Population and Study Design

We prospectively studied patients with a wide range of GLS (age 67 ± 16 years, 58% men, body surface area 2.0 ± 0.26; 73% cardiomyopathy or heart failure, 43% systemic hypertension, 23% pulmonary arterial hypertension, 27% coronary artery disease, 7% simple congenital heart disease, and 7% end-stage renal disease). The study was approved by the Institutional Review Board, and informed consent was obtained from each patient.

Protocol 1 included 44 patients referred for clinically indicated transthoracic echocardiography who qualified for contrast echocardiographic studies (two nonvisualized contiguous segments in the apical four-chamber view), in whom noncontrast STE was still possible, despite suboptimal image quality. In this protocol, GLS was measured in the contrast-enhanced as well as nonenhanced images, and the latter measurements were used as the reference for comparisons. Protocol 2 included 20 patients referred to CMR for clinical reasons who agreed to have an echocardiogram, which was performed immediately after the CMR study. These patients had poor quality of noncontrast echocardiographic images (defined as poor endocardial visualization in two or more contiguous segments in the apical four-chamber view). In this protocol, GLS was measured in the contrast-enhanced echocardiographic images and compared against CMR feature tracking-derived strain. Exclusion criteria were complex congenital heart disease, arrhythmia during acquisition, pacemaker, or defibrillator leads (in protocol 2 only).

Echocardiographic Imaging and Analysis

Transthoracic imaging was performed in the apical four-chamber view with the patient in the left lateral decubitus position (IE33 or EPIQ systems, Philips Healthcare, Andover, MA) with an X5-1 transducer. Before each acquisition, images were optimized for endocardial visualization by adjusting the gain, compress, and time-gain compensation controls.

Acquisition settings for echocardiographic contrast-enhanced imaging were optimized in a series of preliminary tests and included (1) approximately half of the manufacturer-recommended dose of a commercial contrast agent randomly assigned to each patient (Definity by Lantheus [North Billerica, MA], Optison by GE Healthcare [Marlborough, MA], and Lumason by Bracco [Monroe Township, NJ]; 1-2 mL diluted in 3 mL saline) and used to provide partial contrast enhancement with lower bubble density than that typically used for LV opacification, resulting in a certain degree of visible swirling; (2) higher than usual mechanical indices (0.6-0.7); (3) focus set at the level of the mitral valve annulus to facilitate accurate tracking of the speckles in the far field; and (4) lowest frequency range for maximal penetration. Every effort was made to avoid throughout the cardiac cycle the inclusion of any portion of the LV outflow tract (LVOT) in the image sector.

Images were stored digitally and used for offline analysis, with the readers blinded to all prior measurements. Both contrast-enhanced and nonenhanced echocardiographic images were analyzed using STE software to measure GLS (EchoInsight, Epsilon Imaging, Ann Arbor MI), which allows myocardial deformation measurements with contrast enhancement from images acquired using the above scheme and tracking within an approximately 5 mm wide region of interest, which is thinner than the default. LV boundaries were initialized in a midsystolic frame and then automatically tracked throughout the cardiac cycle. In both techniques, manual corrections were performed as needed to optimize boundary tracking throughout the cardiac cycle ( Figure 1 ).

Echocardiographic LV longitudinal strain analysis. (A) An example of a bad-quality nonenhanced apical four-chamber image precluding accurate speckle-tracking. (B-E) Images obtained in the same patient during contrast enhancement (B) , with tracings in end diastole (C) and end systole (D) , with the derived strain curves throughout the cardiac cycle (E) . Solid curves of different colors represent six myocardial segments, while the dashed red curve represents GLS measured in the four-chamber view.

CMR Imaging and Analysis

CMR imaging was performed on a 1.5 T scanner (Philips, Best, Netherlands) with a five-channel cardiac coil. Steady-state free-precision dynamic gradient-echo sequence was used to obtain cine loops, during approximately 5-second breath holds (repetition time 2.9 ms, echo time 1.5 ms, flip angle 60°, and temporal resolution ∼30-40 ms). Images were analyzed offline using commercial software (2D CPA MR, a module of TomTec-Arena, TomTec Imaging Systems, Unterschleissheim, Germany). Similar to echocardiographic GLS analysis, manual tracing in a midsystolic frame allowed feature tracking initialization, performed by an investigator trained in CMR-based chamber quantification (SCMR level III training) who was blinded to echocardiographic data ( Figure 2 ).

Global LV longitudinal strain analysis by CMR feature tracking. Example of a four-chamber view in the same patient in Figure 1 showing feature tracking analysis with endocardial tracings (A) and GLS curve throughout the cardiac cycle (B) .

Statistics

Echocardiographic GLS measurements with contrast enhancement were compared with those without contrast enhancement (protocol 1) and with CMR feature tracking-derived GLS (protocol 2). Comparisons included linear regression with Pearson correlation coefficients. These comparisons also included paired t- tests, and values of P < .05 were considered significant. In addition, Bland-Altman analysis was performed to assess the bias and limits of agreement between the echocardiographic GLS measurement with contrast enhancement and the corresponding reference in each protocol.

Reproducibility Assessment

The reproducibility of echocardiographic GLS measurements with and without contrast enhancement was tested using repeated measurements in 30 patients randomly selected from the group enrolled in protocol 1 and in all of the 20 patients enrolled in protocol 2. These repeated measurements were performed on the same image loops by investigators blinded to all prior measurements. To determine the intraobserver variability, images were reanalyzed one month later by the same investigator, while interobserver variability was assessed by comparing these measurements with those performed by a second independent reader. Both intra- and interobserver variability were quantified by calculating intraclass correlation coefficients (ICCs), as well as the absolute difference between the corresponding pair of repeated measurements as a percentage of their mean. The latter was calculated for each patient and then averaged over the entire group.