The aim of this study was to investigate the role of segmental longitudinal strain for the diagnosis of coronary artery disease (CAD) assessed by automated function imaging.
One hundred fifty-two subjects (mean age, 63 ± 12 years; 77 men) referred for assessment of cardiac function under suspicion of CAD were recruited for this study. Patients with left ventricular dysfunction or with acute coronary syndromes were excluded.
Peak systolic global longitudinal strain (GLS) was significantly decreased in patients with CAD. Peak segmental longitudinal strain difference (LSD) and its ratio to peak systolic GLS were significant higher in patients with CAD. The areas under receiver operating characteristic curves for the diagnosis of CAD were 0.813 for peak systolic GLS, 0.851 for the number of abnormal segments, 0.805 for peak segmental LSD, and 0.862 for the ratio of peak segmental LSD to peak systolic GLS. Using 1.0 as a cutoff point for the ratio of peak segmental LSD to peak systolic GLS, sensitivity was 77.3% and specificity 79.2%.
This study suggests that it may be possible to assess CAD with strain by automated function imaging, but further larger scale studies are needed to confirm this.
Assessment of regional left ventricular (LV) myocardial function is important in patients with coronary artery disease (CAD). However, echocardiographic assessment of regional LV function on the basis of visual analysis is subjective and observer dependent. Measurements of myocardial deformation by strain and strain rate imaging have provided alternatives for the assessment of LV function. Strain Doppler imaging has been proven to be useful in the assessment of regional LV myocardial function in patients with CAD. Speckle-tracking echocardiography has been introduced as a method for angle-independent quantification of myocardial strain on the basis of grayscale images. Myocardial deformation imaging has great promise to improve the objective quantification and characterization of regional function in patients with CAD. Strain has been demonstrated to be superior to velocity and displacement for the quantification of regional myocardial function during acute myocardial ischemia. One study showed that LV longitudinal and radial strain decreased earlier in patients with heart failure with normal ejection fractions than circumferential strain and twist. In another study on the effects of myocardial infarction on different directions of myocardial deformation, longitudinal strain and strain rate were decreased in both transmural and subendocardial infarction, but radial and circumferential strain and strain rate were relatively preserved in subendocardial infarction. All of these results suggest that longitudinal deformation of the myocardium is the earliest and most sensitive to LV dysfunction.
Recently, a novel algorithm of speckle-tracking echocardiography, called automated function imaging (AFI), has been developed to facilitate the clinical application of LV longitudinal strain. By using AFI, peak systolic global longitudinal strain (GLS) and peak systolic segmental longitudinal strain can be obtained simply and automatically. Peak systolic GLS has been noted to be closely correlated with LV ejection fraction. GLS has been also noted to be significantly correlated with infarction size in patients with chronic CAD.
In this study, we evaluated the diagnostic value of segmental longitudinal strain obtained by AFI in patients with CAD. Our previous study showed that subtle systolic dysfunction could be detected by decreased GLS in patients with heart failure and preserved ejection fractions. We hypothesized that subtle changes of ischemic segment could be detected by longitudinal strain, even without stress echocardiography. We also hypothesized that longitudinal strain could be improved in nonischemic segments because of LV remodeling and that the gradient of longitudinal strain between segments would be improved in patients with chronic CAD. The ability to diagnose CAD by longitudinal strain could therefore be increased, even without stress echocardiography, by the combination of different strain parameters, such as the peak segmental longitudinal strain difference (LSD).
We recruited 152 consecutive patients who were referred to our echocardiographic laboratory for the evaluation of LV function under suspicion of CAD (with symptoms of chest pain or histories of CAD). CAD was diagnosed if a patient had (1) previous coronary angiography with significant coronary stenosis and without complete revascularization and (2) significant (>75% stenosis in at least one coronary artery) CAD by coronary angiography after entry into the study. Patients with negative coronary angiographic results or atypical chest pain with negative treadmill exercise results were used as the control group. Patients were excluded if (1) their LV ejection fractions were <50%, (2) they had hypertrophic cardiomyopathy, (3) they had significant valvular heart disease, (4) they had acute coronary syndromes, or (5) there was an equivocal diagnosis of CAD by noninvasive examination. The study was approved by the human research committee of our hospital, and informed consent was obtained from all subjects.
