Validation of Global Longitudinal Strain and Strain Rate as Reliable Markers of Right Ventricular Dysfunction: Comparison with Cardiac Magnetic Resonance and Outcome

Patient characteristics

We retrospectively screened 72 consecutive patients with ICM who underwent conventional and strain echocardiography and CMR from September 2009 to October 2011. Inclusion criteria of ICM patients were 1) dyspnea on exertion (New York Heart Association functional class ≥ II), 2) impaired LV function (LV ejection fraction < 40%), and 3) significant epicardial coronary artery stenosis (≥ 50%) by the conventional coronary angiography. All patients had a history of previous myocardial infarction (MI). Exclusion criteria included age > 80 years, severe tricuspid regurgitation (TR), atrial fibrillation, RV pacing, or poor image quality of RV echocardiographic images without inclusion of whole RV free wall. The study was approved by the Institutional Review Board of Cleveland Clinic.

Echocardiography

Standard echocardiographic examinations with Doppler studies were performed using commercially available echocardiographic systems (Vivid 7, GE Vingmed, Horten, Norway or ie33, Philips, Andover, MA, USA). The echocardiographic images of all subjects were obtained from the parasternal and apical views. Studies were stored digitally and analyzed off-line. RV fractional area change (RVFAC) was calculated from the apical 4-chamber view using the percentage change in areas of the end-diastolic and end-systolic areas of the RV.11) Tricuspid annular plane systolic excursion (TAPSE) was measured the distance of systolic movement of RV annular segment along its longitudinal plane.11) RV myocardial performance (Tei) index was defined as the ratio of isovolumic relaxation time (IVRT) and isovolumic contraction time (IVCT) divided by ejection time (ET) of RV, or (IVRT + IVCT) / ET.11)

Pulmonary artery systolic pressure was estimated from the maximal continuous-wave Doppler velocity of the TR jet plus estimated central venous pressure.11) An index of pulmonary vascular resistance (PVR) was derived by dividing the maximal velocity of the TR jet by the RV outflow tract velocity-time integral.12) An average of 3 measurements was used.

RVGLS was measured off-line with velocity vector imaging (VVI; Axius, Siemens Medical Solutions, Mountain View, CA, USA). The B-mode images of apical 4-chamber views were used for strain measurement. They were obtained and digitally stored in cineloop format with about 60 frames/sec for the offline analysis. After manual tracing of the endocardial border of RV over one frame, then endocardial borders were automatically tracked throughout the cardiac cycle (Fig. 1). Myocardial velocity is derived as the ratio between frame-to-frame displacement of the speckles and the time interval.10) RVGLS was obtained and averaged for all 6 segments of RV free wall and septum. Global longitudinal strain of RV free wall was measured for 3 RV segments (basal, mid, and apex) and ventricular septum for 3 septal segments. The same apical 4-chamber images used for conventional RV measurements were analyzed with VVI by an investigator who was blinded to the conventional echocardiography and CMR data.

Fig. 1
Representative figure for the measurement of right ventricular longitudinal strain with velocity vector imaging software.

• Click for larger image

CMR studies

CMR examinations in the patient with sinus rhythm were performed on standard 1.5 T MR scanners (Sonata or Avanto, Siemens Medical Solutions, Erlangen, Germany) with electrocardiographic gating. The time difference between echocardiography and CMR acquisition was 6 ± 10 days. Scout images were acquired initially to identify the cardiac axes. Subsequently, we assessed global RV ejection fraction (RVEF) using multiple long- and short-axis views. RV dysfunction determined by CMR was defined as RVEF < 50%.

Subsequently, delayed hyperenhancement images were obtained to identify scar lesion of the RV in the same long- and short-axis orientations as the above-described images approximately 20 minutes after injection of 0.2 mmol/kg of gadolinium dimeglumine.

Outcomes

Patients were followed over 15 ± 9 months for a composite of death, cardiac hospitalization due to heart failure, and heart transplantation.

Reproducibility

Intraobserver and interobserver variabilities of the RVGLS were evaluated in 15 random subjects by two investigators (J.P. and K.N.) and measured by calculating the intraclass correlation coefficients.

Statistical analysis

The data were analyzed using standard software (SPSS version 18.0, SPSS Inc., Chicago, IL, USA). Summary data were expressed as mean values ± standard deviation or percentage of patients. Linear regression analysis was performed to evaluate the relationship between RVGLS and RVGLSR and other variables. The optimal cutoff value for predicting RV dysfunction defined by CMR RVEF was determined by the receiver-operating characteristic curve analysis. Comparison of areas under the curve (AUC) was done with the method suggested by Hanley and McNeil.13) To assess prognostic variables, we performed a Cox proportional hazards regression with the combined end point of death, hospital admission for heart failure and heart transplantation. To exclude over-fit of the analysis model, we calculated EuroSCORE II in these patients and used it to the analysis.14)

Cumulative event-free survival curves for adverse events were constructed according to the Cox-proportional hazard regression analysis. A p value less than 0.05 was considered statistically significant.

