Background
Although previous studies have established the ability of mitral annular velocities and velocity dispersion indices to differentiate between ischemic cardiomyopathy (ICM) and nonischemic cardiomyopathy, prospective data are lacking on both the use of heterogeneity of mitral annular velocities to predict the ischemic etiology in patients with left ventricular dysfunction and further cardiovascular prognosis.
Methods
A total of 232 patients with left ventricular ejection fractions < 40% were admitted between 2008 and 2010. Doppler tissue imaging was performed on six mitral annular sites for three consecutive beats and then averaged for each site. Systolic (Vs′) and early (Ve′) and late (Va′) diastolic mitral annular velocity dispersion indices among the six mitral annular sites were calculated.
Results
Ve′ was a significant predictor ( P < .01) of ICM in multivariate logistic regression models adjusted for clinical variables and conventional echocardiography. The optimal cutoff value for predicting ICM was Ve′ ≥ 16.7 with an area under the receiver operating characteristic curve of 0.92. Its sensitivity and specificity were 87% and 85%, respectively. During follow-up (median, 32 months), 64 participants experienced cardiac mortality. The adjusted hazard ratio in Cox proportional-hazards analysis for death in the third tertile in comparison with the first tertile of Ve′ was 2.92 ( P = .02).
Conclusions
A high degree of heterogeneity of e′, expressed as Ve′, provides incremental value over clinical variables and conventional echocardiography to predict the prevalence of low left ventricular ejection fractions patients with ICM. Furthermore, elevated Ve′ could also identify patients at a high risk for cardiac mortality.
Differentiating between ischemic cardiomyopathy (ICM) and nonischemic cardiomyopathy (NICM) using noninvasive diagnostic modalities is clinically important because of their possibly different therapeutic strategies (e.g., no need for percutaneous coronary intervention for NICM and aggressive antiplatelet therapy for ICM ). Patients with low left ventricular (LV) ejection fractions (LVEFs) are often treated with coronary angiography because of the difficulty in differentiating between ICM and NICM when performing noninvasive examinations, such as clinical history, electrocardiography, or stress 201 Tl myocardial scintigraphy. Positron emission tomography, electron-beam computed tomography, magnetic resonance imaging, and transthoracic high-frequency coronary echocardiography are effective in differentiating ICM from NICM but are more complex and less portable approaches. Therefore, invasive coronary angiography remains the conventional means of diagnosing ICM in patients with LV systolic dysfunction. Nevertheless, this invasive procedure poses the risk of serious complications.
As a noninvasive procedure, Doppler tissue imaging (DTI) can assess LV regional and global systolic and diastolic function. More specifically, DTI allows clinicians to assess regional performance and dynamics in ischemic heart disease. In addition to decreases in mitral annular velocities in patients with impaired LV systolic function, previous studies have observed changes in mitral annular velocities in patients with coronary artery disease (CAD). Duncan et al. and Plewka et al. found that stress-induced change in early diastolic velocity and the regional heterogeneity of velocities among the six walls of the left ventricle could distinguish ICM from NICM, respectively. However, differentiating between ischemic and nonischemic etiologies in patients with advanced LV dysfunction by DTI has still seldom been addressed. Prospective data on the heterogeneity of mitral annular velocities to predict ischemic etiology in patients with LV dysfunction and further cardiovascular prognosis are still lacking.
In this study, we examined the feasibility of predicting ischemic etiology of disease in patients with cardiomyopathy using a high degree of heterogeneity on DTI at different sites around the mitral annulus. Moreover, we hypothesized that cardiac mortality might be related to the level of heterogeneous values of mitral annular velocities.
Methods
Study Populations
We prospectively enrolled adult patients with histories of symptomatic heart failure (New York Heart Association [NYHA] functional class ≥ II) who were scheduled to undergo diagnostic coronary angiography to investigate ischemic etiology of cardiomyopathy between 2008 and 2010. Inclusion criteria were the presence of globally reduced contractility (LVEF < 40%) by echocardiography using the modified Simpson’s method, sinus rhythm, the absence of aortic or mitral stenosis, prosthetic valves, severe mitral annular calcification, or the presence of pathologic Q waves on electrocardiography. Sixty-two patients were excluded because of atrial fibrillation (13 patients), significant valvular diseases (12 patients), and myocardial infarction (37 patients), leaving 232 patients (79%) for analysis. The study protocol was reviewed and received approval from the hospital ethics committee. Informed written consent was obtained from each patient before participating in the study. The study was conducted on the basis of the rules outlined in the Declaration of Helsinki.