Standard echocardiography was performed with Doppler studies (Vivid 7; GE Vingmed Ultrasound AS, Horten, Norway) with a 3.5-MHz multiphase-array probe in subjects lying in the left lateral decubitus position. The chamber dimensions were measured by the two-dimensionally guided M-mode method, and LV ejection fraction was measured using the biplane Simpson’s method according to the recommendations of the American Society of Echocardiography. LV mass was measured by the M-mode method. Transmitral Doppler flow velocity was obtained from an apical four-chamber view, and peak early filling velocity (E), peak atrial velocity (A), E/A ratio, and mitral deceleration time were recorded. Pulse Doppler tissue imaging was performed from the medial annulus, and peak systolic annular velocity (Sa), early diastolic annular velocity (Ea), and atrial annular velocity (Aa) were measured. Two-dimensional images focusing on the left ventricle were acquired from apical four-chamber, two-chamber, and long-axis views for three cardiac cycles and digitally stored with frame rates of 60 to 100 frames/sec. The images were analyzed offline using computer software (EchoPAC version 6.0; GE-Vingmed Ultrasound AS).
Images stored in digital loops were analyzed by AFI software at a workstation (EchoPAC workstation, BT08). Analysis was performed for each of the apical views, with the operator manually identifying three points: two on each side of the mitral valve and a third at the apex of the left ventricle. The software automatically detected the endocardium and tracked myocardial motion during the entire cardiac cycle. U-shaped regions of interest (ROIs) were created on all three apical views, and the width of the ROI was adjusted to cover the whole thickness of the LV wall. The software automatically checked the tracking quality within the ROI, and we double checked visually. Tracking quality was determined by the software automatically. If tracking was poor, the operator could repeat the tracking procedure or adjust the ROI by moving the endocardial lining or by changing the ROI width to achieve better tracking. After this adjustment, the software would recheck tracking quality until the tracking quality was satisfactory. GLS and segmental longitudinal strain were obtained automatically by the software after identified the timing of aortic valve closure. The left ventricle was divided into 18 segments, and peak systolic longitudinal strain of each segment was obtained. GLS was calculated from the average of segmental strain, and LSD was defined as the difference of longitudinal strain between highest strain and lowest strain among all of the segments. According to a previous study, abnormal segments were defined as those segments with longitudinal strain > −15%. Previous studies have shown that postsystolic shortening is potentially useful in the diagnosis of CAD. We also tested the diagnostic value of postsystolic shortening in this study. The postsystolic index (PSI) was calculated as [(postsystolic peak longitudinal strain − end-systolic strain)/end-systolic strain] × 100%. Total PSI from all segments was used for analysis.
To assess intraobserver and interobserver variability, GLS was reevaluated in 10 randomly selected patients using Bland-Altman limits of agreement and interclass correlation coefficients.
Differences between the CAD and control groups were compared using independent t tests for continuous variables and χ 2 tests for categorical variables. Correlations between LV longitudinal strain with each other or clinical factors were assessed using Pearson’s correlation test. Significant factors associated with impaired strain, including age, gender, hypertension, diabetes, and LV mass index, were added in multivariate analysis. Multiple backward stepwise logistic regression analysis controlling for these factors was used for the assessment of the independency of strain parameters associated with CAD. Receiver operating characteristic curve analysis was used to select cutoff values to distinguish patients with CAD from controls. One-way analysis of variance was used to test the differences between one-vessel, two-vessel, and three-vessel disease. All data are presented as mean ± SD. A P value < .05 was considered statistically significant. All analysis was performed with SPSS version 11.5 for Windows (SPSS, Inc., Chicago, IL).