Patient characteristics

Initially, we screened 72 patients with ICM. Of them, 11 patients were excluded for poor echocardiographic image quality and 4 were excluded because of the presence of atrial fibrillation. A total 57 patients (43 males, mean age 64 ± 12 years old) were included in this study. Their baseline clinical, routine echocardiographic and CMR data (Table 1) are typical for an ICM population, with multiple risk factors, LV enlargement and moderate to severe LV dysfunction. RV dysfunction was present in 63% of patients.

Table 1
Baseline characteristics

• Click for larger image

RV strain findings

Global and regional two-dimensional strain measurements are listed in Table 1. RVGLS showed significant correlations with CMR RVEF (r = -0.797, p < 0.001), RVFAC (r = -0.530, p < 0.001), and TAPSE (r = -0.547, p < 0.001) (Fig. 2). RVGLSR showed significant correlations between CMR RVEF (r = -0.668, p < 0.01), RVFAC (r = -0.394, p < 0.01), and TAPSE (r = -0.435, p < 0.01) (Fig. 3A, B, and C). RVGLS and RVGLSR showed significant correlations with PVR (r = 0.527, Fig. 2D and r = 0.500, p < 0.01, Fig. 3D, respectively), but not pulmonary artery systolic pressure.

Fig. 2
Correlations between right ventricular global longitudinal strain (RVGLS) and cardiac magnetic resonance image (CMR) and echocardiographic parameters. RVGLS shows good negative correlation with CMR RV ejection fraction (A), RV fractional area change (RVFAC, B), and tricuspid annular plane systolic excursion (TAPSE, C). Also, RVGLS demonstrates significant correlation with pulmonary vascular resistance (D). WU: wood unit.

• Click for larger image

Fig. 3
Correlations between right ventricular global longitudinal strain rate (RVGLSR) and cardiac magnetic resonance image (CMR) and echocardiographic parameters. RVGLSR shows good negative correlation with CMR RV ejection fraction (A), RV fractional area change (RVFAC, B), and tricuspid annular plane systolic excursion (TAPSE, C). Also, RVGLSR demonstrates significant correlation with pulmonary vascular resistance (D). WU: wood unit.

• Click for larger image

CMR findings

CMR RVEF showed significant correlations with conventional echocardiographic RV parameters including RVFAC (r = 0.500, p < 0.01), RV Tei index (r = -0.446, p < 0.01) and TAPSE (r = 0.360, p < 0.01). There were 36 patients with RV dysfunction determined by CMR (RVEF less than 50%). Patients with RV scar on CMR showed poor RV systolic function, lower strain and strain rate (Table 2). The best cutoff value of RVGLS for detection of RV dysfunction was -15.4% (AUC = 0.955, p < 0.01) with a sensitivity of 81% and specificity 95%. The best cutoff value for RVGLSR was -0.94 (AUC = 0.871, p < 0.01), sensitivity 72%, specificity 86%. RVGLS was more accurate in the detection of RV dysfunction than RVFAC [difference = 0.229, 95% confidential interval (CI): 0.094-0.364, p < 0.01], TAPSE (difference = 0.215, 95% CI: 0.086-0.344, p < 0.01), RV Tei index (difference = 0.252, 95% CI: 0.097-0.406, p < 0.01), and RVGLSR (difference = 0.083, 95% CI: 0.010-0.157, p = 0.03) (Table 3).

Table 2
Comparison of conventional and strain echocardiographic data according to the presence of scar in the right ventricle

• Click for larger image

Table 3
Best cut off value and area under the curve to detect right ventricular dysfunction according to the parameters

• Click for larger image

Follow-up

During the follow-up of 15 ± 9 months, there were 12 adverse events including 5 deaths, and 7 hospitalization for heart failure. After univariate analysis by Cox-proportional hazard regression analysis, PVR, decreased TAPSE (< 1.72 cm), and impaired RVGLS (≥ -15.4%) were associated with adverse clinical outcome. After adjustment with EuroSCORE II, decreased RVGLS and RVGLSR were statistically significant (Table 4). Patients with decreased RVGLS and RVGLSR showed significantly lower 1 year event-free survival rate (Fig. 4).

Fig. 4
EuroSCORE II adjusted event free survival by Cox proportional regression analysis. Patients with impaired right ventricular global longitudinal strain (RVGLS) shows lower 1 year event-free survival (93.0% vs. 67.2%, p = 0.030, A). Also, impaired right ventricular global longitudinal strain rate (RVGLSR) group reveals lower 1 year event-free survival (90.5% vs. 67.4%, p = 0.044, B).

