICs or SHFM improved risk stratification in HFREF (see also Supplementary File 1).

Follow-up Study and Outcomes

After the assessment of RHI, patients with HFREF were followed prospectively until August 2013 or an endpoint occurred. The primary endpoint was a HF-related event, including cardiovascular death or hospitalization for worsening HF. All patients were followed up at the outpatient clinics or by primary care physicians every month, and by annual telephone contact with each patient. HF-related events were reported by the physicians and confirmed from a review of medical records and direct contact with the patients, their families, and physicians. An Events Committee comprising 3 independent physicians reviewed all events. Cardiovascular death was defined as death within 30 days of MI, death from HF, documented sudden death without evident noncardiovascular cause, or death from stroke. Hospitalization for HF decompensation was defined if the patient was admitted for at least an overnight stay in hospital because of HF with typical symptoms and had objective signs of worsening HF requiring intravenous drug administration.

Statistical Analysis

Data are expressed as mean±standard deviation for normally distributed variables according to the Shapiro-Wilk test, and those with non-normally skewed distributions are expressed as the median value with interquartile range (IQR). Categorical data are presented as frequencies and percentages. Differences between groups were determined using Fisher’s exact test for categorical variables. Differences in continuous variables were analyzed by the unpaired t-test or the Mann-Whitney U test, as appropriate. Regression adjustment by propensity scores is useful to address a large number of factors to a small number of events per variable, and can reduce the selection bias and confounders. The propensity score is comprehensively calculated by many potential covariates, and can be used as an independent variable to derive adjusted estimates in regression models.^22,23 The propensity score was calculated for each patient using a logistic regression model, with higher Ln-RHI as the dependent variable. The Kaplan-Meier method was used to estimate the probability of HF-related events up to 3 years after the RHI assessment, with the lower-RHI and higher-RHI groups dichotomized by the median cut-off value of Ln-RHI (0.49). The log-rank test was used to compare distributions of event-free times between the higher-RHI group and the lower-RHI group. The Cox proportional hazards model was used to estimate HF-related events hazard ratios (HR) by univariate and multivariable analyses with forced inclusion modeling including Ln-RHI as a continuous variable. The HR and 95% confidence intervals (CI) are presented. The estimates of C-statistics in the Cox proportional hazards regression models were compared after the addition of Ln-RHI to MAGGICs or SHFM. We also assessed the incremental effects of adding Ln-RHI to MAGGICs or SHFM to predict HF-related events using the net reclassification index (NRI) and integrated discrimination improvement (IDI). The NRI demonstrates the global net improvement in reclassification by assessing the shift of an individual’s clinical risk category (low, moderate, high) with a new model. The IDI assesses the change in the estimation prediction probabilities as a continuous variable with and without the new factors. The IDI assessment does not require the categories for models.²⁴ To assess reclassification improvement, we defined 3 risk categories among the population, based on the tertile distribution of MAGGICs (low risk: <17.5%; intermediate risk: 17.5–29.2%; high risk: ≥29.2%) and SHFM (low risk: >92.4%; intermediate risk: 85.9–92.4%; high risk: ≤85.9%).⁵ Power analysis was performed to estimate the required number of patients based on previous study.¹³ The number of enrolled patients in this study (n=362) was appropriate compared with the estimated number by power analysis (required sample size n=253, α error 0.05, power 0.9, event-free rate 64.2%, HR 0.665). A P-value <0.05 was considered statistically significant. Statistical analyses were performed using SPSS 20 for Macintosh (SPSS Inc, Tokyo, Japan), SAS version 9.1.3 program for Windows (SAS Institute Inc, Cary, NC, USA), and STATA version 11.0 (Stata Corp, College Station, TX, USA). (See also Supplementary File 1.)

Results

Clinical Characteristics of the Study Population

Baseline characteristics of the enrolled patients with HFREF (n=362) are shown in Tables 1, S1. Ischemic heart disease accounted for 46% of the population.

