Relationship between Left Ventricular Longitudinal Deformation and Clinical Heart Failure during Admission for Acute Myocardial Infarction: A Two-Dimensional Speckle-Tracking Study


Heart failure (HF) complicating acute myocardial infarction (MI) is an ominous prognostic sign frequently caused by left ventricular (LV) systolic dysfunction. However, many patients develop HF despite preserved LV ejection fractions. The aim of this study was to test the hypothesis that LV longitudinal function is a stronger marker of in-hospital HF than traditional echocardiographic indices.


A total of 548 patients with acute MIs were evaluated (mean age, 63.2 ± 11.7 years; 71.6% men). Within 48 hours of admission, comprehensive echocardiography with assessment of global longitudinal strain (GLS) was performed, along with measurements of N-terminal pro–brain natriuretic peptide.


A total 89 patients (16.2%) had in-hospital HF assessed by Killip class > 1 in whom GLS was significantly impaired compared with patients without in-hospital HF (Killip class 1) (−14.6 ± 3.3% vs −10.1 ± 3.5%, P < .0001). In stepwise multiple logistic regression analysis including age, known HF, three-vessel disease, involvement of the left anterior descending coronary artery, episodes of atrial fibrillation, renal function, N-terminal pro–brain natriuretic peptide, troponin T level, LV ejection fraction, wall motion score index, and diastolic dysfunction indices, GLS emerged as the strongest marker of clinical HF (odds ratio, 1.47; 95% confidence interval [CI], 1.33–1.62; P < .0001). GLS remained independently associated with in-hospital HF in patients with LV ejection fractions > 40% (odds ratio, 1.33; 95% CI, 1.14–1.54; P < .05) and improved the C-statistic over other important covariates significantly (0.87 [95% CI, 0.82–0.91] vs 0.82 [95% CI, 0.76–0.89], P = .02).


Global longitudinal function assessed by GLS is significantly impaired in patients with MIs with in-hospital HF, and multivariate analysis suggests that reduced GLS is the single most powerful marker of manifest LV hemodynamic deterioration in the acute phase of MI.

Modern reperfusion strategies have significantly decreased the loss of viable myocardium associated with acute myocardial infarction (MI), but in-hospital congestive heart failure (HF) remains a significant predictor of poor short-term and long-term prognoses. Acute HF develops as a result of myocyte loss with depressed cardiac output and/or because of abnormally elevated filling pressure.

Acute HF complicating MI despite preserved left ventricular (LV) ejection fraction (LVEF) is associated with a doubling in the risk for all-cause mortality. The discrepancy between apparently minor acute myocardial injury and overt HF is poorly understood, but the burden of comorbid conditions adversely affecting myocardial relaxation properties, including hypertension, diabetes, and diffuse atherosclerosis, has been proposed to decrease the tolerance of even a minor loss of contractile function that is not reflected by decreased LVEF. Deformation analysis using two-dimensional speckle-tracking allows the quantification of systolic longitudinal fiber shortening, which may be expressed as regional and global longitudinal strain (GLS). The longitudinal fibers in the subendocardial layer are more sensitive to ischemia and wall stress and can exhibit abnormal contraction patterns in the setting of apparently normal LVEF. Furthermore, deformation properties of the myocardium in the setting of MI have been shown to correlate with infarct size.

Echocardiographic findings in relation to in-hospital HF complicating MI with contemporary revascularization management are not well characterized. Recently, deformation analysis has been related to adverse prognosis in patients admitted with acute HF, in those with ST-segment elevation MI (STEMI), and in stable patients with chronic HF. Furthermore, recent data suggest that patients with HF with preserved ejection fraction are characterized by abnormal global LV longitudinal deformation. To test the hypothesis that impaired GLS reflects in-hospital HF to a greater extent than traditional measures of systolic and diastolic dysfunction, we prospectively performed comprehensive echocardiographic analyses and neurohormonal assessment in a large cohort of patients with MIs. Furthermore, we assessed the importance of GLS in relation to in-hospital HF in patients with preserved LVEFs.