Coronary Angiography
CAD was considered present when an obstruction of the vessel lumen exceeded 50%. For the accurate ascertainment of ischemic etiology of cardiomyopathy, ICM was diagnosed for patients with ≥75% luminal diameter stenoses of the left main or proximal left descending anterior coronary artery, as well as two or more major epicardial coronary arteries; otherwise, patients were diagnosed with NICM.
Conventional M-Mode, Two-Dimensional, and Doppler Echocardiography
Each subject underwent echocardiography independently by an experienced physician with transthoracic M-mode, two-dimensional, and Doppler imaging using commercially available echocardiographic units (Vivid 7, GE Healthcare, Milwaukee, WI; or Philips Sonos 7500, Philips Medical Imaging, Best, The Netherlands) <2 days before coronary angiography was performed. LV end-diastolic volume index (LVEDVI), LV end-systolic volume index (ESVI), and LVEF were assessed on apical two-chamber and four-chamber views using the modified Simpson’s rule. LV mass was determined using the area-length method and was also indexed to body surface area.
Transmitral early (E) and late (A) diastolic flow velocities, deceleration time of early diastolic flow velocity, isovolumic relaxation time, and myocardial performance index were measured using conventional Doppler echocardiography. The LV wall motion score index (WMSI) was determined as recommended by the American Society of Echocardiography.
Pulsed-Wave DTI (PWDTI)
In each patient, PWDTI used image-guided spectral Doppler analysis, which was performed immediately after conventional two-dimensional echocardiography with the patient in the left lateral decubitus position and breathing normally. The values of mitral annular velocities were determined at six sites for three consecutive beats using three apical views (two-chamber, four-chamber, and apical long-axis views). The six mitral annular sites were septal, lateral, inferior, anterior, posterior, and anteroseptal area at the central part of the junction of left atrium and the left ventricle on each mitral annular site. Nyquist limits of 15 and 20 cm/sec were adjusted for velocity recording with the optimal gain setting. The sample volume used in this study was either 5.9 mm in length for the GE Vivid 7 system or 5.6 mm in length for the Philips Sonos 7500 system; it was positioned parallel to the transducer without angle correction on each mitral annular site. The velocity acquired by PWDTI during the expiration phase was recorded and stored on videotape, digitized, and transferred to a magneto-optical disk for offline analysis. The two parameters calculated for each patient included the following: (1) average values of s′, e′, and a′ of the six mitral annular sites and (2) the velocity dispersion indices of s′, e′, and a′ (Vs′, Ve′, and Va′), defined as the ratio of the standard deviation to the average value of DTI velocity of the six mitral annular sites.
CAD was assessed using the velocity dispersion indices by dividing patients into four subgroups: group 1 (without ≥50% stenosis of any epicardial vessel), group 2 (with ≥50% stenosis of one epicardial vessel), group 3 (with ≥50% stenosis of two epicardial vessels), and group 4 (with ≥50% stenosis of three epicardial vessels).
End Point and Follow-Up
All patients were followed up unless the end point, defined as cardiac death, included pump failure or sudden cardiac death. Deaths resulting from a deterioration of congestive heart failure, with progression of congestive symptoms, were classified as pump failure. Sudden cardiac death was defined as witnessed cardiac arrest or death <1 hour after the onset of acute symptoms or as an unexpected, unwitnessed death (i.e., during sleep). Cause of cardiac mortality was ascertained either by a physician or by personal contact with a family member.
Statistical Analysis
Continuous variables are expressed as mean ± SD and categorical variables as numbers and percentages. Differences in clinical characteristics between the two groups (ICM and NICM) were tested using unpaired t tests (for continuous variables normally distributed), Mann-Whitney U tests (for continuous variables that violated the assumption of normal distribution), and χ 2 tests (for categorical variables). For comparisons of continuous variables among multiple groups, one-way analysis of variance and least significant difference post hoc tests were used.