Table 1. Baseline Characteristics of Study Group of Patients With HFREF

	Total (n=362)
Age (years)	65.9±12.2
Male, n (%)	252 (70)
Body mass index (kg/m2)	23.7±4.0
NYHA functional class II and III	234 (65)
Etiology of HF
Ischemic heart disease, n (%)	167 (46)
Dilated cardiomyopahty, n (%)	93 (26)
Other, n (%)	102 (28)
Hypertension, n (%)	235 (65)
Dyslipidemia, n (%)	232 (64)
Diabetes mellitus, n (%)	124 (34)
Current smoker, n (%)	72 (20)
Coronary artery disease, n (%)	167 (46)
Atrial fibrillation, n (%)	83 (23)
COPD, n (%)	14 (4)
Heart rate (beats/min)	70.1±13.8
Systolic blood pressure (mmHg)	117.0±19.6
Diastolic blood pressure (mmHg)	70.0±13.2
Hemoglobin (g/dl)	13.2±2.1
Serum sodium (mEq/L)	139.1±3.4
Fasting plasma glucose (mg/dl)	94.0 (85–113)
Hemoglobin A1c (%)	6.0 (5.7–6.6)
Total cholesterol (mg/dl)	168.0 (144.8–197.3)
HDL cholesterol (mg/dl)	49.0 (39.0–58.0)
LDL cholesterol (mg/dl)	99.0 (81.0–127.0)
Triglycerides (mg/dl)	102.0 (76.0–139.0)
Uric acid (mg/dl)	6.4 (5.2–7.6)
eGFR (ml/min/1.73 m2)	58.9±19.8
BNP (pg/ml)	189.1 (80.9–488.0)
LVEF (%)	38.3 (30.8–45.8)
Cardiac output (L/min)	3.5 (2.8–4.2)
E/e’	14.0 (10.8–19.0)
Left ventricular mass index (g/m2)	162.8 (132.0–208.3)
Left atrium diameter (mm)	41.0 (36.0–46.5)
Medications
β-blocker, n (%)	296 (82)
ACEI or ARB, n (%)	327 (91)
Diuretic, n (%)	217 (60)
Aldosterone antagonist, n (%)	154 (43)
Calcium-channel blocker, n (%)	88 (24)
Statin, n (%)	192 (53)
Antidiabetic drugs, n (%)	81 (22)
Allopurinol, n (%)	78 (22)
ICD, n (%)	13 (4)
CRT-D, n (%)	16 (4)
MAGGICs (%)	25.4±13.1
SHFM (%)	86.2±11.8

Data are mean±standard deviation (SD), median (25–75th percentile), or number (%). ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BNP, B-type natriuretic peptide; COPD, chronic obstructive pulmonary disease; CRT-D, cardiac resynchronization therapy defibrillator; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; HF, heart failure; HFREF, HF with reduced left ventricular ejection fraction; ICD, implantable cardioverter-defibrillator; LDL, low-density lipoprotein; LVEF, left ventricular ejection fraction; MAGGICs, Meta-Analysis Global Group in Chronic Heart Failure integer risk score; NYHA, New York Heart Association; SHFM, Seattle Heart Failure Model.

HF-Related Events in the Lower- and Higher-RHI Groups

Data of 362 patients were available for analysis of HF-related events. There was no loss to follow-up. The median follow-up period was 32.0 (IQR: 16.0–37.0) months. Overall, 82 HF-related events were observed during the follow-up period: 12 cardiovascular deaths and 70 patients hospitalized for HF decompensation (Table S2). None of patients who received a LV assist device or heart transplant were observed. Ln-RHI was significantly lower in the HF-related events group compared with the event-free group (0.42±0.18 vs. 0.54±0.21, P<0.001). The frequency of HF-related events was significantly higher in the lower-RHI group (cut-off value: 0.49) compared with the higher-RHI group (30.9% vs. 14.4%, P<0.001, Table S2, Figure A). Hospitalization for HF decompensation was also significantly higher in the lower-RHI group than in the higher-RHI group (27.1% vs. 11.6%, P<0.001, Table S2). The lower-RHI group had a significantly higher probability of HF-related events than the higher-RHI group according to the Kaplan-Meier analysis (log-rank test: P<0.001, Figure A). We also divided the whole population into 2 groups according to the cut-off level of BNP 200 pg/ml. Figure B shows the Kaplan-Meier curves for cardiovascular events among the 4 subgroups according to the combination of BNP and RHI. The combination of Higher-BNP and Lower-RHI had significantly higher probability of cardiovascular events. Even in the lower-BNP HF patients, a lower-RHI significantly demonstrated higher probability of HF-related events (Figure B).

Figure.

Kaplan-Meier analysis of the probability of future events in patients with HFREF. (A) Probability of heart failure-related events in the lower-RHI and higher-RHI groups, and (B) heart failure-related events when combining RHI and BNP, (C) combining RHI and MAGGICs, and (D) combining RHI and SHFM in patients with HFREF. BNP, B-type natriuretic peptide; HFREF, heart failure with reduced left ventricular ejection fraction; MAGGICs, the integer risk score for heart failure of the Meta-analysis Global Group in Chronic Heart Failure score; RHI, reactive hyperemia peripheral arterial tonometry index; SHFM, Seattle Heart Failure Model.

We also divided the whole population into 2 groups according to a MAGGICs cut-off value of 19.1% (the lower- and higher-MAGGICs groups), which was suggested for low-risk patients in the MAGGICs study.⁵ Figure C shows the Kaplan-Meier curves for HF-related events among the 4 subgroups stratified by a combination of MAGGICs and RHI. The combination of a higher-MAGGICs and a lower-RHI was associated with a significantly higher probability of HF-related events (P<0.001 by log-rank test, Figure C). The combination of a higher-SHFM and a lower-RHI was also associated with a significantly higher probability of HF-related events (P<0.001 by log-rank test, Figure D). Even for lower-MAGGICs or lower-SHFM HF patients, the lower-RHI group demonstrated a significantly higher probability of HF-related events (Figures C,D).

