Journal of Echocardiography

, Volume 10, Issue 2, pp 56–64

A novel global strain diastolic index correlates with plasma NT-proBNP levels in asymptomatic hypertensive patients with preserved left ventricular ejection fraction

  • Shuo-Ju Chiang
  • Masao Daimon
  • Katsuhisa Ishii
  • Sakiko Miyazaki
  • Yoko Koiso
  • Hiromasa Suzuki
  • Katsumi Miyauchi
  • Bei Yang
  • Mei-Hsiu Yeh
  • Betau Hwang
  • Hiroyuki Daida
Original Investigation

DOI: 10.1007/s12574-012-0122-4

Cite this article as:
Chiang, SJ., Daimon, M., Ishii, K. et al. J Echocardiogr (2012) 10: 56. doi:10.1007/s12574-012-0122-4
  • 99 Downloads

Abstract

Background

The strain imaging diastolic index (SI-DI) was reported to be a sensitive marker of regional left ventricular (LV) delayed relaxation induced by ischemia. However, the clinical usefulness of the global SI-DI has not been evaluated. N-terminal pro-brain natriuretic peptide (NT-proBNP) is a sensitive biomarker for the detection of asymptomatic diastolic LV dysfunction. This study investigated the ability of a novel parameter, the global SI-DI, obtained using 2D speckle tracking imaging (2DSI) to correlate with the plasma NT-proBNP levels in asymptomatic hypertensive patients with preserved LV ejection fraction.

Methods

We performed 2D echocardiography and 2DSI in 83 asymptomatic hypertensive patients with preserved ejection fraction (>45 %) and in 37 control subjects. In 2DSI, the LV longitudinal peak strain and the SI-DI of 18 LV segments were measured. NT-proBNP was measured in all subjects. The data were compared between hypertensive patients and normal controls.

Results

The average peak strain and global SI-DI of 18 LV segments were significantly reduced in hypertensive patients compared with control subjects (P < 0.05); however, only the global SI-DI was significantly correlated with log10 NT-proBNP (r = −0.469, P = 0.001). In Pearson’s correlation analyses, log10 NT-proBNP was significantly correlated with E/e′, E/A ratio, early diastolic mitral annular velocity (e′), global peak strain, deceleration time of the E-wave, and LV ejection fraction. In the multiple stepwise regression analysis, the global SI-DI was the strongest independent determinant of log10 NT-proBNP (β = −0.386, P = 0.008).

Conclusions

The global SI-DI derived from 2DSI correlates well with plasma NT-proBNP levels and may have prognostic value in asymptomatic hypertensive patients with preserved ejection fraction.

Keywords

Hypertension Echocardiography Brain natriuretic peptides Ultrasonic diagnosis 

Introduction

Hypertension is the most common risk factor associated with heart failure (HF) in the general population [1]. In the process of left ventricular (LV) remodeling, changes in LV diastolic characteristics occur that precede changes in LV systolic function. In fact, more than half of the patients with HF present with normal ejection fraction and diastolic dysfunction [2]. Thus, it is important to identify asymptomatic hypertensive patients who are at risk of developing diastolic HF among those with normal LV systolic function. Strain imaging derived from 2D speckle tracking echocardiography enables the quantification of regional wall motion without tethering effect and Doppler angle dependency [3, 4, 5]. A number of previous investigations have emphasized the clinical value of systolic strain imaging in assessing myocardial systolic dysfunction in patients with HF [6, 7, 8, 9, 10, 11]. Although some investigations have assessed diastolic function with strain imaging, the clinical value of diastolic strain has not yet been firmly established [12, 13, 14].

Recently, the strain imaging diastolic index (SI-DI) was reported to be a sensitive marker to detect regional LV delayed relaxation during early diastole induced by myocardial ischemia [15, 16]. However, no previous studies have examined the clinical usefulness of the global SI-DI. We hypothesized that a novel parameter, the global SI-DI, may be useful in identifying asymptomatic hypertensive patients with a preserved ejection fraction who are at high risk of developing diastolic HF. Therefore, we examined the value of the global SI-DI for assessing latent LV diastolic dysfunction in asymptomatic hypertensive patients with preserved ejection fraction. Since plasma N-terminal pro-brain natriuretic peptide (NT-proBNP) is a powerful neurohumoral prognostic factor in HF with preserved ejection fraction [17, 18, 19], we also determined the ability of the global SI-DI to correlate with plasma NT-proBNP levels.