Study Design and Patient Population

We conducted a prospective study of patients referred to our center for invasive coronary angiography for either STEMI or non–ST-segment elevation MI (NSTEMI) from September 2009 to October 2010. All patients provided written informed consent before transthoracic echocardiographic examination and blood sampling. Exclusion criteria were age < 18 years, noncardiac disease with a life expectancy < 1 year, or inability to provide written informed consent.

On the basis of hospital records obtained at admission, diabetes mellitus, hypertension, history of ischemic heart disease, prior MI, and preexisting congestive HF were registered. Findings in relation to coronary angiography, including culprit lesion, number of diseased vessels, left main coronary artery involvement, and type of revascularization (percutaneous coronary intervention [PCI], coronary artery bypass grafting, or no intervention) were registered. Clinical events from the arrival of emergency services and during hospitalization were recorded, including the occurrence of supraventricular arrhythmias. An episode of atrial fibrillation (AF) during hospitalization was registered as a complication related to the MI if there was no medical history of AF.

Peripheral samples of plasma were obtained within 24 hours of echocardiography. Analysis of N-terminal pro–brain natriuretic peptide (NT-proBNP) was performed on the commercially available Modular Analytics E170 NT-proBNP immunoassay (Roche Diagnostics GmbH, Mannheim, Germany) immediately after blood sampling. Additional biochemical workup included creatinine, hemoglobin, and peak troponin T during the hospital stay. Estimated glomerular filtration rate was measured using the four-variable Modification of Diet in Renal Disease equation.

The primary outcome variable was in-hospital HF assessed according to the Killip classification as follows: no sign of HF (Killip class 1), basilar rales and/or radiologic signs of pulmonary congestion (Killip class 2), pulmonary edema (Killip class 3), and cardiogenic shock (Killip class 4). Evidence of HF was assessed continuously throughout the admission both during daily rounds and by the trained staff members in the coronary care unit. An independent reviewer without knowledge of echocardiographic results adjudicated the diagnosis of HF. Timing of HF was classified as HF at presentation if patients had objective HF on admission or incident if HF developed after admission. For the present study, in-hospital HF was considered in all patients with Killip class > 1, whether on presentation or incident during hospitalization. Patients with known stable chronic HF, regardless of diuretic therapy before admission, were not included in the in-hospital HF group unless they experienced worsening decompensation. The study was approved by the regional scientific ethics committee (reference number H-D-2009-063).

Echocardiography and Two-Dimensional Speckle-Tracking

Echocardiography was performed within 48 hours of admission to our institution. Echocardiographic cine loops were obtained by recording three consecutive heart cycles. All examinations were performed using a Vivid e9 (GE Vingmed Ultrasound AS, Horten, Norway). Images were obtained at a frame rate of ≥60 frames/sec and digitally transferred to a remote workstation for offline analysis (EchoPAC BT 11.1.0; GE Vingmed Ultrasound AS). All analyses were performed by a single experienced operator (M.E.) without knowledge of Killip class and blinded to clinical, biochemical, and coronary angiographic information.

Two-dimensional parasternal images were used to determine LV dimensions and wall thickness. Maximum left atrial volume index (LAVi) was determined from the biplane area0length method just before the opening of the mitral valve (MV), and LV volumes were determined using the biplane Simpson model. Wall motion scoring was performed by dividing the left ventricle into 16 segments, and each segment was assigned a score on the basis of myocardial thickening (1 = normal or hyperkinesis, 2 = hypokinesis, 3 = akinesis). Wall motion score index (WMSI) was calculated from the average score of all segments. LV mass was calculated from the LV linear dimensions in the parasternal view. Volumetric and dimensional measurements of the left ventricle and left atrium were indexed to body surface area when appropriate. All volumetric analyses were performed in accordance with European Association of Echocardiography and American Society of Echocardiography recommendations.