Cox Proportional Hazards Analysis for HF-Related Events

The results of the univariate and multivariable Cox proportional hazards analysis for HF-related events are shown in Table 2A. Univariate Cox proportional hazards analysis identified systolic blood pressure (HR: 0.98, 95% CI: 0.97–0.99, P=0.03), estimated glomerular filtration rate (HR: 1.54, 95% CI: 1.01–2.34, P=0.045), serum sodium level (HR: 0.88, 95% CI: 0.84–0.92, P<0.001), Ln[BNP] (per 1.0, HR: 1.64, 95% CI: 1.37–1.96, P<0.001), LVEF (HR: 0.96, 95% CI: 0.94–0.98, P<0.001), and Ln-RHI (per 0.1, HR: 0.78, 95% CI: 0.70–0.87, P<0.001) as significant factors associated with HF-related events. Multivariable Cox proportional hazards analysis using these 6 significant factors from the univariate analysis identified Ln-RHI (per 0.1, HR: 0.85, 95% CI: 0.75–0.95, P=0.005), serum sodium level (HR: 0.92, 95% CI: 0.87–0.98, P=0.004), and Ln[BNP] (per 1.0, HR: 1.38, 95% CI: 1.12–1.70, P=0.002) as the independent predictors for HF-related events (Table 2A). Previous studies have reported that LV systolic or diastolic dysfunction and left atrial size are associated with clinical outcome in HF. We assessed a multivariable Cox model consisting of echocardiographic parameters (LVEF, E/e,’ left atrial diameter (LAD), and cardiac output) and Ln-RHI. Univariate Cox proportional hazards analysis identified LVEF (HR: 0.96, 95% CI: 0.94–0.98, P<0.001), E/e’ (HR: 1.05, 95% CI: 1.03–1.08, P<0.001) and LAD (HR: 1.05, 95% CI: 1.02–1.08, P=0.001) as significant factors associated with HF-related events. Multivariable Cox proportional hazards analysis identified Ln-RHI (HR: 0.82, 95% CI: 0.72–0.94, P=0.003), E/e’ (HR: 1.04, 95% CI: 1.01–1.06, P=0.03), and LAD (HR: 1.05, 95% CI: 1.02–1.08, P=0.004) as the independent and significant predictors for HF-related events (Table S3). Univariate Cox proportional hazard analysis identified MAGGICs as a significant factor for HF-related events (HR: 1.03, 95% CI: 1.01–1.04, P<0.001). Multivariable Cox proportional hazard analysis by the forced inclusion model with MAGGICs and Ln-RHI identified that each factor independently and significantly predicted future HF-related events (MAGGICs, HR: 1.02, 95% CI: 1.01–1.04, P=0.004; Ln-RHI (per 0.1), HR: 0.80, 95% CI: 0.72–0.89, P<0.001, Table 2B). Multivariable Cox proportional hazard analysis using the forced inclusion model with SHFM and Ln-RHI identified that each factor independently and significantly predicted future HF-related events (Ln-RHI (per 0.1), HR: 0.83, 95% CI: 0.74–0.93, P=0.002; SHFM, HR: 0.96, 95% CI: 0.94–0.97, P<0.001, Table 2C). Estimated HR of Ln-RHI adjusted by propensity score was 0.86 per 0.1 (95% CI: 0.76–0.98, P=0.02). We evaluated the effect of nitrate and nitroglycerin on the results. Ln-RHI did not differ between the patients with or without nitrate or nitroglycerin use (nitrate or nitroglycerin use: 0.51±0.19, no nitrate or nitroglycerin use: 0.52±0.21, P=0.98). The relative risk according to Ln-RHI between patients with and without nitrate or nitroglycerin use was similar in this interaction analysis (P for interaction=0.47) (Figure S3).

Table 2. (A) Cox Proportional Hazard Analysis for HF-Related Events in Patients With HFREF, (B) Cox Proportional Hazard Analysis With MAGGICs and Ln-RHI for for HF-Related Events in Patients With HFREF, (C) Cox Proportional Hazard Analysis With SHFM and Ln-RHI for HF-Related Events in Patients With HFREF