Materials and methods

Patient selection

The study included 83 consecutive asymptomatic patients with essential hypertension from an outpatient clinic population. All patients fulfilled the following criteria: (1) a past history of systolic blood pressure (BP) ≥140 mmHg and/or diastolic BP ≥90 mmHg in two or more hospital visits at 2-week intervals without antihypertensive medication; (2) continued hypertensive treatment for ≤12 months; (3) no clinical, laboratory, or echocardiographic evidence of congestive HF, regional LV wall motion abnormalities, significant valvular heart disease, coronary artery disease, secondary causes of hypertension, diabetes mellitus, chronic renal failure, or other important concomitant disease; (4) LV ejection fraction ≥45 %; and (5) good-quality echocardiographic images. Any subjects who had atrial or ventricular arrhythmias, pacemaker implantation, bundle branch block, significant valvular heart disease, or a past history of coronary artery disease or congestive HF were excluded from the study. We also excluded subjects with poor and inadequate images for strain analysis. A total of 37 age-matched subjects without a history of cardiovascular disease served as the control group. The control subjects also had normal clinical and laboratory findings, and no abnormal findings on conventional echocardiography or exercise stress testing. The study protocol was approved by the Institutional Review Board of Juntendo University Hospital and Taipei City Hospital.

Analysis of the strain diastolic index

Echocardiographic images were obtained using an ultrasound system (IE-33, Philips) with a 2.5-MHz phased array transducer in the apical 2-, 3-, or 4-chamber views and a high frame rate (45 ± 5 frames/s). Five stable and well-defined consecutive cardiac cycles were acquired for each view during a transient breath hold at end-expiration and stored on a disk for offline analysis. Longitudinal strain analysis [15, 16] was performed offline on a PC workstation using commercially available software (QLAB, Advanced Ultrasound Quantification Software, Release 7.0; Philips Ultrasound, Bothell, WA, USA) [20, 21, 22, 23]. End diastole was defined as the time of the R-wave peak in the electrocardiogram, and end systole was defined as the time of aortic valve closure on 2D echocardiography. Cardiac cycles with atrial or ventricular premature beats were excluded. From each apical view, the septal and lateral points on the mitral annulus and a point on the apical endocardium were manually initialized in the end-diastolic frame. Following initialization, the computer automatically formed a region-of-interest (ROI) including the entire transmural LV wall in all six segments in each apical view, and the software selected markers moving with the tissue. Manual adjustment of the ROI was performed when necessary. Subsequently, automatic frame-by-frame tracking of these markers using a block matching algorithm in the whole layer of six ROIs during the cardiac cycle yielded a measure of strain in each of the six segments of the myocardium in the 2-, 3-, and 4-chamber views. The strain curve was generated by strain computation throughout the cardiac cycle. Automatic frame-by-frame tracking of these markers during the cardiac cycle yielded a measure of strain in each of the six segments of the myocardium in the 2-, 3-, and 4-chamber views. Each segment was individually analyzed, and the SI-DI was calculated in each segment using a strain curve [15, 16]. The peak strain was defined as the highest strain value that was acquired in the longitudinal direction throughout the cardiac cycle. The end-systolic strain value at the closure of the aortic valve (A) and at a time that represented one-third of the diastolic duration (B) were measured. The SI-DI value was determined as: (A − B)/A × 100 % to assess the regional delayed LV relaxation [15, 16] (Fig. 1). Then, the global SI-DI was calculated as the average of the SI-DI values that were obtained from the three apical views. The peak strain was also measured in each segment in the three apical views (Fig. 1), and the average value was calculated in order to obtain the global peak strain. All strain measurements were acquired in three cardiac cycles and the averaged data were calculated. When we obtained the strain curve, we confirmed whether the three strain curves during three consecutive cardiac cycles were uniform and if these curves corresponded with the transverse motion of the myocardium visually. We excluded segments from analysis in which three uniform strain curves could not be obtained during three consecutive cardiac cycles or in which the strain curves were considered to be inappropriate for analysis. All analyses of strain imaging and conventional echocardiography were performed by investigators who were blinded to the results of the plasma NT-proBNP measurements and all clinical data. For the measurement of the SI-DI, the interobserver variability was 5.6 % and the intraobserver variability was 4.8 %.
Fig. 1