Risk factors for ICM were initially assessed using unadjusted logistic regression analysis. To explore the incremental value in predicting ICM to other risk markers, a series of logistic regressions were performed. Hierarchical logistic regression (including three blocks) was then fitted with variables that reached significance ( P < .10) in the previous unadjusted analysis. Clinical variables were first fitted in the logistic model (block 1), and then conventional echocardiographic variables were incorporated into model (block 2). Finally, Ve′ was introduced to the model with the presence of other covariates (block 3). With regard to predictors of cardiac death, a similar model-building strategy was used unless the model was based on Cox proportional-hazards analysis. For the cardiac mortality model, clinical and conventional echocardiographic variables plus ischemic etiology were introduced into blocks 1 and 2 separately. Ve′ was finally estimated in block 3 of the Cox model. The predictive accuracy of each model was illustrated using receiver operating characteristic curves. We also evaluated the prognosis for the etiologies of pump failure death and sudden cardiac death with variables that reached significance ( P < .10) in the previous unadjusted analysis for cardiac mortality.
Cumulative survival curves were established using the Kaplan-Meier method, and the curves were compared using log-rank tests. Receiver operating characteristic curve analysis was performed to identify an optimal cutoff value for predicting patients with ICM. P values < .05 were considered to indicate statistical significance. SPSS version 17.0 for Windows (SPSS, Inc., Chicago, IL) was used for data analysis.
Reproducibility of Measurements
Ten patients were selected to determine the reproducibility of the measurements by two independent physicians for interobserver variability and by the same physician on two separate occasions for intraobserver variability. Variability was determined as the absolute difference between two sets of measurements divided by the mean of the measurements, as expressed in percentage form.
Results
Patient Characteristics
Table 1 lists the clinical characteristics of the 232 study patients. Of the patients, 132 were diagnosed with ICM, and 100 were diagnosed with NICM on the basis of the angiographic results.
| Variable | All patients ( n = 232) | ICM ( n = 132) | NICM ( n = 100) | P |
|---|---|---|---|---|
| Age (y) | 63 ± 14 | 66 ± 13 | 59 ± 15 | <.01 ∗ |
| Men | 177 (76%) | 110 (83%) | 67 (67%) | .01 † |
| NYHA functional class III or IV | 100 (43%) | 61 (46%) | 39 (39%) | .28 † |
| Complete LBBB | 33 (14%) | 11 (8%) | 21 (21%) | .01 † |
| Diabetes mellitus | 108 (47%) | 65 (49%) | 43 (43%) | .42 † |
| Hypertension | 115 (50%) | 69 (52%) | 46 (46%) | .36 † |
| Total cholesterol (mg/dL) | 177 ± 48 | 183 ± 51 | 171 ± 45 | .09 ∗ |
| Current smokers | 54 (23%) | 34 (26%) | 20 (20%) | .32 † |
| Stroke | 22 (10%) | 12 (9%) | 10 (10%) | .94 † |
| Chronic renal insufficiency | 75 (32%) | 48 (36%) | 27 (27%) | .11 † |
| Peripheral vascular disease | 25 (11%) | 17 (13%) | 8 (8%) | .27 † |
| Typical angina | 114 (49%) | 75 (57%) | 39 (39%) | .02 † |
| CAD ‡ | 148 (64%) | 132 (100%) | 16 (16%) | <.01 † |
| Medications | ||||
| ACE inhibitors or ARBs | 190 (82%) | 111 (84%) | 79 (79%) | .32 † |
| β-blockers | 203 (88%) | 116 (88%) | 87 (87%) | .85 † |
| Spironolactone | 86 (37%) | 53 (40%) | 33 (33%) | .26 † |
‡ CAD was considered present for an obstruction of the vessel lumen of >50%.