Variable	Simple regression			Multiple regression using significant factors in the univariate analysis
	HR	95% CI	P value	HR	95% CI	P value
(A)
Age (per year)	1.02	1.00–1.04	0.06	–	–	–
Male (yes)	1.30	0.80–2.13	0.29	–	–	–
Body mass index <25 kg/m2 (yes)	0.62	0.38–1.03	0.07	–	–	–
NYHA functional class II or III vs. I	1.44	0.90–2.32	0.13	–	–	–
Diabetes mellitus (yes)	1.51	0.97–2.34	0.07	–	–	–
Dyslipidemia (yes)	0.89	0.57–1.40	0.62	–	–	–
Hypertension (yes)	0.75	0.48–1.17	0.20	–	–	–
Current smoker (yes)	1.00	0.58–1.73	1.00	–	–	–
Ischemic etiology (yes)	0.73	0.46–1.13	0.16	–	–	–
Atrial fibrillation (yes)	1.49	0.92–2.39	0.10	–	–	–
Heart rate (per beats/min)	1.01	0.99–1.02	0.34	–	–	–
Systolic blood pressure (per mmHg)	0.98	0.97–0.99	0.03	0.99	0.98–1.00	0.14
Hemoglobin <10.0 g/dl (yes)	1.03	0.45–2.37	0.94	–	–	–
eGFR <60 ml/min/1.73 m2 (yes)	1.54	1.01–2.34	0.045	1.18	0.75–1.86	0.47
Serum sodium (per mEq/L)	0.88	0.84–0.92	<0.001	0.92	0.87–0.98	0.004
Ln(BNP) (per 1.0)	1.64	1.37–1.96	<0.001	1.38	1.12–1.70	0.002
LVEF (per %)	0.96	0.94–0.98	<0.001	1.00	0.98–1.02	0.91
Cardiac output (per L/min)	0.90	0.71–1.14	0.40	–	–	–
Ln-RHI (per 0.1)	0.78	0.70–0.87	<0.001	0.84	0.75–0.95	0.005
(B)
Ln-RHI (per 0.1)	0.78	0.70–0.87	<0.001	0.80	0.72–0.89	<0.001
MAGGICs (per 1.0)	1.03	1.01–1.04	<0.001	1.02	1.01–1.04	0.004
(C)
Ln-RHI (per 0.1)	0.78	0.70–0.87	<0.001	0.83	0.74–0.93	0.002
SHFM (per 1.0)	0.95	0.94–0.96	<0.001	0.96	0.94–0.97	<0.001

95% CI, 95% confidence interval; HR, hazard ratio; Ln-RHI, logarithmic-scaled reactive hyperemia peripheral arterial tonometry index. Other abbreviations as in Table 1.

C-Statistics and NRI for the Cox Hazard Model to Predict HF-Related Events

We estimated the C-statistic for MAGGICs alone (0.612, 95% CI: 0.535–0.689). Separate combinations of Ln-RHI with MAGGICs increased the C-statistic for HF-related events prediction (from 0.612 to 0.670, Table 3A). Good calibration for the analysis was confirmed by Grønnesby and Borgan statistics (P=0.91). The value of the C-statistic for MAGGICs with Ln (BNP) was 0.647 and 95% CI was 0.572–0.721. Separate combinations of MAGGICs and Ln (BNP) with Ln-RHI increased the C-statistic for HF-related events prediction from 0.647 to 0.689 (Table 3A). Good calibration for the analysis was confirmed by Grønnesby and Borgan statistics (P=0.114). We also reclassified the risk for the study patients using MAGGICs and Ln-RHI. The NRI and IDI showed the significance of inclusion of Ln-RHI as a continuous variable (NRI: 9.76% for patients with HF-related events, 10.36% for those without HF-related events, and 20.11% overall, P=0.02; IDI: 5.29%, P<0.001, Table 3A). We also estimated the C-statistics for SHFM alone (0.662, 95% CI: 0.571–0.753). Separate combinations of Ln-RHI with SHFM increased the C-statistic for HF-related events prediction (from 0.662 to 0.695, Table 3B). Good calibration for the analysis was confirmed by Grønnesby and Borgan statistics (P=0.558). The value of the C-statistic for SHFM with Ln (BNP) was 0.680 and 95% CI was 0.592–0.767. Separate combinations of SHFM and Ln (BNP) with Ln-RHI increased the C-statistic for HF-related events prediction from 0.680 to 0.707 (Table 3B). Good calibration for the analysis was confirmed by Grønnesby and Borgan statistics (P=0.076). We also reclassified the risk for the study patients using SHFM and Ln-RHI. The NRI and IDI showed the significance of inclusion of Ln-RHI as a continuous variable (NRI: 4.88% for patients with HF-related events, 20.0% for those without HF-related events, and 24.88% overall, P<0.001; IDI: 3.94%, P<0.001, Table 3B).

Table 3. (A) C-Statistics and Net Reclassification Improvement for Cox Hazard Model to Predict HF-Related Events in Patients With HFREF by Addition of Ln-RHI to MAGGICs, (B) C-Statistics and Net Reclassification Improvement for Cox Hazard Model to Predict HF-Related Events by Addition of Ln-RHI to SHFM

