Longitudinal comparison of dyssynchrony correction and ‘strain’ improvement by conduction system pacing: LEVEL-AT trial secondary findings

Abstract

Aims

Longitudinal dyssynchrony correction and ‘strain’ improvement by comparable cardiac resynchronization therapy (CRT) techniques is unreported. Our purpose was to compare echocardiographic dyssynchrony correction and ‘strain’ improvement by conduction system pacing (CSP) vs. biventricular pacing (BiVP) as a marker of contractility improvement during 1-year follow-up.

Methods and results

A treatment-received analysis was performed in patients included in the LEVEL-AT trial (NCT04054895), randomized to CSP or BiVP, and evaluated at baseline (ON and OFF programming) and at 6 and 12 months (n = 69, 32% women). Analysis included intraventricular (septal flash), interventricular (difference between left and right ventricular outflow times), and atrioventricular (diastolic filling time) dyssynchrony and ‘strain’ parameters [septal rebound, global longitudinal ‘strain’ (GLS), LBBB pattern, and mechanical dispersion). Baseline left ventricular ejection fraction (LVEF) was 27.5 ± 7%, and LV end-systolic volume (LVESV) was 138 ± 77 mL, without differences between groups. Longitudinal analysis showed LVEF and LVESV improvement (P < 0.001), without between-group differences. At 12-month follow-up, adjusted mean LVEF was 46% with CSP (95% CI 42.2 and 49.3%) vs. 43% with BiVP (95% CI 39.6 and 45.8%), (P = 0.31), and LVESV was 80 mL (95% CI 55.3 and 104.5 mL) vs. 100 mL (95% CI 78.7 and 121.6 mL), respectively (P = 0.66). Longitudinal analysis showed a significant improvement of all dyssynchrony parameters and GLS over time (P < 0.001), without differences between groups. Baseline GLS significantly correlated with LVEF and LVESV at 12-month follow-up.

Conclusion

CSP and BiVP provided similar dyssynchrony and ‘strain’ correction over time. Baseline global longitudinal ‘strain’ predicted ventricular remodelling at 12-month follow-up.

Graphical Abstract

Graphical Abstract.

Conduction system pacing (CSP) and biventricular pacing (BiVP) provided a similar dyssynchrony and ‘strain’ correction over time in patients who successfully received the device (treatment-received analysis). LVESV, left ventricular end-systolic volume.

Introduction

Conduction system pacing (CSP)—with His bundle pacing (HBP) or left bundle branch pacing (LBBP)—has emerged as an alternative^1,2 to biventricular pacing (BiVP) in patients with cardiac resynchronization therapy (CRT) indication. Despite similar results to BiVP, CSP has been evaluated in a limited number of studies,^3–5 a fraction of them randomized.^6–10 The LEVEL-AT⁸ trial—‘Left Ventricular Activation Time Shortening with Conduction System Pacing vs Biventricular Resynchronization Therapy’ (NCT04054895)—has shown similar degrees of cardiac electrical resynchronization with CSP and BiVP, defined as decreased LV activation time (LVAT) as measured with electrocardiographic imaging (ECGI).

Since the beginning of CRT, several parameters for dyssynchrony analysis at different levels have been proposed and can improve our understanding of the pathophysiology of mechanical dyssynchrony and regional contraction abnormalities secondary to an electrical conduction disorder.¹¹ These parameters have shown predictive and prognostic capabilities in numerous single-centre studies,^12–14 but conflicting results have been reported in multicentre trials.¹⁵

The aim of this study was to test our hypothesis that CSP and BiVP provide similar dyssynchrony and ‘strain’ correction over time, as well as similar improvement in LV function. The main goals were to compare three data points in the two treatment groups during the 12-month follow-up: (i) change in atrioventricular (AV), intraventricular, and interventricular (VV) dyssynchrony, measured with classical echocardiographic parameters and ‘strain’ techniques with speckle tracking; (ii) reverse ventricular remodelling through changes in LV ejection fraction (LVEF) and LV end-systolic volume (LVESV); and (iii) increase in global longitudinal ‘strain’ (GLS).