Comparison Between Patients With CAD or Angiographically Normal Coronary Arteries
In total, 152 patients (mean age, 63 ± 12 years; 77 men) were included in this study, 75 (49%) of whom had CAD. Comparing between patients with CAD or angiographically normal coronary arteries, patients with CAD were older and had higher serum creatinine and more hypertension and diabetes ( Table 1 ). Among general echocardiographic parameters, patients with CAD had higher LV mass indexes, larger left atrial dimensions, higher mitral A velocities, longer deceleration time, lower Ea values, and higher E/Ea ratios ( Table 2 ).
|Variable||Patients with CAD||Controls||P|
|Age (years)||67 ± 11||60 ± 10||<.001|
|Men||44 (59%)||33 (43%)||.051|
|Hypertension||56 (75%)||36 (47%)||<.001|
|Diabetes mellitus||22 (29%)||11 (14%)||.021|
|Smoking||25 (33%)||17 (22%)||.107|
|Dyslipidemia||27 (36%)||20 (26%)||.161|
|Body weight (kg)||65.5 ± 12.5||64.6 ± 12.3||.705|
|Systolic blood pressure (mm Hg)||139 ± 22||135 ± 22||.365|
|Diastolic blood pressure (mm Hg)||80 ± 15||78 ± 16||.476|
|Glucose (mg/dL)||138 ± 81||119 ± 43||.151|
|Creatinine (mg/dL)||1.41 ± 1.32||0.88 ± 0.42||.009|
|Cholesterol (mg/dl)||185 ± 54||188 ± 32||.767|
|Triglycerides (mg/dL)||149 ± 86||125 ± 77||.194|
|High-density lipoprotein (mg/dL)||49 ± 10||54 ± 17||.202|
|Low-density lipoprotein (mg/dL)||117 ± 43||109 ± 33||.361|
|Variable||Patients with CAD||Controls||P|
|LV end-diastolic volume (mL) ∗||70.3 ± 24.1||65.0± 15.8||.127|
|LV end-systolic volume (mL) ∗||30.0 ± 12.9||18.6 ± 6.9||.056|
|LV ejection fraction (%) ∗||70.1 ± 8.5||71.7 ± 6.9||.210|
|LV mass index (g/m 2 )||109.1 ± 40.2||91.4 ± 26.7||.003|
|Left atrial dimension (cm)||3.6 ± 0.6||3.3 ± 0.5||.001|
|E (m/s)||0.69 ± 0.23||0.74 ± 0.17||.237|
|A (m/s)||0.92 ± 0.22||0.74 ± 0.17||.006|
|E/A ratio||0.77 ± 0.28||0.93 ± 0.26||.003|
|Mitral deceleration time (ms)||220.5 ± 74.7||188.2 ± 51.2||.009|
|Sa (m/s)||0.08 ± 0.01||0.09 ± 0.02||.096|
|Ea (m/s)||0.06 ± 0.03||0.08 ± 0.03||.006|
|Aa (m/s)||0.10 ± 0.03||0.25 ± 1.08||.373|
|E/Ea ratio||13.01 ± 5.20||10.00 ± 2.93||.002|
|GLS (%)||−16.4 ± 4.9||−20.6 ± 2.8||<.001|
|LSD (%)||26.6 ± 8.9||17.2 ± 5.0||<.001|
|LSR||1.61 ± 0.97||0.84 ± 0.24||<.001|
|Number of abnormal segments||7.0 ± 4.2||2.4 ± 2.6||<.001|
|PSI (%)||182 ± 161||81 ± 63||<.001|
Differences in LV Longitudinal Strain Between Patients With CAD or Angiographically Normal Coronary Arteries