(A)
C-statistics for Cox proportional hazard model to predict HF-related events
HFREF patients (n=362)	C-statistic (95% CI)		Increment in C-statistic
MAGGICs	0.612 (0.535–0.689)
MAGGICs+Ln-RHI	0.670 (0.605–0.736)		0.058 (−0.007 to 0.123)
MAGGICs+Ln-BNP	0.647 (0.572–0.721)
MAGGICs+Ln-BNP+Ln-RHI	0.689 (0.626–0.751)		0.042 (−0.009 to 0.093)
NRI models to predict HF-related events
Risk category by MAGGICs	Classification by MAGGICs alone	New risk category using MAGGICs+Ln-RHI
		Low risk	Intermediate risk	High risk
Patients without HF-related events
Low risk	87 (31.07%)	61 (21.79%)	23 (8.21%)	3 (1.07%)
Intermediate risk	139 (49.64%)	51 (18.21%)	61 (21.79%)	27 (9.64%)
High risk	54 (19.29%)	2 (0.71%)	29 (10.36%)	23 (8.21%)
Patients with HF-related events
Low risk	10 (12.20%)	4 (4.88%)	5 (6.10%)	1 (1.22%)
Intermediate risk	50 (60.98%)	10 (12.20%)	21 (25.61%)	19 (23.17%)
High risk	22 (26.83%)	0 (0.00%)	7 (8.54%)	15 (18.29%)
(B)
C-statistics for Cox proportional hazard model to predict HF-related events
HFREF patients (n=362)	C-statistic (95% CI)		Increment in C-statistic (95% CI)
SHFM	0.662 (0.571–0.753)
SHFM+Ln-RHI	0.695 (0.621–0.769)		0.033 (−0.012 to 0.077)
SHFM+Ln-BNP	0.680 (0.592–0.767)
SHFM+Ln-BNP+Ln-RHI	0.707 (0.635–0.779)		0.027 (−0.019 to 0.073)
NRI models to predict HF-related events
Risk category by SHFM	Classification by SHFM alone	New risk category using SHFM+Ln-RHI
		Low risk	Intermediate risk	High risk
Patients without HF-related events
Low risk	8 (2.86%)	6 (2.14%)	1 (0.36%)	1 (0.36%)
Intermediate risk	42 (15.0%)	12 (4.29%)	20 (7.14%)	10 (3.57%)
High risk	230 (82.1%)	5 (1.79%)	51 (18.2%)	174 (62.1%)
Patients with HF-related events
Low risk	27 (32.9%)	24 (29.3%)	3 (3.66%)	0 (0.00%)
Intermediate risk	18 (22.0%)	2 (2.44%)	10 (12.2%)	6 (7.32%)
High risk	37 (45.1%)	0 (0.00%)	3 (3.66%)	34 (41.5%)

(A) MAGGICs was calculated for HF-related events over 3 years. Low risk was <17.5%, intermediate risk was 17.5–29.2%, and high risk was ≥29.2%. The overall NRI was 20.11%, P=0.02 after adding Ln-RHI to MAGGICs. (B) SHFM was calculated for HF-related events over 3 years. Low risk was >92.4%, intermediate risk was 85.9–92.4%, and high risk was ≤85.9%. The overall NRI was 24.88%, P<0.001 after adding Ln-RHI to SHFM. NRI, net reclassification index. Other abbreviations as in Table 1.

Discussion

This is the first study to demonstrate the incremental prognostic significance of peripheral microvascular endothelial dysfunction volumetrically assessed by RH-PAT in patients with HFREF. Peripheral microvascular endothelial dysfunction was independently associated with HF-related events in patients with HFREF, even after adjustment for propensity score with calculation of multiple risk factors that has been reported previously.^3–8,25 We also demonstrated that MAGGICs and SHFM were powerful novel prognostic predictors in HF patients, with Ln-RHI incrementally predicting near-future HF-related events using the Cox model C-statistic. We also demonstrated that Ln-RHI combined with MAGGICs or SHFM could significantly improve risk determination as shown by the NRI. We clearly demonstrated that peripheral microvascular endothelial dysfunction is an important pathophysiological factor in the prognosis of HFREF.

Endothelial dysfunction is associated with the pathogenesis and progression of HF.⁹ Treasure et al reported that endothelium-dependent vasodilator function was impaired in the coronary microvasculature of patients with dilated cardiomyopathy.²⁶ Kubo et al demonstrated that endothelium-dependent vasodilation was attenuated in HF patients, using invasive measurement of forearm blood flow responses to intra-arterial administration of methacholine, which stimulates endothelium-derived relaxing factor.²⁷ Those studies demonstrated the presence of endothelial dysfunction in patients with HF. However, the tests used in those studies are difficult to perform as routine practical examinations in patients with HF because of their invasiveness. Several studies have reported that endothelial function noninvasively assessed by FMD predicts the prognosis of HF. Katz et al reported that endothelial function assessed by FMD was associated with prognosis in ischemic and nonischemic chronic HF patients (n=259),¹⁵ however, clinical factors in their study were limited and they did not include well-established factors (BNP, renal function, anemia, etc.) for prognosis of HF in their analysis. Meyer et al¹⁴ also indicated that endothelial function assessed by FMD predicted cardiovascular death and United Network for Organ Sharing status 1. Although BNP was included in their study, their study population was small (n=75) and limited to those with LVEF <30%. We investigated the relationship between peripheral microvascular endothelial function assessed by RH-PAT and prognosis of HF in a relatively large population of HFREF patients treated with optimal therapies and with LVEF <50%, which is a generally acceptable cut-off value for HFREF. In the Framingham Heart Study, FMD was associated with traditional risk factors, whereas RH-PAT was more sensitive to metabolic risk factors,^19,28 and it was reported that RH-PAT was not statistically associated with FMD. Several meta-analyses have indicated that the test for microvascular or conduit arterial endothelial function has independent prognostic value for cardiovascular events;^29,30 however, 1 report has recently demonstrated that vasodilation in the microvasculature, but not the conduit artery, was significantly associated with cardiovascular outcomes.³¹ The target vasculature for FMD and RH-PAT is essentially different, and each test might reflect different pathophysiologic conditions in cardiovascular disease. We demonstrated that Ln-RHI was significantly associated with HF-related events independent of BNP, MAGGICs, and SHFM. RH-PAT is the Food and Drug Administration-approved device for assessing peripheral microvascular endothelial function in real-world clinical practice. RH-PAT could comprehensively reflect various clinical conditions of HFREF patients. It has been reported that RH-PAT can detect endothelial dysfunction caused by depression and sleep apnea syndrome, which are poor prognostic factors of HF, in addition to coronary risk factors.^32,33 Several studies have already confirmed that RH-PAT examination is a noninvasive, less operator-dependent, and digitally reproducible automated test for peripheral microvascular endothelial function.^17,18 And we have confirmed good day-to-day, intra-, and interobserver reproducibility (see Results in Supplementary File 1, Figures S1, S2).