Patients with ICM were older and had higher percentages of male gender and typical angina than patients with NICM. The NICM group had a higher proportion of complete left bundle branch block (LBBB) than the ICM group. The two groups also did not significantly differ in other clinical characteristics. Additionally, the prevalence of CAD was 16% in patients with NICM. Moreover, the two groups were compatible in terms of medications, including angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers, β-blockers, and spironolactone.
Conventional Echocardiography and PWDTI
Patients with ICM had higher LVEFs and lower LVEDVIs, LVESVIs, LV mass indexes, and WMSIs than patients with NICM ( Table 2 ).
| Variable | All patients ( n = 232) | ICM ( n = 132) | NICM ( n = 100) | P |
|---|---|---|---|---|
| LVEDVI (mL/m 2 ) | 113.7 ± 38.6 | 107.8 ± 38.8 | 121.2 ± 37.1 | .01 ∗ |
| LVESVI (mL/m 2 ) | 85.1 ± 35.9 | 80.1 ± 35.7 | 91.5 ± 35.4 | .02 † |
| LVEF (%) | 26.8 ± 8.4 | 27.8 ± 8.5 | 25.4 ± 8.2 | .03 ∗ |
| LV mass index (g/m 2 ) | 175.7 ± 61.1 | 165.4 ± 58.3 | 188.6 ± 62.4 | .01 ∗ |
| WMSI | 1.90 ± 0.33 | 1.84 ± 0.39 | 1.99 ± 0.21 | .01 † |
| Heart rate (beats/min) | 74 ± 15 | 73 ± 14 | 76 ± 17 | .14 ∗ |
| Mitral inflow velocities | ||||
| E (cm/sec) | 82.7 ± 33.0 | 82.6 ± 33.0 | 82.7 ± 33.1 | 1.00 † |
| E/A | 1.30 ± 0.93 | 1.30 ± 1.00 | 1.31 ± 0.84 | .92 † |
| Deceleration time (msec) | 188.7 ± 71.3 | 187.8 ± 67.5 | 189.8 ± 76.2 | .84 † |
| Isovolumic relaxation time (msec) | 104.1 ± 31.2 | 104.1 ± 30.8 | 104.2 ± 31.8 | .98 ∗ |
| Myocardial performance index | 0.90 ± 0.68 | 0.89 ± 0.67 | 0.91 ± 0.70 | .87 ∗ |
| Velocity dispersion indices | ||||
| Vs′ | 18.3 ± 6.2 | 18.4 ± 6.6 | 18.2 ± 5.8 | .85 † |
| Ve′ | 20.2 ± 13.4 | 27.7 ± 12.3 | 10.5 ± 7.1 | <.01 † |
| Va′ | 21.4 ± 9.7 | 21.2 ± 9.4 | 21.5 ± 10.2 | .84 † |
| Mean mitral annular velocities | ||||
| s′ (cm/sec) | 4.39 ± 1.29 | 5.04 ± 1.48 | 4.81 ± 1.06 | .28 ∗ |
| e′ (cm/sec) | 4.43 ± 1.42 | 4.70 ± 1.46 | 4.09 ± 1.30 | <.01 ∗ |
| a′ (cm/sec) | 6.55 ± 2.39 | 6.73 ± 1.46 | 6.33 ± 2.01 | .22 ∗ |
| LV diastolic filling pressure | ||||
| E/e′ | 20.2 ± 9.9 | 19.3 ± 10.5 | 21.3 ± 9.0 | .13 † |
| Pulmonary wedge pressure (mm Hg) | 22.7 ± 8.9 | 21.8 ± 9.4 | 23.4 ± 8.5 | .35 † |
Significantly higher Ve′ (27.7 ± 12.3 vs 10.5 ± 7.1 P < .01), and e′ (4.70 ± 1.46 vs 4.09 ± 1.30 cm/sec, P < .01) were seen in patients with ICM than in those with NICM. In Figure 1 , the top tracing is for a patient with ICM who presented with Ve′ of 39.3 and died of pump failure. Conversely, the bottom tracing reveals Ve′ of 9.8 in a surviving patient with NICM. Moreover, the values of Vs′, Va′, s′, and a′ were compatible in patients with ICM and those with NICM.