2D strain and strain analyses. The 2-dimensional (2D) speckle tracking image in the apical 3-chamber view (top); measurement of the strain imaging diastolic index (SI-DI) (lower left) and the longitudinal peak strain (lower right). The SI-DI value was determined as: (A − B)/A × 100 % in each segment

Conventional 2D and Doppler echocardiography

Conventional echocardiographic measurements were also performed in all subjects. The LV diameter was measured at both end systole and end diastole by M-mode echocardiography. The thickness of the interventricular septum and the LV posterior wall were measured at end systole. These parameters were used in a previously reported equation to calculate the LV mass [24]. The LV end-diastolic volume and end-systolic volume were determined from biplane images using a modification of Simpson’s method. The ejection fraction (%) was calculated by the following equation: 100 × (end-diastolic volume − end-systolic volume)/end-diastolic volume. The left atrial volume was measured using the biplane area–length method. Each parameter obtained from the chamber quantification was indexed for the body surface area (BSA), when appropriate.

For assessing conventional diastolic parameters, mitral inflow and tissue Doppler imaging were also examined [25]. The peak early diastolic velocity (E), the deceleration time from the peak of the early diastolic wave to baseline (E-Dec time), the peak atrial systolic velocity (A), and the E/A ratio were assessed. The mitral annular motion velocity was recorded at the medial mitral annulus site in the apical 4-chamber view by pulsed tissue Doppler echocardiography. The peak early diastolic motion velocity (e′), peak motion velocity during atrial systole (a′), and the ratio of the peak early diastolic transmitral flow velocity E to e′ (E/e′) were calculated [26].

Blood test for NT-proBNP

Blood samples for the measurement of NT-proBNP were drawn in all hypertensive patients and normal controls at rest on the same day as echocardiography. Plasma NT-proBNP levels were determined using a commercially available sandwich immunoassay (Elecsys™) and an automated bench-top analyzer (Elecsys 2010, Roche Diagnostics, Germany). All echocardiographic measurements were obtained without knowledge of the NT-proBNP data.

Statistical analysis

Values are expressed as the mean ± standard deviation. Comparison of variables between the hypertensive and control groups was performed using an unpaired t test or one-way analysis of variance (ANOVA) with Tukey’s post hoc test. When the data were not normally disturbed, a Kruskal–Wallis test was performed. Categorical data were compared between the two groups with a Chi-squared test. Pearson’s linear correlation analysis was used to determine the correlations between log10 NT-proBNP and the echocardiographic variables. Multiple linear regression analysis was also performed to determine the independent determinants of log10 NT-proBNP among the echocardiographic parameters. All statistical analyses were performed with SPSS software version 17.0 (SPSS Inc., Chicago, IL, USA). A P value less than 0.05 was considered to be statistically significant for all analyses.

Results

Patient characteristics and conventional echocardiography

The patient characteristics are summarized in Table 1. There were no significant differences between the two groups in height, BSA, and heart rate, whereas weight, body mass index, and systolic and diastolic BP were greater in hypertensive patients than in normal controls.
Table 1

Characteristics of hypertensive patients and normal control subjects

 

Control (n = 37)

Hypertension (n = 83)

P value

Gender (men/women)

13/24

48/35

NS

Age (years)

52.9 ± 17.8

57.5 ± 11.5

NS

Height (cm)

164.8 ± 7.5

163.5 ± 7.5

NS

Weight (kg)

59.4 ± 9.0

65.1 ± 10.5

<0.05

BMI (kg/m2)

21.8 ± 2.5

24.3 ± 3.0

<0.05

BSA (m2)

1.7 ± 0.2

1.7 ± 0.2

NS

Creatinine

0.8 ± 0.5

0.9 ± 0.7

NS

Heart rate (beats per minute)

72.8 ± 6.7

73.8 ± 9.8

NS

Systolic blood pressure (mmHg)

118.9 ± 10.3

133.5 ± 7.6

<0.05

Diastolic blood pressure (mmHg)

67.2 ± 7.1

77.8 ± 7.9

<0.05

Smoking

7

33

<0.001

Hyperlipidemia

1

37

<0.001

Stroke

0

1

NS

Medications

 Beta-blocker

0

31

 

 CCB

0

57

 

 ARB

0

38

 

 ACE-I

0

2

 

 Diuretic

0

3

 

 Alpha-blocker

0

2

 