GLS was significantly impaired in patients with CAD (−16.4 ± 4.9% vs −20.6 ± 2.8%, P < .001). LSD was increased in patients with CAD (26.5 ± 8.9% vs 17.2 ± 5.0%, P < .001) and significantly correlated with GLS ( r = 0.229, P = .005). For the correction of the influence of GLS on LSD, we used the ratio of LSD to GLS (LSR). LSR was significantly higher in patients with CAD (1.61 ± 0.97 vs 0.84 ± 0.24, P < .001). The number of abnormal segments and PSI were higher in patients with CAD ( Table 2 ). Age was significantly correlated with impaired strain ( r = 0.206, P = .011), and strain was impaired in patients with hypertension (GLS, −17.9 ± 4.2% vs −19.5 ± 4.8%, P = .040) and diabetes (GLS, −16.6 ± 6.3% vs −19.1 ± 3.7%, P = .040). We further compared strain parameters between patients with CAD or angiographically normal coronary arteries in older patients (aged ≥ 60 years), in patients with hypertension, and in patients with diabetes. Most of the strain parameters were impaired in patients with CAD compared with control subjects whenever patients were older, had hypertension, or had diabetes ( Table 3 ). Using multivariate logistic regression analysis controlling for age, LV mass index, sex, hypertension, and diabetes, we found that GLS, LSD, the number of abnormal segments, LSR, and PSI were all independently associated with CAD ( Table 4 ).
|Variable||CAD ( n = 58)||Control ( n = 38)||CAD ( n = 56)||Control ( n = 36)||CAD ( n = 22)||Control ( n = 11)|
|GLS (%)||−16.6 ± 3.8||−20.3 ± 3.4 †||−16.4 ± 4.2||−20.2 ± 9.1 †||−15.1 ± 6.5||−19.7 ± 4.7 ∗|
|LSD (%)||26.8 ± 9.3||18.1 ± 6.1 †||26.6 ± 9.1||18.1 ± 5.1 †||26.3 ± 9.0||17.6 ± 2.9 †|
|LSR||1.68 ± 0.71||0.90 ± 0.30 †||1.72 ± 0.72||0.91 ± 0.27 †||1.53 ± 1.51||0.93 ± 0.22|
|Number of abnormal segments||7.3 ± 4.0||2.7 ± 3.2 †||7.5 ± 4.4||2.8 ± 3.3 †||7.1 ± 4.1||3.6 ± 5.1 ∗|
|PSI (%)||187 ± 162||86 ± 66 †||196 ± 179||89 ± 70 †||154 ± 99||77 ± 51 ∗|
|Variable||Odds ratio per 1-SD increment||95% confidence interval|
|Number of abnormal segments||4.66||2.32–9.34|
Diagnostic Value of LV Longitudinal Strain in CAD
The areas under receiver operating characteristic curves were 0.647 for hypertension, 0.578 for diabetes, 0.684 for age, 0.813 for GLS, 0.805 for LSD, 0.851 for the number of abnormal segments, 0.862 for LSR, and 0.763 for PSI ( Figure 1 ). Using >−19% as a diagnostic value for GLS, sensitivity was 74.7% and specificity 80.5% for the diagnosis of CAD. Using ≥20% for LSD, sensitivity was 74.7% and specificity 74.0%. Using >1.0 for LSR, sensitivity was 77.3% and specificity 79.2%. Using four or more abnormal segments as an index for CAD, sensitivity was 77.3% and specificity 79.2%. Using >89% for PSI, sensitivity was 70.8% and specificity 68.5% ( Table 5 ). We then combined different parameters to improve the diagnostic rate. A combination of LSR > 1.0 and four or more abnormal segments increased specificity to 89.6% but decreased sensitivity to 68.0%. A combination of GLS > −19% and PSI > 89% improved specificity to 95.9% but decreased sensitivity to 55.4%. A combination of LSD ≥ 20% and PSI > 89% increased sensitivity to 86.5% and specificity 61.6% ( Table 5 ). The presence of visual wall motion abnormality had 18.7% sensitivity and 100% specificity for CAD.