We have previously reported impaired peripheral microvascular endothelial function in HF patients with preserved LVEF,¹⁸ and an increase in endothelium-derived microparticles, which are a quantitative plasma marker for endothelial dysfunction, in peripheral blood samples from HF patients.³⁴ Other studies have also reported an attenuated response of forearm blood flow and reactive hyperemia to vasodilator drugs or exercise in HF patients.³⁵ An insufficient increase in blood flow in skeletal muscle during exercise, caused by endothelial dysfunction, leads to exercise intolerance.^35,36 Our study demonstrated that impaired peripheral microvascular endothelial function was significantly associated with the prognosis of HFREF patients, which is considered to be related to decreased production of endothelium-derived NO, elevation of oxidative stress, and an activated renin-angiotensin-aldosterone system. In fact, dietary supplementation with L-arginine and physical exercise has been shown to improve endothelium-dependent vasodilation in HFREF patients.^37,38 Another study demonstrated that angiotensin-converting enzyme inhibitors prevented HF-induced endothelial dysfunction and induced vascular remodeling in a rat model of HF.³⁹ We previously reported that cardiac resynchronization therapy improved Ln-RHI in patients with HF.⁴⁰ All these reports support the concept that peripheral endothelial function could be an important therapeutic target in HFREF. Although we need further investigation of the mechanism, endothelial dysfunction is associated with prognosis in both HFREF and HF patients with preserved LVEF. We should recognize that the vascular component is important in HF. RH-PAT could be a useful tool for risk stratification and assessment of disease status in HFREF patients, besides conventional clinical risk factors. After identification of high-risk HFREF patients, future studies are needed to investigate effective therapeutic approaches for these HF patients with poor cardiovascular prognosis.

Study Limitations

First, the study did not have a large sample size, and comprised only 2 centers in Japan. More large, multicenter studies with various ethnic groups are required to confirm our result. Second, endothelial function is influenced by various factors such as age, sex, atherosclerotic risk factors, and drugs. We tried to exclude bias from various factors that may affect peripheral microvascular endothelial function and prognosis of HF using a multivariable model and propensity score adjustment, but this may have been insufficient. Third, our study excluded patients with HFREF who were in an unstable condition despite maximal therapy (NYHA class IV). Peripheral microvascular endothelial function testing with RH-PAT is difficult to perform in unstable conditions and the results may be unreliable. Fourth, the predictive role of a change in RHI over time still remains unknown. Studies are required to establish whether improvement in RHI after therapeutic investigation might be used as a surrogate endpoint for HF-related events. Further prospective clinical trials are needed. An observational study cannot determine causality.

Conclusions

This study established that peripheral microvascular endothelial dysfunction volumetrically assessed with RHI is a pathophysiologically important component in the cardiovascular prognosis of HFREF, and predicted near-future HF-related cardiovascular events in HFREF patients.

Acknowledgments

Funding: This study was supported in part by a Grant-in-Aid for Scientific Research (No. C25461086 for S. Sugiyama) from the Ministry of Education, Culture, Sports, Science, and Technology in Japan.

Disclosures

All authors declare that they have no conflict of interest.

Supplementary Files

Supplementary File 1

Methods

Results

Figure S1. Bland-Altman plot for intraobserver repeatability of reactive hyperemia peripheral arterial tonometry (RH-PAT).

Figure S2. Bland-Altman plot for intrer-observer repeatability of reactive hyperemia peripheral arterial tonometry (RH-PAT).

Figure S3. Risk of HF-related events according to RHI with or without nitrate or nitroglycerin use.