 ARB + diuretic

0

7

 

Data are shown as the mean ± standard deviation (SD) or number of patients (%)

ACE-I angiotensin-converting enzyme inhibitor, ARB angiotensin receptor blocker, BMI body mass index, BSA body surface area, CCB calcium channel blocker, NS not significant

The results of the conventional echocardiography are summarized in Table 2. Both interventricular septal wall thickness and LV posterior wall thickness were larger in the hypertensive group than in the control group, whereas the LV end-diastolic and end-systolic diameters were similar between the two groups. The left atrial volume index and the LV mass index in the hypertensive group were greater than in the control group. The hypertensive group had a smaller E/A ratio and e′, but a greater E-Dec time and E/e′ compared with the control group. These results indicated decreased diastolic function in hypertensive patients. The LV ejection fraction was similar in the two groups.
Table 2

Results of the conventional echocardiography

 

Control (n = 37)

Hypertension (n = 83)

P value

IVS thickness (cm)

0.9 ± 0.1

1.1 ± 0.1

<0.001

PW thickness (cm)

0.9 ± 0.1

1.1 ± 0.1

<0.001

LVEDD (cm)

4.4 ± 0.5

4.4 ± 0.5

NS

LVESD (cm)

2.9 ± 0.4

2.7 ± 0.5

NS

LAV index (ml/m2)

17.4 ± 3.4

19.5 ± 4.7

<0.05

LV mass index (g/m2)

81.4 ± 11.0

90.6 ± 24.1

<0.05

Ejection fraction (%)

59.3 ± 8.4

56.5 ± 7.3

NS

E-wave (cm/s)

75.8 ± 17.3

65.9 ± 15.4

<0.01

A-wave (cm/s)

61.2 ± 18.1

75.2 ± 17.7

<0.001

E/A

1.4 ± 0.6

0.9 ± 0.3

<0.001

E-Dec time (ms)

187.2 ± 35.6

224.2 ± 44.1

<0.001

E/e

7.9 ± 1.9

9.3 ± 2.5

<0.01

e

10.0 ± 2.5

7.5 ± 1.9

<0.001

IVS interventricular septum; LAV left atrial volume, LVEDD left ventricular end-diastolic diameter, LVESD left ventricular end-systolic diameter, NS not significant, PW posterior wall

SI-DI and peak strain

In patients with hypertension, strain images derived from 1,469 of all 1,494 segments (98.3 %) were suitable for analysis. In control subjects, strain images derived from 659 of all 666 segments (98.9 %) were suitable for analysis. The SI-DI and peak strain were measured in each segment of the apical 4-, 3-, and 2-chamber views (Fig. 2). The hypertensive patients had significantly reduced SI-DI in every segment in the 4-, 3-, and 2-chamber views compared with the controls (Fig. 3). As a result, the calculated global SI-DI was significantly reduced in hypertensive patients compared with the controls (0.4 ± 0.1 vs. 0.7 ± 0.1, P < 0.001), suggesting impaired LV relaxation in hypertensive patients.
Fig. 2

Strain in the apical 4-, 3-, and 2-chamber views in a hypertensive patient and a control subject. a Apical 4-chamber view, b apical 3-chamber view, c apical 2-chamber view, left column normal strain, right column hypertension (HTN) strain, AVC aortic valve closure, AVO aortic valve opening

Fig. 3

The SI-DI in the apical 4- (left), 3- (middle), and 2- (right) chamber views. In these views, the SI-DI of every segment is lower in hypertensive patients than in normal controls

The peak strain was also reduced in every segment in the 4-, 3-, and 2-chamber views in hypertensive patients compared with the controls. Thus, the global peak strain was significantly lower in the hypertensive group than in the control group (−17.9 ± 3.3 vs. −20.8 ± 4.5, respectively, P < 0.001).