Methods

Study design and population

Treatment-received analysis was performed in patients included in the LEVEL-AT randomized clinical trial [n = 69 (CSP n = 29; BiVP n = 40), 32% women]. One patient in the LEVEL-AT trial (n = 70) was excluded because he did not receive pacing due to implant failure of both leads. Follow-up was performed for 1 year (at 6 and 12 months; (Figure 1). LEVEL-AT inclusion criteria⁸ were age ≥18 years and symptomatic patients with heart failure on optimal medical treatment with LVEF ≤ 35% and wide QRS (LBBB ≥ 130 ms or QRS ≥ 150 ms in non-LBBB) or an indication for CRT due to AV block and cardiac dysfunction (LVEF ≤ 40%). Exclusion criteria were myocardial infarction, unstable angina, or cardiac revascularization within 3 months before assessment, as well as pregnancy or participation in other clinical trials.

Figure 1.

Study flowchart. BiVP, biventricular pacing; CSP, conduction system pacing.

The LEVEL-AT protocol was approved by the Hospital Clinical Research Ethics Committee (Reg.HCB/2019/0576) and complied with the Declaration of Helsinki. All patients provided written informed consent.

Implant procedure and optimization

Implant procedure and optimization have been previously described.⁸ Optimization in the BiVP cohort was performed with the fusion-optimized interval (FOI) method as previously described and according to the usual practice in our centre.¹⁶ Briefly, FOI is defined as the AV and VV intervals that obtain the narrowest QRS complex. The FOI method has shown greater LV remodelling compared with nominal settings.¹⁷ Neither automated optimization algorithms nor multipoint pacing was used.

In patients with LBBP, the AV interval was optimized with electrocardiography whenever intrinsic conduction was preserved. This aimed to shorten the QRS interval by allowing fusion with intrinsic right bundle ventricular activation, adjusting the AV interval that resulted in the shortest QRS.⁸

Echocardiographic measurements

Patients were evaluated with echocardiography (Vivid E95, GE-Vingmed, Boston, MA) at baseline, 15 days post-implantation (ON and OFF programming), and 6 and 12 months. All echocardiograms were completed by operators blinded to the device programming (R.J.-A. or C.G.-R.) and supervised by a third blinded echocardiographer (A.D.) who reviewed all the studies. A minimum of three cardiac cycles were acquired in each imaging plane. Frame rate of images obtained for speckle tracking analysis was optimized to range between 50 and 80 Hz. Special attention was given to obtaining a recording of optimal research quality. Analysis included AV (diastolic filling time), intraventricular [septal flash (SF)], and VV (difference between left and right ventricular outflow times) dyssynchrony, as well as ‘strain’ parameters (septal rebound stretch, GLS, LBBB pattern, and mechanical dispersion) assessed with speckle tracking.

LV volumes and LVEF were calculated by the Simpson rule from two- and four-chamber apical views. ‘Strain’ myocardial deformation of the left ventricle was quantified offline from 2D echocardiography using speckle tracking (2D-strain, EchoPac, version 202.41.0, GE Healthcare Milwaukee, WI). For the analysis, the region of interest (ROI) along the endocardial border was manually traced on a single 2D echocardiography frame in end-diastole; the software automatically tracked all ventricular segments throughout the cardiac cycle. In cases where tracking was deemed incorrect, the ROI was adjusted manually. Segments that were poorly visualized or not tracked correctly despite manual correction were excluded from analysis. Studies with two or more non-assessable segments were excluded. The long-axis cine images (two-, three-, and four-chamber views) were used to determine GLS.

The dyssynchrony parameters evaluated were as follows (Figure 2):

Figure 2.

Methodology: echocardiographic dyssynchrony assessment. Analysis of cardiac dyssynchrony using standard techniques and techniques derived from myocardial deformation analysis. (A) AV dyssynchrony assessment based on the relationship between transmitral filling time (b) and the R–R interval (a). (B) Measurement of VV dyssynchrony based on the difference between pulmonary and aortic pre-ejection times, defined as the time from the beginning of the QRS complex to the onset of ejection determined from pulsed Doppler register of the outflow tracts. (C) Quantification of SF as a marker of left intraventricular dyssynchrony using an M-mode register passing through the VV septum, measuring the distance between the maximum protosystolic excursion of the septum and its baseline position. (D) Analysis of septal rebound stretch (arrow) as a marker of intraventricular dyssynchrony, defined as the presence of myocardial stretch prior to aortic valve closure (positive deflection, marked as *) preceded by shortening (negative deflection) in septal segments. (E) Longitudinal ‘strain’ curves showing a typical pattern of LBBB, with early shortening followed by stretching of a septal segment (arrow) and early stretching with a post-systolic peak of contraction in a lateral segment (arrowheads).