Table S1. Baseline characteristics between the lower-RHI and higher-RHI groups of HFREF patients

Table S2. Cardiovascular events in patients with HFREF in the lower- and higher-RHI groups

Table S3. Cox proportional hazard analysis for HF-related events in patients with HFREF according to echocardiographic parameters and RHI

Please find supplementary file(s);

http://dx.doi.org/10.1253/circj.CJ-15-0671

References

• 1.    Sakata  Y,  Shimokawa  H. Epidemiology of heart failure in Asia. Circ J 2013; 77: 2209–2217.
• 2.    McMurray  JJ,  Petrie  MC,  Murdoch  DR,  Davie  AP. Clinical epidemiology of heart failure: Public and private health burden. Eur Heart J 1998; 19(Suppl P): P9–P16.
• 3.    Braunwald  E. Biomarkers in heart failure. N Engl J Med 2008; 358: 2148–2159.
• 4.    Tsutamoto  T,  Wada  A,  Maeda  K,  Hisanaga  T,  Maeda  Y,  Fukai  D, et al. Attenuation of compensation of endogenous cardiac natriuretic peptide system in chronic heart failure: Prognostic role of plasma brain natriuretic peptide concentration in patients with chronic symptomatic left ventricular dysfunction. Circulation 1997; 96: 509–516.
• 5.    Pocock  SJ,  Ariti  CA,  McMurray  JJ,  Maggioni  A,  Kober  L,  Squire  IB, et al. Predicting survival in heart failure: A risk score based on 39 372 patients from 30 studies. Eur Heart J 2013; 34: 1404–1413.
• 6.    Groenveld  HF,  Januzzi  JL,  Damman  K,  van Wijngaarden  J,  Hillege  HL,  van Veldhuisen  DJ, et al. Anemia and mortality in heart failure patients a systematic review and meta-analysis. J Am Coll Cardiol 2008; 52: 818–827.
• 7.    From  AM,  Leibson  CL,  Bursi  F,  Redfield  MM,  Weston  SA,  Jacobsen  SJ, et al. Diabetes in heart failure: Prevalence and impact on outcome in the population. Am J Med 2006; 119: 591–599.
• 8.    Scrutinio  D,  Mastropasqua  F,  Guida  P,  Ammirati  E,  Ricci  V,  Raimondo  R, et al. Renal dysfunction and accuracy of N-terminal pro-B-type natriuretic peptide in predicting mortality for hospitalized patients with heart failure. Circ J 2014; 78: 2439–2446.
• 9.    Marti  CN,  Gheorghiade  M,  Kalogeropoulos  AP,  Georgiopoulou  VV,  Quyyumi  AA,  Butler  J. Endothelial dysfunction, arterial stiffness, and heart failure. J Am Coll Cardiol 2012; 60: 1455–1469.
• 10.    Deanfield  JE,  Halcox  JP,  Rabelink  TJ. Endothelial function and dysfunction: Testing and clinical relevance. Circulation 2007; 115: 1285–1295.
• 11.    Flammer  AJ,  Anderson  T,  Celermajer  DS,  Creager  MA,  Deanfield  J,  Ganz  P, et al. The assessment of endothelial function: From research into clinical practice. Circulation 2012; 126: 753–767.
• 12.    Matsuzawa  Y,  Guddeti  RR,  Kwon  TG,  Lerman  LO,  Lerman  A. Secondary prevention strategy of cardiovascular disease using endothelial function testing. Circ J 2015; 79: 685–694.
• 13.    Fischer  D,  Rossa  S,  Landmesser  U,  Spiekermann  S,  Engberding  N,  Hornig  B, et al. Endothelial dysfunction in patients with chronic heart failure is independently associated with increased incidence of hospitalization, cardiac transplantation, or death. Eur Heart J 2005; 26: 65–69.
• 14.    Meyer  B,  Mortl  D,  Strecker  K,  Hulsmann  M,  Kulemann  V,  Neunteufl  T, et al. Flow-mediated vasodilation predicts outcome in patients with chronic heart failure: Comparison with B-type natriuretic peptide. J Am Coll Cardiol 2005; 46: 1011–1018.
• 15.    Katz  SD,  Hryniewicz  K,  Hriljac  I,  Balidemaj  K,  Dimayuga  C,  Hudaihed  A, et al. Vascular endothelial dysfunction and mortality risk in patients with chronic heart failure. Circulation 2005; 111: 310–314.
• 16.    Kuvin  JT,  Karas  RH. Clinical utility of endothelial function testing: Ready for prime time? Circulation 2003; 107: 3243–3247.
• 17.    Bonetti  PO,  Barsness  GW,  Keelan  PC,  Schnell  TI,  Pumper  GM,  Kuvin  JT, et al. Enhanced external counterpulsation improves endothelial function in patients with symptomatic coronary artery disease. J Am Coll Cardiol 2003; 41: 1761–1768.
• 18.    Akiyama  E,  Sugiyama  S,  Matsuzawa  Y,  Konishi  M,  Suzuki  H,  Nozaki  T, et al. Incremental prognostic significance of peripheral endothelial dysfunction in patients with heart failure with normal left ventricular ejection fraction. J Am Coll Cardiol 2012; 60: 1778–1786.
• 19.    Hamburg  NM,  Palmisano  J,  Larson  MG,  Sullivan  LM,  Lehman  BT,  Vasan  RS, et al. Relation of brachial and digital measures of vascular function in the community: The Framingham heart study. Hypertension 2011; 57: 390–396.