Global SI-DI, echocardiographic parameters, and log10 NT-proBNP

The hypertensive patients had higher plasma NT-proBNP levels than the control subjects (39.5 ± 60.8 vs. 9.9 ± 11.0 pg/ml, respectively, P < 0.05). The global SI-DI was significantly correlated with log10 NT-proBNP (P = 0.001) (Fig. 4), whereas the peak strain was not (P = 0.401). Among the echocardiographic parameters, E/e′, E/A ratio, e′, global peak strain, deceleration time of the E-wave, and LV ejection fraction were also significantly correlated with log10 NT-proBNP, as well as the global SI-DI (Table 3). Furthermore, multivariate stepwise linear regression analysis showed that the global SI-DI was the strongest independent determinant of log10 NT-proBNP (β = −0.386, P = 0.008) (Table 3).
Fig. 4

The linear regression of the global 18-segment SI-DI and log10 N-terminal pro-brain natriuretic peptide (NT-proBNP). Y = 2.702 − 2.849X, r = −0.469, P = 0.001

Table 3

The effect of echocardiographic parameters on log 10 N-terminal pro-brain natriuretic peptide (NT-proBNP)

 

Pearson’s r

P value

Multivariate β

P value

LV mass index (g/m2)

0.266

0.05

 

NS

LAV index (ml/m2)

0.129

NS

  

E/e

0.492

0.001

0.267

0.04

E/A ratio

−0.272

0.04

 

NS

e′ (cm/s)

−0.405

0.002

−0.281

0.03

Ejection fraction (%)

−0.382

0.04

 

NS

Global peak strain

−0.302

0.02

 

NS

Deceleration time of the E-wave

0.263

0.03

 

NS

Global SI-DI

−0.469

0.001

−0.374

0.005

Discussion

In the current study, we demonstrated the usefulness of a novel parameter, the global SI-DI, to find the correlation with the plasma NT-proBNP levels in asymptomatic hypertensive patients with preserved ejection fraction. Presumably, these asymptomatic patients with elevated NT-proBNP are at increased risk of developing diastolic HF in the future. Furthermore, the global SI-DI was a more sensitive marker for predicting elevated NT-proBNP than conventional diastolic parameters obtained from mitral inflow and tissue Doppler imaging, which have been recognized as standard prognostic indicators in diastolic HF. Our results indicate that the global SI-DI might serve as a novel echocardiographic parameter for assessing diastolic function.

In patients with hypertension, the deterioration of diastolic function precedes the impairment of systolic function and plays a major role in the pathophysiology of HF. Thus, it is clinically important to detect the deterioration of diastolic function accurately at an early stage of HF. The SI-DI has been proposed as a sensitive marker to detect regional LV delayed relaxation induced by myocardial ischemia [15, 16]. Furthermore, LV delayed relaxation is a characteristic of early-stage diastolic dysfunction that might lead to restrictive physiology [25]. In addition, recent studies showed that the reproducibility of global strain parameters was better than that of regional strain parameters [27]. Thus, we presume that the global SI-DI might be useful in assessing diastolic function at an early stage in asymptomatic patients with latent LV dysfunction.

Although several studies have established the clinical value of systolic strain, there is much less information on diastolic strain. In the current study, we examined the clinical value of diastolic strain in asymptomatic patients who had not yet developed HF, whereas previous studies evaluated diastolic strain in patients with overt HF [13, 14]. NT-proBNP was significantly elevated even in our asymptomatic hypertensive patients compared with normal controls. NT-proBNP is a powerful prognostic factor [17, 18, 19] that is strongly correlated with the LV filling pressure, as determined by invasive methods [15]. Thus, we assume that the global SI-DI might predict the risk of the development of HF with elevated filling pressure in asymptomatic hypertensive patients.

Interestingly, both the global peak strain and the global SI-DI were reduced in hypertensive patients compared with normal control subjects. However, the presence of hypertension did not appear to alter the LV ejection fraction. Some investigations [7, 11] reported decreased systolic function assessed by 2D speckle tracking in patients with preserved ejection fraction. These results indicate that LV contraction assessed by 2D speckle tracking may be superior to the conventional LV ejection fraction for evaluating LV systolic dysfunction. Despite the reduction in global strain in hypertensive patients with preserved ejection fraction in the present study, only the SI-DI was a major independent determinant of NT-proBNP. This finding emphasizes the importance of assessing diastolic function in patients with normal ejection fraction, regardless of the detection of systolic dysfunction by 2D speckle tracking.