• Click for larger image

Table 4
Unadjusted hazard ratio and EuroSCORE II adjusted hazard ratio in the prediction of adverse clinical events

• Click for larger image

Variability

Interobserver variability of RVGLS was small [intraclass correlation coefficient was 0.909 (95% CI = 0.753-0.968), p < 0.001], and similar to intraobserver [0.904 (95% CI = 0.746-0.966), p < 0.01].

Detection of RV dysfunction

LV systolic function is a well-known predictor of future cardiovascular morbidity and mortality in patients with MI.15) The presence of RV dysfunction is associated with poor clinical outcome in patients with LV dysfunction after an acute MI.4) Moreover, quantitative measurement of RV size and function is important in the prediction of clinical outcomes in a number of pulmonary diseases.11) However, quantitative assessment of RV function is often neglected and the assessment of RV by conventional echocardiography remains difficult because of the complex shape of the chamber.11) Although several two-dimensional echocardiographic measurements of RV function have been proposed, including RV diameter and two-dimensional area.16) RVFAC is a widely used index of RV function, but despite its relatively good correlation with CMR and radionuclide RVEF,16), 17) the feasibility of RVFAC is limited because of the challenge of tracing the entire RV endocardium in end-diastole and end-systole.16), 17), 18) In our study, RVFAC showed significant correlations with RVEF by CMR, but the degree of correlation was moderate (r = 0.500, p < 0.001).

In the present study, CMR was used as reference in the assessment of RV function. CMR has been validated in several previous studies.19), 20) CMR demonstrates an excellent reproducibility for assessment of RV function and RV volumes.6), 20) Despite the excellent image quality and reproducibility, long acquisition time, time-consuming analysis and not suitable for some patients unable to take examination are major disadvantages of CMR.21)

Strain imaging

The application of speckle tracking to assessment of global chamber function is based on the understanding that GLS is the result of the sum of all segmental deformation. The use of speckle tracking for the assessment of regional and global LV function has been validated by sonomicrometry,22), 23) and this method has been shown to provide prognostic information that is incremental to EF.24)

Speckle strain has been introduced for the assessment of the RV in some disease categories,10), 25), 26) but has not previously been validated in patients with ICM. Because the muscle fibers of the RV predominantly run in a longitudinal direction and the major contractility of RV occurs in the longitudinal plane,6), 27) GLS can represent RV function accurately. Moreover, strain has the advantage of distinguishing true contractility of the free wall rather than tethering or translational motion which can influence indices like TAPSE. This may be important not only in ICM, but also in idiopathic dilated cardiomyopathy with longitudinal rotation of the heart (swinging motion of the heart).25), 28) In this study, we found RVGLS of -15.4% and RVGLSR of -0.94 s^-1 to be the optimal cut-point for the detection of RV systolic dysfunction. Using these criteria, the detection of RV systolic dysfunction was more accurate than using other conventional RV echocardiographic parameters including RVFAC, TAPSE, and RV Tei index. Decreased RVGLS (≥ -15.4%) and RVGLSR (≥ -0.94 s^-1) was associated with lower 1 year event free survivals compared with preserved group irrespective of the LV systolic function.

There are several proprietary approaches to the measurement of myocardial strain. VVI is a validated technique for assessment of LV function,10) which has several advantages in the assessment of ventricular function. Like all 2-dimensional strain approaches, its measurement of myocardial contractility is angle-independent.10), 22) It can determine global and regional myocardial function simultaneously and is able to analyze archived digital images.10) VVI seems to be less sensitive than other methods to wall thickness, possibly because of a greater contribution of wall tracking to the strain algorithm. We have previously shown that VVI may be superior to other speckle strain techniques in the assessment of atrial function,29) perhaps reflecting less dependence of tracking on a thick wall.

Like all contraction-phase measurements, RVGLS is significantly influenced by afterload. In this study, this was evidenced by correlation with PVR. Increased PVR, as a result of increased afterload, is associated with increases in RV size and RV dysfunction. Indeed, these findings may be concordant with previously published data showing decreased event-free survival in patients with increased PVR.7)

Limitations

This is an observational study with analysis of relatively small numbers of stored digital images. Bias may have been introduced from patient selection. Analysis may have been affected by image quality-to minimize this potential contribution to error, we excluded the patients with poor image quality and analyzed standard apical four chamber images. Second, we used echocardiographic images of different machines from two vendors in this strain analysis. Therefore, cut-off value of RVGLS and RVGLSR provided in this study cannot be applied to all speckle tracking analysis results in other clinical situations. Nonetheless, a prospective study with large number of patients will be needed to confirm the correlations and the clinical impact of this measurement.

Conclusion

RVGLS and RVGLSR correlate better with CMR RVEF than conventional echocardiographic parameters. Lower RVGLS and RVGLSR are associated with adverse clinical outcome in patients with ICM.