1. AV dyssynchrony, assessed with LV filling time. The LV filling time was measured from the onset of the E-wave to the end of the A-wave, and the R–R interval was measured to calculate the percentage of filling time relative to the cardiac cycle (LV filling time/RR, %).¹²

2. Intraventricular dyssynchrony, assessed with SF,^12,18,19 a quick inward–outward movement of the VV septum in early systole. Using M-mode in parasternal short- and long-axis views, SF was quantified (in millimetres) as the highest amplitude of the early inward motion measured from the resting position prior to the onset of septal contraction. The pair of SF measures (baseline and final) was obtained at the axis with the highest baseline SF.

Moreover, intraventricular dyssynchrony was qualitatively assessed by three ‘strain’ analysis methods:

1. Septal rebound stretch, evaluating the presence of septal systolic stretching after initial shortening of the septum.²⁰

2. LBBB contraction as described previously by Risum and colleagues.^13,21 All three of the following criteria were required to be read as a typical LBBB pattern from longitudinal ‘strain’ curves in the four-chamber view: (1) early shortening of at least one basal or mid-ventricular segment in the septal wall and early stretching in at least one basal or mid-ventricular segment in the lateral wall; (2) early septal peak shortening (within the first 70% of the ejection phase); and (3) lateral wall peak shortening after aortic valve closure. If any one of these criteria was unmet, the patient was categorized as having an atypical pattern.

3. Mechanical dispersion²² was calculated as the standard deviation of the segmental time intervals between the onset of the R-wave of the ECG and the peak of longitudinal ‘strain’ (including post-systolic peaks) and was expressed in milliseconds. A 17-segment model was used (see Supplementary data online, Figure A).

• c) VV dyssynchrony (in milliseconds), quantified using pulsed Doppler and calculated as the time difference between QRS onset and the onset of the flow wave in the right and left outflow tracts.

Longitudinal analysis of LVEF and LVESV through a 12-month follow-up was prespecified as a secondary endpoint of the LEVEL-AT trial. Echocardiographic response was defined as >5%^5,10 improvement in LVEF at 12-month follow-up. Echocardiographic dyssynchrony analysis (AV, intraventricular, and VV) and GLS were predefined in the study protocol. Correction of dyssynchrony by ‘strain’ analysis was retrospectively reviewed.

Statistical analysis

Baseline continuous variables were presented as the mean value ± SD, and Student’s t-test was used to compare means between groups. Categorical variables were expressed as total number with percentages and compared between groups using Pearson χ² test.

We analysed trajectories between and within groups in the longitudinal follow-up using generalized multilevel mixed effect for repeated measures with group per time as longitudinal interaction fixed effects. In all models, participant identification was treated as a random effect to model the within-participant correlations. Missing data for the overall follow-up was 5.6% and can be considered at random. To fit the models, we used Lme4 R package (v.1.1.30). Residual plots were used to perform model validation. Statistical comparisons and marginal means were computed with emmeans library (v. 1.8.1-1).

To assess the performance of each dyssynchrony parameter as predictor for response (increase in LVEF > 5%), we used ROC analysis, and sensitivity, specificity, and positive and negative predictive values were computed. The optimal threshold cut-off was chosen with Youden criteria. ORs for diagnoses were reported.

To assess the rank of agreement between observers, we selected a random sample of 30 patients. Lin’s concordance coefficient and Bland–Altman plot were used for continuous outcomes and Cohen’s kappa for the categorical ones.

All analyses were addressed considering a two-tailed type 1 error of 5%. Statistical analysis was performed using R version 4.2.1 for Windows (R Project for Statistical Computing).

Results

Sixty-nine patients were included in the study (CSP n = 29; BiVP n = 40; Figure 1). Baseline characteristics of the patients are shown in Table 1, without differences between groups: 32% (22/69) were women; baseline LVEF was 27.6 ± 7% and LVESV 138 ± 77 mL; and 13.8% (4/29) patients received HBP and 86.2% (25/29) LBBP.

Table 1.