In the present study, we showed that the global SI-DI was the strongest independent determinant of log10 NT-proBNP compared to conventional echocardiographic parameters. We examined patients with preserved ejection fraction in this study and diastolic dysfunction was considered to be a major factor contributing to elevated NT-proBNP. Conventional echocardiographic parameters using tissue Doppler imaging carry inherent limitations of tissue Doppler imaging. Tissue Doppler imaging is angle-dependent and is affected by tethering and myocardial translation. Other conventional echocardiographic parameters, such as the E/A ratio and deceleration time of the E-wave, have a limitation of pseudo-normalization, and are known to be inferior to those parameters using tissue Doppler imaging [26]. In contrast, 2D speckle tracking echocardiography used for determining the SI-DI offers advantages over the Doppler-derived parameters given their angle independence. In addition, the global SI-DI, using the average of strain measurements of the whole left ventricle, allows the assessment of the global LV diastolic function. These factors were considered to contribute to the advantage of the global SI-DI over conventional echocardiographic parameters.

In the current study, we used longitudinal strain analyses, although transverse strain was used in previous studies [15, 16]. This is because the reproducibility of longitudinal strain measurements was reported to be superior to those of other directional strain measurements [28]. In addition, it was also reported that longitudinal strain was impaired earlier than other directional strain in hypertensive patients [29]. The accuracy of transverse, longitudinal, or other directional SI-DI for assessing diastolic function needs further evaluation.

Limitations

An important limitation of our study is that we did not measure tau or LV diastolic filling pressure by invasive methods. This was because cardiac catheterization was not indicated in our study population of asymptomatic patients from outpatient clinics. However, NT-proBNP has been reported to be strongly correlated with the filling pressure measured invasively [18]. Thus, we believe that the global SI-DI might represent a good index of diastolic function in our population. Additional studies are needed that compare the global SI-DI with invasive measures of diastolic function.

In this study, strain analysis was performed using QLAB software. The ability of 2D speckle tracking using QLAB was validated in comparison with tagged cardiac magnetic resonance [20], and many investigations using 2D speckle tracking echocardiography with QLAB have been reported [21, 22, 23]. However, strain measurements using this software have not been well validated as far as we know. There is a need for further investigation in order to validate this method.

Although we excluded patients with regional LV wall motion abnormalities, we could not exclude patients with latent coronary artery disease. Thus, an effect of latent coronary artery disease on strain measurements cannot be ruled out.

The plasma NT-proBNP level could be significantly influenced by the estimated glomerular filtration rate (GFR) [30]. Although we excluded patients with chronic renal failure, we regrettably did not examine the data on the estimated GFR in this study. However, previous investigations, conducted among similar patients to our study, showed the NT-proBNP level to be correlated with the pulmonary capillary wedge pressure measured by catheterization [17] and predicted cardiovascular outcome [19], making us confident of our results in this study.

Conclusions

The global strain imaging diastolic index (SI-DI) was an independent predictor of elevated N-terminal pro-brain natriuretic peptide (NT-proBNP) in hypertensive patients with preserved ejection fraction. The global SI-DI may serve as a novel and sensitive indicator in asymptomatic hypertensive patients who are at risk of developing diastolic heart failure (HF).

Acknowledgments

This work was partially supported by a Grant-in-Aid for Scientific Research (20500427) from the Japan Society for the Promotion of Science (Masao Daimon), partially supported by a Grant-in-Aid for Scientific Research (20231501) from the Ministry of Health, Labour and Welfare (Masao Daimon), and was partially supported by a Grant for Scientific Research (097XDAA00141) from the Department of Health, Taipei City Government (Shuo-Ju Chiang and Betau Hwang). We are grateful to Ms. Mei-Hsiu Yeh and Ms. Maiko Hirano for their research assistance.

Conflict of interest

There is no relationship with any industry.

Copyright information

© Japanese Society of Echocardiography 2012

Authors and Affiliations

  • Shuo-Ju Chiang
    • 1
    • 2
  • Masao Daimon
    • 2
  • Katsuhisa Ishii
    • 3
  • Sakiko Miyazaki
    • 2
  • Yoko Koiso
    • 2
  • Hiromasa Suzuki
    • 2
  • Katsumi Miyauchi
    • 2
  • Bei Yang
    • 2
  • Mei-Hsiu Yeh
    • 1
  • Betau Hwang
    • 1
  • Hiroyuki Daida
    • 2
  1. 1.Division of Cardiology, Department of Internal Medicine and PediatricsTaipei City Hospital Zhongxiao Branch, National Yang-Ming UniversityTaipeiTaiwan
  2. 2.Department of CardiologyJuntendo University School of MedicineTokyoJapan
  3. 3.Department of CardiologyKansai Electric Power HospitalOsakaJapan

Personalised recommendations