Color Doppler examination of the MV was performed in the apical window and if more than trivial mitral regurgitation (MR) was present, it was quantified by calculating the effective regurgitant orifice area using the proximal isovelocity surface area method. Effective regurgitant orifice area < 0.20 cm 2 was considered mild, 0.20 to 0.40 cm 2 moderate, and >0.40 cm 2 severe MR. If the effective regurgitant orifice area could not be determined, MR was considered mild when regurgitant jet area occupied >5% and <20% of the left atrial area, moderate when regurgitant jet area occupied >20% and <40%, and severe when regurgitant jet area occupied >40%. The presence of an eccentric jet raised the grade by one degree. Doppler recordings of mitral inflow were performed by placing a 2.5-mm sample volume at the tip of the MV leaflets during diastole and recording the pulsed-wave Doppler signal. Peak velocities of early (E) and atrial (A) diastolic filling and MV deceleration time were measured, and the E/A ratio was calculated. Continuous-wave Doppler recordings of the LV outflow tract were obtained, and aortic valve opening and closure times were measured. Pulsed-wave Doppler tissue imaging recordings were performed at the lateral and medial mitral annulus using a 2.5-mm sample volume with measurements of myocardial peak early (e′). The mean E/e′ ratio was calculated from an average of lateral and medial values of E/e′.

Two-dimensional speckle-tracking was performed using a semiautomatic algorithm (Automated Function Imaging; GE Healthcare, Waukesha, WI). Briefly, manual positioning of three points (two annular and one apical) was performed in each of the three apical projections, enabling the software to semiautomatically track the myocardium throughout the heart cycle. Each ventricular wall was subsequently divided into three segments, for a total of 17 segments covering the entire myocardium. Careful inspection of tracking and manual correction, if needed, was performed, and in case of unsatisfactory tracking, the segment was excluded from the analysis. Longitudinal strain curves were generated for each segment, and from the average of all maximum values, GLS was calculated.

Statistical Analysis

All data are reported as mean ± SD or as median (interquartile range [IQR]). NT-proBNP was logarithmically transformed (log 10 ) to stabilize the variance before entering the models. Categorical variables (presented as absolute values and percentages) were compared using χ 2 tests (or Fisher’s exact tests when indicated). Continuous variables were compared using Student’s t tests. All tests were two sided, and statistical significance was defined as P < .05. The variables included in multivariable logistic regression analysis to identify factors with independent associations with in-hospital HF were age as a continuous variable, sex, diabetes, known ischemic heart disease, known chronic HF, episodes of AF, presence of ST-segment elevation, left anterior descending coronary artery involvement, three-vessel disease and/or left main coronary artery involvement, peak troponin T, log NT-proBNP, estimated glomerular filtration rate, WMSI, LVEF, MR severity, E/e′, LAVi, MV deceleration time, and GLS. We performed the multiple logistic regression analyses using stepwise, forward, and backward elimination (stepwise method, with P < .10 for inclusion and P < .05 for retention). The performance of the final parsimonious model was assessed with the C-statistic. We also evaluated the added model performance with sequential addition to a baseline clinical model consisting of age, known chronic HF, episodes of AF, left anterior descending coronary artery involvement, three-vessel disease, peak troponin T, estimated glomerular filtration rate, and log NT-proBNP in the following sequence: (1) clinical model + LVEF + LAVi, (2) clinical model + LVEF + LAVi + E/e′ + MV deceleration time, and (3) clinical model + LVEF + LAVi + E/e′ + MV deceleration time + GLS. The incremental model performance was assessed with the Akaike information criterion and −2 log likelihood.

Furthermore, additional separate multiple logistic regression analyses were performed to identify echocardiographic and neurohormonal factors independently associated with in-hospital HF in patients with preserved LVEFs (>40%). Two separate multiple regression models were constructed for GLS and LVEF, but because of the small number of patients with in-hospital HF in this group, we restricted the number of covariates to adjustment for age, NT-proBNP, troponin T, LAVi, and episodes of AF.

Internal validation was performed on the total study population by creating 200 bootstrap samples with random replacement. The stepwise model mentioned above was trained on each bootstrap sample, and the performance of each of these models was assessed in the original data set. The deviation between the C-statistic in the bootstrap sample and in the original data set was averaged for all 200 bootstrap samples and considered the average optimism. The C-statistic of the original parsimonious model was then corrected for the optimism and considered a nearly unbiased estimate of the internal validity. The variables selected by the stepwise modeling for each bootstrap sample were counted to identify factors that were consistently associated with in-hospital HF and to evaluate the stability of the modeling. SAS version 9.2 (SAS Institute, Inc., Cary, NC) was used for all data analyses.