Velocity Dispersion Indices in Patients with CAD: Subgroup Analysis
Ve′ in group 1 was lower than in groups 2, 3, and 4. Although elevated Ve′ in group 4 in comparison with groups 2 and 3 was noted, these groups did not significantly differ from one another. Also, the four groups did not differ in the values of Vs′ and Va′ ( Table 3 ).
| Variable | Group 1 ( n = 84) | Group 2 ( n = 35) | Group 3 ( n = 47) | Group 4 ( n = 66) | ANOVA P |
|---|---|---|---|---|---|
| Vs′ | 18.4 ± 7.4 | 19.3 ± 6.9 | 18.5 ± 7.1 | 18.0 ± 7.5 | .94 |
| Ve′ | 9.9 ± 7.0 | 23.8 ± 10.6 | 26.3 ± 12.3 | 27.1 ± 13.7 | <.01 ∗ , † , ‡ |
| Va′ | 22.3 ± 10.1 | 23.4 ± 9.5 | 20.8 ± 9.4 | 21.7 ± 8.5 | .65 |
∗ Group 1 versus group 2, P < .05 in post hoc analysis.
† Group 1 versus group 3, P < .05 in post hoc analysis.
Incremental Value of DTI to Clinical and Conventional Echocardiographic Parameters for Predicting ICM
Unadjusted analysis revealed age, male gender, complete LBBB, and typical angina to be significant clinical predictors of ICM. Additionally, LVEF, LVEDVI, LVESVI, LV mass index, and WMSI, as well as Ve′, were significant echocardiographic parameters ( P < .05). Multivariate analysis revealed that Ve′, male gender, and typical angina were significantly independent factors in predicting ICM ( Table 4 ).
| Variable | Unadjusted | Multivariate | ||
|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | |
| Clinical | ||||
| Age (per 10-y increment) | 1.48 (1.21–1.80) | <.01 | 1.41 (0.93–2.14) | .10 |
| Male gender | 2.33 (1.27–4.24) | .01 | 5.85 (1.26–27.21) | .02 |
| NYHA functional class III or IV | 1.32 (0.80–2.19) | .28 | ||
| Complete LBBB | 0.31 (0.13–0.74) | .01 | 0.55 (0.09–3.33) | .55 |
| Diabetes mellitus | 1.24 (0.73–2.10) | .42 | ||
| Hypertension | 1.29 (0.75–2.22) | .36 | ||
| Typical angina | 2.08 (1.14–3.77) | .02 | 3.69 (1.14–11.94) | .03 |
| Echocardiography | ||||
| LVEDVI | 0.991 (0.983–0.998) | .01 | 0.98 (0.91–1.05) | .59 |
| LVESVI | 0.989 (0.980–0.997) | .01 | 1.03 (0.94–1.13) | .55 |
| LVEF | 1.04 (1.01–1.07) | .03 | 1.02 (0.89–1.17) | .76 |
| LV mass index | 0.99 (0.98–0.99) | .01 | 0.99 (0.98–1.00) | .09 |
| WMSI | 0.20 (0.07–0.53.) | <.01 | 0.29 (0.06–1.53) | .15 |
| Velocity dispersion index | ||||
| Ve′ | 1.27 (1.19–1.34) | <.01 | 1.33 (1.21–1.45) | <.01 |
Moreover, the incremental value of conventional echocardiography and DTI to clinical risk markers of the risk for ICM was examined by fitting a logistic model that combined clinical information (age, male gender, functional classes, complete LBBB, diabetes, hypertension, and typical angina; area under the curve [AUC], 0.75; 95% confidence interval, 0.68–0.81). Furthermore, information from the conventional echocardiographic examination (LVEF, LVEDVI, LVESVI, LV mass index, and WMSI) was incorporated subsequently, improving the model fit significantly (change in likelihood ratio [ΔLR] χ 2 = 13.34, P < .01; AUC, 0.81; 95% CI, 0.75–0.87). Finally, adding information from Ve′ to the model significantly improved the model fit (ΔLR χ 2 = 97.72, P < .01; AUC, 0.96; 95% CI, 0.91–0.98; Figure 2 A).