• 20.    McKee  PA,  Castelli  WP,  McNamara  PM,  Kannel  WB. The natural history of congestive heart failure: The Framingham study. N Engl J Med 1971; 285: 1441–1446.
• 21.    Levy  WC,  Mozaffarian  D,  Linker  DT,  Sutradhar  SC,  Anker  SD,  Cropp  AB, et al. The Seattle Heart Failure Model: Prediction of survival in heart failure. Circulation 2006; 113: 1424–1433.
• 22.    D’Agostino  RB Jr. Propensity scores in cardiovascular research. Circulation 2007; 115: 2340–2343.
• 23.    Vansteelandt  S,  Daniel  RM. On regression adjustment for the propensity score. Stat Med 2014; 33: 4053–4072.
• 24.    Miller  TD,  Askew  JW. Net reclassification improvement and integrated discrimination improvement: New standards for evaluating the incremental value of stress imaging for risk assessment. Circ Cardiovasc Imaging 2013; 6: 496–498.
• 25.    Hillege  HL,  Girbes  AR,  de Kam  PJ,  Boomsma  F,  de Zeeuw  D,  Charlesworth  A, et al. Renal function, neurohormonal activation, and survival in patients with chronic heart failure. Circulation 2000; 102: 203–210.
• 26.    Treasure  CB,  Vita  JA,  Cox  DA,  Fish  RD,  Gordon  JB,  Mudge  GH, et al. Endothelium-dependent dilation of the coronary microvasculature is impaired in dilated cardiomyopathy. Circulation 1990; 81: 772–779.
• 27.    Kubo  SH,  Rector  TS,  Bank  AJ,  Williams  RE,  Heifetz  SM. Endothelium-dependent vasodilation is attenuated in patients with heart failure. Circulation 1991; 84: 1589–1596.
• 28.    Hamburg  NM,  Keyes  MJ,  Larson  MG,  Vasan  RS,  Schnabel  R,  Pryde  MM, et al. Cross-sectional relations of digital vascular function to cardiovascular risk factors in the Framingham Heart Study. Circulation 2008; 117: 2467–2474.
• 29.    Ras  RT,  Streppel  MT,  Draijer  R,  Zock  PL. Flow-mediated dilation and cardiovascular risk prediction: A systematic review with meta-analysis. Int J Cardiol 2013; 168: 344–351.
• 30.    Xu  Y,  Arora  RC,  Hiebert  BM,  Lerner  B,  Szwajcer  A,  McDonald  K, et al. Non-invasive endothelial function testing and the risk of adverse outcomes: A systematic review and meta-analysis. Eur Heart J Cardiovasc Imaging 2014; 15: 736–746.
• 31.    Lind  L,  Berglund  L,  Larsson  A,  Sundstrom  J. Endothelial function in resistance and conduit arteries and 5-year risk of cardiovascular disease. Circulation 2011; 123: 1545–1551.
• 32.    Borel  JC,  Roux-Lombard  P,  Tamisier  R,  Arnaud  C,  Monneret  D,  Arnol  N, et al. Endothelial dysfunction and specific inflammation in obesity hypoventilation syndrome. PLoS One 2009; 4: e6733, doi:10.1371/journal.pone.0006733.
• 33.    Osika  W,  Montgomery  SM,  Dangardt  F,  Wahrborg  P,  Gan  LM,  Tideman  E, et al. Anger, depression and anxiety associated with endothelial function in childhood and adolescence. Arch Dis Child 2011; 96: 38–43.
• 34.    Nozaki  T,  Sugiyama  S,  Sugamura  K,  Ohba  K,  Matsuzawa  Y,  Konishi  M, et al. Prognostic value of endothelial microparticles in patients with heart failure. Eur J Heart Fail 2010; 12: 1223–1228.
• 35.    Zelis  R,  Sinoway  LI,  Musch  TI,  Davis  D,  Just  H. Regional blood flow in congestive heart failure: Concept of compensatory mechanisms with short and long time constants. Am J Cardiol 1988; 62: 2E–8E.
• 36.    Borlaug  BA. Mechanisms of exercise intolerance in heart failure with preserved ejection fraction. Circ J 2014; 78: 20–32.
• 37.    Hambrecht  R,  Hilbrich  L,  Erbs  S,  Gielen  S,  Fiehn  E,  Schoene  N, et al. Correction of endothelial dysfunction in chronic heart failure: Additional effects of exercise training and oral L-arginine supplementation. J Am Coll Cardiol 2000; 35: 706–713.
• 38.    Hambrecht  R,  Fiehn  E,  Weigl  C,  Gielen  S,  Hamann  C,  Kaiser  R, et al. Regular physical exercise corrects endothelial dysfunction and improves exercise capacity in patients with chronic heart failure. Circulation 1998; 98: 2709–2715.
• 39.    Mulder  P,  Elfertak  L,  Richard  V,  Compagnon  P,  Devaux  B,  Henry  JP, et al. Peripheral artery structure and endothelial function in heart failure: Effect of ACE inhibition. Am J Physiol 1996; 271: H469–H477.
• 40.    Enomoto  K,  Yamabe  H,  Toyama  K,  Matsuzawa  Y,  Yamamuro  M,  Uemura  T, et al. Improvement effect on endothelial function in patients with congestive heart failure treated with cardiac resynchronization therapy. J Cardiol 2011; 58: 69–73.