Global longitudinal strain is a more reproducible measure of left ventricular function than ejection fraction regardless of echocardiographic training

We invited 126 surviving patients from a previous study admitted with suspected non-ST elevation acute coronary syndrome (NSTE-ACS) [5, 14] to a five-year follow-up study with echocardiographic examination. Eleven patients declined further participation and 10 did not respond. Patients from this cohort were prospectively included to undergo a double echocardiographic examination by a trainee and an expert examiner, and 47 patients constitute the basis of the present study. The study was approved by the regional ethical committee.

To standardize and increase quality of transthoracic echocardiographic examinations (TTE) both the European Association of Echocardiography (EAE) and American College of Cardiology (ACC)/American Heart Association (AHA) have published guidelines regarding echocardiographic and clinical competence [15, 16]. The defined expert in this study was a physician qualified as advanced or level 3-examiner with special interest in speckle tracking echocardiography. This implied a minimum of 12 months of training, 300 performed examinations and 750 interpreted examinations. The trainee was a physician qualified to basic or level 2 which required a minimum of 6 months of training, 150 examinations and 300 interpreted examinations.

Both examiners obtained echocardiographic examinations using a Vivid 7 Scanner (GE Ultrasound, Horten, Norway) with images and cineloops stored digitally. Both sets of echocardiographic examination were performed during the same consultation and the examiners were blinded for each other’s recordings and findings when examining the patients. Three consecutive cycles from three apical and three parasternal image planes were recorded using 2-dimensional gray scale echocardiography. Frame rates were between 55 and 95 frames/second.

Both examiners used the same commercially available software (EchoPAC version 112, GE Ultrasound) for post hoc analysis of echocardiographic recordings on a separate work station. The examiners were blinded for clinical data and echocardiographic findings before analyzing both sets of images. Closure of the aortic valve defined the end of systole and was determined by pulsed Doppler flow in the left ventricular outflow tract (LVOT). Peak systolic strain was defined as the maximum value of peak negative strain (myocardial shortening) or peak positive strain (myocardial lengthening) during systole. GLS by speckle tracking echocardiography was measured manually in a 18-segments LV model as the average segmental value based on three apical imaging planes. LVEF was calculated using Simpson’s biplane method. Image quality for both expert and trainee images was evaluated by the trainee echocardiographer. Image quality was rated poor when < 60% of endocardial border was visible in any standard apical image plane, fair when 60–74% was visualized and good when > 75% of the endocardium was visible. Left ventricular end diastolic diameter was measured perpendicular to the left ventricle in parasternal long axis image measuring the distance from the septal endocardium to the endocardium of posterior wall in the end diastole at the level of the tips of the mitral valve. E wave express early diastolic mitral inflow velocity measured by pulsed doppler. E’ represent early diastolic mitral annular velocity.

Continuous data are presented as mean ± SD or median (inter quartile range). Categorical data are presented as numbers (percentage). All statistical analyses were performed using SPSS 22.0 (SPSS Inc., Chicago, Illinois). Intra class correlation coefficients (ICC) were obtained using two-way mixed model with measures of absolute agreement to describe test reproducibility. ICC of LVEF and GLS were compared using Z-scores. Bland-Altman plots were created to demonstrate inter-observer agreement and were performed by Graphpad Prism ver. 6.05. Linear regression analysis of difference vs mean was used to identify proportional interrater bias. Fixed bias indicates a systemically difference in results between observers, expressed as mean difference analyzed by paired samples T-test.

The clinical data of the 47 patients are summarized in Table 1.

GLS and LVEF data in all assessment scenarios are displayed in Table 2. None of the image recordings for LVEF calculation or GLS measurement were rejected due to image quality. Single LV segments were excluded due to suboptimal image quality or poor tracking when performing GLS measurements. More than 95% of all segments were included in the analyses and there was no significant difference in feasibility between the analyses or examiners. Analyzing trainee images, trainee analyzed 17.3 (±1.2) segments versus expert 17.1 (±1.5), p = 0.45. In expert images the trainee analyzed 17.4 (±1.0) versus expert 17.6 (±1.1), p = 0.38. No images were rejected for LVEF calculation. Results from evaluation of image quality for both echocardiographers are reported in Table 3.

Reproducibility was excellent in general for GLS regardless of level of training in both image acquisition and analysis. It was weakest when trainee and expert compared results after analyzing their own images with an ICC of 0.89 (0.74–0.95) (Table 2, scenario C and Fig. 3). The strongest reproducibility was seen when both analyzed the expert images where ICC was 0.94 (0.84–0.97) (Table 2, scenario A and Fig. 1).

Reproducibility was good for LVEF and was best when both examiners analyzed trainee images with an ICC of 0.76 (0.57–0.87) (Table 2, scenario B and Fig. 2). When the trainee analyzed both set of images the ICC was only 0.13 (− 0.45–0.49) (Table 2, scenario D and Fig. 4).

There was a significant difference in reproducibility between GLS and LVEF in favor of GLS as listed in Table 2 in all echocardiographer-analyst scenarios.

Furthermore, there was no proportional bias for GLS or LVEF, which indicate that the measurements agreed equally through the entire range of measurements. Fixed bias was present in all scenarios indicating a systematic difference in results of both GLS and LVEF (Fig. 1, 2, 3, 4, 5). There was a significant difference in fixed bias between GLS and LVEF when both examiners analyzed trainee images (Table 2, scenario B and Fig. 2) and when both examiners analyzed their own images (Table 2, scenario C and Fig. 3). In both scenarios, the fixed bias was higher in LVEF analysis.

When trainee comparing left ventricular end diastolic diameter in expert and trainee images ICC was 0.91 (0.85–0.95) with a systematic bias of − 0.43 mm. ICC of E/e’ was 0.92 (0.85–0.95) with a systematic bias of 0.25 (Fig. 6).