Baseline characteristics

	CSP n = 29	BiVP n = 40	P
Female, % (n)	38% (11)	28% (11)	0.36
Age, years	65.8 ± 9.5	67.9 ± 8.7	0.37
Ischaemic cardiomyopathy, % (n)	34% (10)	30% (12)	0.69
Glomerular filtration <45 mL/min, % (n)	17.2% (5)	20% (8)	0.77
Hypertension, % (n)	58.6% (17)	70% (28)	0.33
Diabetes, % (n)	24% (7)	42% (17)	0.11
BMI, kg/m2	27.5	27.7	0.84
QRS width, ms	173 ± 22	181 ± 21	0.12
LBBB, % (n)	65.5% (19)	60% (24)	0.64
Nonspecific intraventricular delay, % (n)	10.3% (3)	12.5% (5)	0.78
AV block (no upgrade), % (n)	6.9% (2)	10% (4)	0.20
Upgrades, % (n)	17.2% (5)	17.5% (7)	0.98
Permanent atrial fibrillation, % (n)	17.2% (5)	22.5% (9)	0.59
NYHA functional class, % (n)			0.19
I	6.9% (2)	7.5% (3)
II	62% (18)	42% (17)
III	20.7% (6)	45% (18)
IV	10% (3)	5% (2)
LVEF, %	27.8 ± 7	27.4 ± 7	0.85
LVESV, mL	135.3 ± 95	139 ± 61	0.81
Medical treatment, % (n)
Beta-blockers	82.7% (24)	87.5% (35)	0.73
ACEI/ARB/ARNI	86.2% (25)	87.5% (35)	0.87
Aldosterone antagonist	72.4% (21)	57.5% (23)	0.20

Abbreviations: ACEIs, angiotensin-converting enzyme inhibitors; ARBs, angiotensin II receptor blockers; ARNIs, angiotensin receptor neprilysin inhibitors; BMI, body mass index; LBBB, left bundle branch block; LVEF, left ventricular ejection fraction; LVESV, left ventricular end-systolic volume; NYHA, New York Heart Association.

Final programmed AV and VV intervals in patients with BiVP were 142 ± 26 and −15 ± 18 ms, respectively.

Dyssynchrony parameters

Regarding the first objective of the study, longitudinal analysis showed a significant improvement over time (P < 0.001) of all echocardiographic dyssynchrony parameters assessed—AV, intraventricular, and VV—but without differences between CSP and BiVP (P-values for group ∗ time > 0.05; Table 2; Graphical Abstract).

Table 2.

Longitudinal analysis of dyssynchrony correction, ‘strain’, and LV remodelling between CSP and BiVP