Patient Population

The total study population consisted of 611 patients. Twenty-two patients were excluded from the analysis because of AF ( n = 18) and ventricular paced rhythm ( n = 4) during echocardiography. Forty-one patients were excluded because of poor image quality or technical limitations of echocardiography causing three or more myocardial segments to be incorrectly tracked by the speckle-tracking algorithm. Thus, 548 patients (90%) were included in the analyses (mean age, 63.2 ± 11.7 years; 71.6% men), of whom 89 (16.2%) had in-hospital HF as assessed by Killip class > 1. Incident HF was seen in 44 patients and HF on presentation in 45 patients. Among patients with STEMIs ( n = 370 [67.5%]) and NSTEMIs ( n = 178 [32.5%]), in-hospital HF was seen in 63 (17%) and 26 (15%) patients, respectively. Patients experiencing in-hospital HF were older, had more extensive myocardial injuries as assessed by peak troponin T, more often had left anterior descending coronary artery involvement, and had more in-hospital AF. The median duration of hospital stay was 5 days (IQR, 4–6 days). Patients with STEMIs were transferred directly to our institution after verification of ST-segment elevation on electrocardiograms taken by the emergency medical response team and electronically transmitted. If patients presented to local hospitals with ST-segment elevation, they were immediately transferred to our institution. In all cases, preprocedural treatment with antithrombotic therapy and heparin was initiated as soon as ST-segment elevation was diagnosed. The median symptom-to-balloon time was 197 min (IQR, 147–310 min). Patients with NSTEMIs were stabilized with antithrombotic therapy, β-blockade, and nitrates if indicated and transferred to our institution within 48 hours. The baseline clinical and echocardiographic characteristics are shown in Table 1 .

Table 1

Clinical and echocardiographic characteristics according to in-hospital congestive HF

Variable All Patients ( n = 548) No HF ( n = 459) HF ( n = 89)
Age (y) 63.2 ± 11.7 61.9 ± 11.4 69.9 ± 10.8
Men 393 (71.6%) 337 (73.4%) 55 (61.8%)
Killip class > 1 89 (16.2%) 0 89 (100%)
Episodes of AF 38 (6.9%) 21 (4.6%) 17 (19.1%)
Medical history
History of hypertension 253 (46.2%) 209 (45.5%) 44 (49.4%)
Smoking 368 (66.7%) 315 (68.6%) 53 (59.5%)
Ischemic heart disease 95 (17.3%) 79 (17.2%) 16 (18.0%)
Diabetes 75 (13.7%) 61 (13.3%) 14 (15.7%)
Known HF 33 (6.0%) 21 (4.6%) 12 (13.6%)
eGFR (mL/min/1.73 m 2 ) 90.6 ± 28.6 92.8 ± 27.8 79.2 ± 30.3
Peak troponin T (pg/L) 2.1 (0.5–5.5) 2.0 (0.5–4.6) 4.6 (0.6–9.7)
NT-proBNP (pmol/L) 115.5 (48.3–246.0) 98.7 (42.4–187.0) 409.0 (187.0–732.0)
Type of infarction
STEMI 370 (67.5%) 307 (66.9%) 63 (70.8%)
NSTEMI 178 (32.5%) 152 (33.1%) 26 (29.2%)
Angiographic findings
LAD involvement 208 (38.1%) 156 (34.0%) 52 (58.4%)
Three-vessel disease or LM culprit 90 (16.2%) 67 (14.6%) 23 (25.8%)
Treatment decision
PCI 103 (18.8%) 82 (17.9%) 21 (23.6%)
Primary PCI 333 (60.8%) 284 (61.9%) 49 (55.1%)
No invasive treatment 112 (20.4%) 93 (20.3%) 19 (21.4%)
Additional CABG 44 (8.1%) 33 (7.2%) 11 (12.4%)
LVEF (%) 50.6 ± 10.7 52.1 ± 9.8 43.2 ± 12.2
GLS (%) −13.9 ± 3.7 −14.6 ± 3.3 −10.1 ± 3.5
WMSI 1.46 ± 0.30 1.41 ± 0.27 1.75 ± 0.30
Moderate to severe MR 3 (0.55%) 2 (0.44%) 1 (0.18%)
LAVi (mL/m 2 ) 35.9 ± 11.0 35.0 ± 10.1 40.7 ± 13.9
MV deceleration time (msec) 179.9 ± 50.7 183.9 ± 49.6 158.8 ± 51.9
E/e′ mean of lateral and medial 11.5 ± 5.3 10.8 ± 4.4 15.3 ± 7.8

