We reviewed medical records of 543 patients who were diagnosed with ant-STEMI and underwent PPCI successfully at proximal or middle LAD artery at Chonbuk National University Hospital from January 2008 to December 2016. Two-dimensional echocardiographic study was performed within 48 hours after percutaneous coronary intervention (PCI) in all patients using machines from a specific vendor because GLS values can significantly vary depending on the study time point after AMI3) or on the software from different vendors.4) Patients with prior history of symptomatic HF, prior history of myocardial infarction (MI) or angina, in-hospital death during index hospitalization, final thrombolysis in myocardial infarction (TIMI) flow grade <3 of the culprit lesion at the end of the procedure, or significant (≥moderate) organic valvular dysfunction were excluded. After primary elimination of 298 patients based on these criteria, 74 patients were additionally excluded because their echocardiographic images were not analyzable due to low frame rate or poor quality. A total of 171 patients with eligible echocardiographic data were eventually enrolled to the study (Figure 1). The study protocol was approved by the local ethics committee of the Institutional Review Board of Chonbuk National University Hospital (IRB No. 2018-10-016) and written informed consent was waived because the study was retrospective and anonymous clinical data were used for analysis.

Patients with ant-STEMI-related HF was divided into 2 categories according to the timing of development: in-hospital and post-discharge HF. Group 1 (in-hospital HF) included patients with signs of HF, rales or pulmonary edema (Killip classification II and III) at the time of admission or during hospitalization after PCI. Patients in Group 1 were subdivided into HF at the time of admission (Group 1A) and new HF after PPCI (Group 1B). Among those who did not develop in-hospital HF, patients who experience HF after discharge were assigned to Group 2 (post-discharge HF). Group 2 involved patients who were discharged without signs of HF at the time of discharge and visited emergency room (ER) or were re-hospitalized due to dyspnea with signs of HF and elevated level of NT-proBNP (≥150 pg/mL for patients aged ≤74 years, ≥450 pg/mL for patients aged ≥75 years) at any time after discharge. Group 3 included patients who did not experience HF during index hospitalization or after discharge. The verification of HF development or hospitalization was performed by an independent reviewer blinded to echocardiographic information.

Patient demographic data, atherosclerotic risk factors, procedural characteristics, and laboratory data were collected based on the medical records. Patients who were lost during follow-up were called to confirm the occurrence of clinical events. Chest pain-to-ER visit time was defined as the time from onset of chest pain to the contact to our ER. Successful PCI was defined as residual diameter stenosis <30% and TIMI flow grade 3 of the culprit lesion at the end of the procedure. No reflow phenomenon was defined as a transient or persistent TIMI flow grade 0 or 1 after relief of coronary occlusion without angiographic evidence of mechanical obstruction.

Two-dimensional echocardiography was performed using a Vivid 7 (GE Vingmed Ultrasound, Horten, Norway) or Vivid q (GE Medical Systems, Haifa, Israel) echocardiography machine with an M4S probe in all study patients. LV dimensions, volumes, and ejection fraction were measured according to the recommended guidelines from American Society of Echocardiography and European Association of Echocardiography.5) Wall motion score index (WMSI) of the LV was analyzed using a 17-segment model. Indexing of modified Simpson formula-derived LV volumes was performed to body surface area for left ventricular end-diastolic volume index (LVEDVI) and left ventricular end-systolic volume index (LVESVI). Mitral regurgitation (MR) was evaluated by distal jet area using color Doppler study and effective regurgitant orifice area using the proximal isovelocity surface area method and MR ≥grade 2 was considered significant. Right ventricular systolic pressure (RVSP) was calculated based on the peak velocity of tricuspid regurgitation and respiratory collapsibility of the inferior vena cava.

Measurement of GLS was performed as an offline analysis of digitally stored cineloop images at a remote workstation (EchoPAC BT 201.54.0; GE Vingmed Ultrasound, Horten, Norway) using a semiautomated software (Automated Function Imaging; AFI) by 2 experienced operators (S. H. Lee and W. H. Kim) who were blinded to patients' clinical information and the mean of the 2 measured values was used. For measurement of GLS, three apical images were selected and a region of interest (ROI) was automatically tracked for each of the 3 images by the software. Position of the ROI was manually adjusted if automatic tracking was not satisfactory. The software generated longitudinal strain curves for each of 17 segments, a color-coded bull's eye plot image, and a calculated value of GLS from the average of peak longitudinal strain values of 17 segments (Figure 2). For measurement of GCS, a mid-LV level image at parasternal short axis view using the papillary muscles as an anatomical landmark was obtained. A value of GCS was automatically calculated by the software from the average of peak circumferential strain values of 6 segments. In case of unsatisfactory tracking even after adjustment of the trace line, the segments were excluded from the analysis. If GLS was not available because of insufficient tracking of speckles in more than three segments, the patient was excepted from the study. The LS_ant was defined as the mean peak systolic longitudinal strain of the 11 segments supplied by LAD and calculated by the operator using the bull's eye plot image (Figure 3).

Continuous variables were presented as mean ± standard deviation and categorical variables as percentages (%). The Chi-square test was used for between-group comparison of categorical variables. A Kruskal-Wallis test and Mann-Whitney test with Bonferroni correction for post hoc analysis were used for continuous variables. Laboratory variables such as high-sensitivity C-reactive protein (hs-CRP) and NT-proBNP were incorporated into the model as a continuous variable. Absolute values were used for statistical analyses of strain values including GLS, GCS, and LS_ant. Univariate and stepwise backward multivariate logistic regression analyses were separately used to determine the factors associated with the development of in-hospital and post-discharge HF. All variables significant on univariate analyses with a p value <0.05 in group-to-group comparison were submitted for multivariate analysis. The PASW Statistics 18.0 for Microsoft Windows^® (SPSS Inc. Chicago, IL, USA) was used for statistical analyses. In all cases, p values <0.05 were considered statistically significant. The predictive power and the best cut-off values of GLS and LS_ant for in-hospital and post-discharge HF were respectively determined in comparison to LVEF and level of NT-proBNP on receiver operating characteristic (ROC) curve analysis using the MedCalc for Windows, version 18.11.3 (MedCalc Software, Ostend, Belgium).

Mean age of overall patients was 62±13 years and 72.5% were male. Median follow-up period was 31.6 (mean, 98.9±76.6) months. In-hospital HF developed in 39 patients (22.8%). Among 132 patients who did not have HF during index hospitalization, 13 patients (9.8% [13/132], 7.6% of overall patients [13/171]) experienced HF during follow-up after discharge. In the total of 171 patients, the segmental feasibility of GLS measurement was 95.0%. Inter-observer and intra-observer variability for overall strain measurement was 0.88 and 0.92, respectively.

Baseline demographic characteristics were comparable except age and history of smoking. Mean age of patients in Group 2 was significantly higher than those in other two groups (65±13 vs. 69±12 vs. 60±13 years, p=0.029). Smokers were more prevalent in Group 3 than in Groups 1 and 2 (38.5% vs. 38.5% vs. 59.7%, p=0.038). The levels of NT-proBNP and hs-CRP were significantly higher in Group 2 and in Group 1 than the other 2 groups, respectively (Table 1). Peak troponin I levels were comparable in the 3 groups.

Echocardiography for strain analysis was 1.5±1.1 days after PPCI. The strain values showed significantly different values in the three groups. GLS (−10.63±2.93% vs. −12.11±2.24% vs. −13.51±3.30%, p<0.001) and GCS (−8.61±3.55% vs. −11.59±2.89% vs. −12.11±3.68%, p=0.004) were more prevalently impaired in Group 1. LS_ant was depressed in both HF groups, Group 1 and Group 2, compared to non-HF group (−6.22±4.69% vs. −5.49±5.17% vs. −9.48±5.44%, p=0.001).

In addition to strain parameters, patients in Group 1 and Group 2 showed poorer LVEF, higher WMSI, LVESVI, and E/E′ ratio, more significant MR, more elevated RVSP, and more extensive coronary artery disease than those of Group 3 (Table 1). Chest pain-to-ER visit time, door-to-balloon time, and frequencies of initial TIMI flow grade 0 and no reflow phenomenon during the procedure were comparable.

Among 39 patients in Group 1, 56.4% (n=22) presented with symptoms or signs of HF at the time of admission (Group 1A) and the remained 43.6% (n=17) developed HF after admission (Group 1B). Patients in these two subgroups were indistinguishable in demographic, laboratory, procedural, and echocardiographic characteristics (see the Supplementary Table 1). The development of post-discharge HF was widely distributed from 1 month to 8 years after ant-STEMI. Mean level of NT-proBNP at the diagnosis of post-discharge HF was 5,007.6±4,835.6 pg/mL.

Table 2 demonstrates the result of post hoc analysis for group-to-group comparison. Levels of NT-proBNP and hs-CRP, traditional echocardiographic parameters including LVEF, WMSI, LVESVI, E/E′ ratio, RVSP, and significant MR, history of smoking and number of diseased vessels showed significant differences between in-hospital-HF and no-HF groups. Patients with post-discharge HF were different in age, levels of NT-proBNP and hs-CRP, E/E′ ratio, and RVSP compared to those of no-HF group.

As regards strain values, GLS and GCS were significantly depressed in Group 1 compared to Group 3. LS_ant was the only strain parameter that showed discrepancy between Group 2 and Group 3. GLS and GCS values were impaired in both HF groups, Group 1 and Group 2, compared to Group 3 but they showed far more depression in Group 1 than in Group 2 (Table 2).

In multivariate analysis to identify predictors of in-hospital HF using GLS, LS_ant, GCS, LVEF, WMSI, LVESVI, significant MR, RVSP, number of diseased vessels, and history of smoking, all of which were proved to be of significance in univariate analysis, GLS was the only and the strongest predictor remained statistically significant (odds ratio [OR], 1.853; 95% confidence interval [CI], 1.207–2.846; p=0.005), and RVSP showed borderline significance (Table 3).

For prediction of post-discharge HF, multivariate analysis including age, LS_ant and E/E′ ratio showed that and LS_ant was the only independent factor associated with post-discharge HF (OR, 1.143; 95% CI, 1.006–1.298; p=0.040).

ROC curve analysis to predict in-hospital HF using the area under the curve (AUC) displayed the AUCs for GLS of 0.753 (95% CI, 0. 682–0.816; p<0.001), for LVEF of 0.735 (95% CI, 0.662–0.800; p<0.001) and for NT-proBNP of 0.600 (95% CI, 0.516–0.679; p=0.072) (Figure 4A). The predictive power of GLS for in-hospital HF was significantly superior than that of NT-proBNP (p=0.012) and comparable with LVEF (p=0.532). The best cut-off value of GLS was −10.9% which achieved a sensitivity of 69.2% and a specificity of 76.5% to predict in-hospital HF. For prediction of post-discharge HF, LS_ant (AUC, 0.715; 95% CI, 0.630–0.790; p=0.010) and NT-proBNP (AUC, 0.720; 95% CI, 0.628–0.801; p=0.005) were comparably fair (p=0.585), but LVEF was not discriminatory (AUC, 0.643; 95% CI, 0.555–0.725; p=0.055) (Figure 4B). The LS_ant cut-off value of −2.36% achieved a sensitivity of 53.9% and a specificity of 89.1% in prediction of post-discharge HF.