CABG , Coronary artery bypass grafting; eGFR , estimated glomerular filtration rate; LAD , left anterior descending coronary artery; LM , left main coronary artery.

Data are expressed as mean ± SD, as number (percentage), or as median (IQR).

P < .001 and P < .05 compared with patients without evidence of HF.

Echocardiographic and Clinical Correlates of In-Hospital HF

Patients with in-hospital HF had significantly poorer longitudinal function (−10.1 ± 3.3% vs −14.5 ± 3.5%, P < .0001), lower LVEFs (43.2 ± 12.2% vs 52.1 ± 9.8%, P < .0001), and higher WMSIs (1.75 ± 0.30 vs 1.41 ± 0.27, P < .0001). Diastolic dysfunction was more frequent in patients with in-hospital HF with lower E/e′ ratios, shorter MV deceleration times, and larger values of LAVi ( Table 1 ). Of the 89 patients with in-hospital HF, 52 were in Killip class 2, 27 in Killip class 3, and 10 in Killip class 4. There was progressive deterioration in GLS with increasing Killip class, and this was significant from Killip class 1 to 2 (−14.6 ± 3.3% vs −11.1 ± 3.3%, P < .0001) and class 2 to 3 (−11.1 ± 3.3% vs −8.3 ± 2.8%, P < .005), but there was no significant difference from class 3 to 4 after adjustment for multiple comparisons with Bonferroni correction. Patients with HF on presentation did not exhibit significantly different GLS (−10.3 ± 4.0% vs −9.8 ± 2.8%, P = NS) or LVEFs (44.2 ± 13.4% vs 41.8 ± 11.1%, P = NS) compared with patients with incident HF. Thus, multivariable modeling was conducted with in-hospital HF at any time as the dependent variable.

Significant independent predictors were consistently, in descending order of significance (on the basis of Wald χ 2 values), GLS, age, troponin T, LAVi, and episodes of AF (overall C-statistic = 0.87). The results of the multivariate logistic regression analysis with the significant independent predictors of in-hospital HF with associated odds ratios and 95% confidence intervals (CIs) are shown in Table 2 . Analyzing STEMI and NSTEMI separately did not alter the results. Internal validation by bootstrapping of the overall model revealed minimal overoptimism (C-statistic = 0.84; optimism correction, 0.03) and confirmed GLS as the most important variable associated with in-hospital HF (selected in 99% of bootstrap samples).

Table 2

Univariate and multivariable logistic regression of in-hospital HF

Variable Univariate Multivariable ∗,†
χ 2 OR 95% CI OR 95% CI χ 2 P
Age (per year) 32.7 1.07 1.04–1.09 1.05 1.02–1.08 9.85 <.05
Female sex 4.9 1.72 1.07–2.77
Known HF 9.6 3.26 1.54–6.91 NS
eGFR (per 10 mL/min/1.73 m 2 decrease) 16.32 1.21 1.10–1.32 NS
Episodes of AF 21.8 5.20 2.60–10.4 3.10 1.2–8.4 5.68 <.05
Three-vessel disease or LM culprit 5.8 1.97 1.14–3.40 NS
LAD involvement 17.1 2.67 1.68–4.25 NS
Peak troponin T (per 1 pg/L increase) 28.7 1.13 1.08–1.18 1.06 1.00–1.12 8.36 <.05
Log NT-proBNP (per 0.1 increase) 66.4 1.30 1.22–1–38 NS
LVEF (per 5.0% decrease) 45.5 1.49 1.33–1.68 NS
WMSI (per 0.1 increase) 72.7 1.54 1.39–1.69 NS
LAVi (per 1 mL/m 2 increase) 17.6 1.04 1.02–1.06 1.04 1.01–1.07 7.09 <.05
E/e′ ratio 39.8 1.14 1.09–1.18 NS
GLS (per absolute % increase) 77.3 1.50 1.37–1.64 1.47 1.33–1.62 45.5 <.0001

eGFR , Estimated glomerular filtration rate; LAD , left anterior descending coronary artery; LM , left main coronary artery.

Odds ratios are per 1-unit increase unless stated otherwise.

C-statistic = 0.88, global χ 2 = 76.0.

Hosmer and Lemeshow goodness-of-fit test, P = .45.

The incremental value of GLS was assessed in four modeling steps shown in Figure 1 . Addition of GLS decreased the Akaike information criterion and −2 log likelihood significantly ( P < .001). In direct comparison the C-statistic of GLS outperformed that of LVEF (0.84 [95% CI, 0.79–0.88] vs 0.74 [95% CI, 0.65–0.78], P < .0001), and when adding NT-proBNP to LVEF, the C-statistic was still lower than for GLS alone, although the difference was not significant (0.83 [95% CI, 0.76–0.87] vs 0.84 [95% CI, 0.79–0.88], P = .26) ( Figure 2 ).

Figure 1

Incremental improvement in model performance as assessed by the −2 log likelihood and Akaike information criterion. Addition of LVEF and LAVi significantly improved a model including clinical information (age, history of heart failure, multivessel disease, left anterior descending coronary artery involvement, episodes of AF, troponin T, estimated glomerular filtration rate, and log NT-proBNP). Doppler indices (E/e′ ratio and MV deceleration time) did not improve the model, whereas the addition of GLS yielded significantly better model performance.

Figure 2

Receiver-operating characteristic curves depicting the performance of LVEF, LVEF plus log NT-proBNP, and GLS in relation to in-hospital HF. AUC , Area under the curve.

In-Hospital HF in Patients with Preserved Ejection Fractions

A total of 464 patients (85%) had LVEFs > 40% (mean age, 62.4 ± 11.7 years; 72% men), of whom 53 (11.2%) experienced in-hospital HF. An example of a patient with a preserved LVEF, in-hospital HF, and impaired GLS is given in Figure 3 . Significantly impaired GLS was seen in patients with in-hospital HF (−11.9 ± 2.9% vs −15.1 ± 3.0%, P < .0001), and median NT-proBNP levels were significantly higher (283.0 pmol/L [IQR, 164.0–599.0 pmol/L] vs 87.4 pmol/L [IQR, 37.2–160.0 pmol/L], P < .0001). Of the 53 patients with in-hospital HF, 38 were in Killip class 2, nine in Killip class 3, and five in Killip class 4. There was no significant difference in GLS from Killip class 2 to 3 (−11.9 ± 3.1% vs −11.3 ± 2.0%, P = NS) or class 3 to 4 (−11.3 ± 2.0% vs −13.1 ± 3.3%, P = NS). Clinical characteristics are shown in Table 3 . Multiple regression analyses of in-hospital HF with age, troponin T level, log NT-proBNP, LAVi, and episodes of AF were performed with the addition of LVEF and GLS in two separate models. LVEF was not significant when log NT-proBNP was in the model, whereas GLS remained significant and diminished the association of log NT-proBNP to borderline significance ( Table 4 ). When adding GLS to a model already consisting of age, troponin T, log NT-proBNP, LVEF, LAVi, and episodes of AF, the C-statistic increased significantly (0.82 [95% CI, 0.76–0.89] vs 0.87 [95% CI, 0.82–0.91], P = .02; Figure 4 ).

Jun 2, 2018 | Posted by in CARDIOLOGY | Comments Off on Relationship between Left Ventricular Longitudinal Deformation and Clinical Heart Failure during Admission for Acute Myocardial Infarction: A Two-Dimensional Speckle-Tracking Study

Full access? Get Clinical Tree

Get Clinical Tree app for offline access