Outcome	CSP				BiVP				Effects
	15 days OFF	15 days ON	6 months	12 months	15 days OFF	15 days ON	6 months	12 months	Group	Time	Group ∗ time
Prevalence (%)	Adjusted prevalence (95% CI)				Adjusted prevalence (95% CI)				P-value	P-value	P-value
Septal rebound (strain)	76.9 (57.2, 89.2)	33.3 (17.6, 53.9)	24.0 (11.1, 44.2)	15.0 (4.9, 37.6)	76.3 (60.4, 87.2)	44.7 (30.0, 60.1)	34.3 (20.6, 51.2)	31.3 (17.7, 49.0)	0.13	<0.001	0.77
SF	75.9 (57.3, 88.0)	44.8 (28.1, 62.8)	31.0 (17.0, 49.7)	31.0 (17.0, 49.7)	75.0 (59.5, 86.0)	37.5 (24.0, 53.2)	23.7 (12.8, 39.6)	10.8 (4.1, 25.5)	0.10	<0.001	0.49
LBBB ‘strain’ pattern	91.9 (62.4, 98.7)	5.4 (0.6, 32.3)	3.6 (0.4, 26.9)	1.5 (0.1, 21.1)	84.0 (54.9, 95.8)	14.3 (3.6, 43.0)	4.6 (0.7, 23.7)	7.0 (1.2, 30.9)	0.81	<0.001	0.48
Mean	Adjusted mean (95% CI)				Adjusted mean (95% CI)				P-value	P-value	P-value
GLS (%)	−9.6 (−11.0, −8.3)	−10.2 (−11.5, −8.8)	−11.2 (−12.5, −9.8)	−11.3 (−12.7, −9.8)	−8.7 (−9.8, −7.6)	−9.4 (−10.5, −8.3)	−10.0 (−11.2, −8.8)	−11.3 (−12.5, −10.1)	0.32	<0.001	0.55
Mechanical dispersion (ms)	114.8 (102.2, 127.5)	77.2 (64.6, 89.9)	76.4 (63.2, 89.6)	66.5 (53.6, 79.4)	113.5 (103.0, 123.9)	85.7 (75.3, 96.2)	78.3 (67.7, 88.9)	73.7 (62.7, 84.6)	0.56	<0.001	0.54
LV filling time/RR (%)	47.6 (44.4, 50.6)	49.9 (46.8, 53.0)	51.2 (48.1, 54.3)	53.2 (50.1, 56.3)	46.0 (43.4, 48.7)	47.1 (44.5, 49.8)	49.1 (46.4, 51.8)	48.8 (46.1, 51.5)	0.10	<0.001	0.57
Difference outflow times (ms)	45.7 (38.8, 52.5)	23.9 (17.0, 30.7)	22.1 (15.2, 28.9)	16.6 (9.8, 23.5)	35.5 (29.7, 41.4)	23.8 (18.0, 29.6)	24.4 (18.4, 30.4)	21.4 (15.3, 27.6)	0.76	<0.001	0.06
LVESV (mL)	135.3 (110.9, 159.7)	115.5 (91.0, 139.9)	96.0 (71.3, 120.6)	79.9 (55.3, 104.5)	139.2 (118.1, 160.4)	129.3 (108.1, 150.5)	107.7 (86.4, 129.1)	100.1 (78.7, 121.6)	0.39	<0.001	0.66
LVEF (%)	27.8 (24.2, 31.3)	33.3 (29.8, 36.9)	41.6 (38.1, 45.2)	45.8 (42.2, 49.3)	27.4 (24.4, 30.4)	33.7 (30.6, 36.7)	39.5 (36.5, 42.6)	42.7 (39.6, 45.8)	0.53	<0.001	0.31

Abbreviations: CI, confidence interval; CSP, conduction system pacing; GLS, global longitudinal ‘strain’; LBBB, left bundle branch block; LV, left ventricular; LVEF, left ventricular ejection fraction; LVESV, Left ventricular end-systolic volume; RR, R–R interval.

AV dyssynchrony, assessed with LV filling time, showed significant improvement with CSP over time (P < 0.001), with no differences between groups (P = 0.57; Figure 3A and Table 2).

Figure 3.

Correction of dyssynchrony with CSP and BiVP over time. Longitudinal analyses showed that the improvement of all echocardiographic dyssynchrony parameters (A-F) over time was significant (P < 0.001) but without differences between groups. LBBB, left bundle branch block.

Similarly, all four intraventricular dyssynchrony variables—SF, septal rebound stretch, LBBB ‘strain’ pattern, and mechanical dispersion—showed significant correction over time (P < 0.001), with no differences between groups in any of the parameters assessed (Figure 3B–E and Table 2). Presence of SF decreased significantly (P < 0.001) from 75.9% (95% CI 57.3 and 88%) to 31% (95% CI 17 and 49.7%) with CSP and from 75% (95% CI 59.5 and 86%) to 10.8% (95% CI 4.1 and 25.5%) with BiVP; no differences between groups were observed (P = 0.49).

The final dyssynchrony parameter analysed (Figure 3F) was VV dyssynchrony, quantified as the difference between LV and right ventricular outflow times. Again, significant improvement was observed over time (P < 0.001), without differences between groups (P = 0.06).

LV reverse remodelling

Longitudinal analysis showed significant improvement (P < 0.001) of LVEF and LVESV with CSP and BiVP, without differences between groups (Figure 4A and B and Table 2). For CSP, baseline LVEF was 27.8% (95% CI 24.2 and 31.3%) increasing to 45.8% (95% CI 42.2 and 49.3%) at 12-month follow-up (P < 0.001). For BiVP, baseline LVEF was 27.4% (95% CI 24.4 and 30.4%) increasing to 42.7% (95% CI 39.6 and 45.8%) at 12-month follow-up (P < 0.001). No differences were detected between groups for LVEF (P = 0.31).

Figure 4.

Left ventricular remodelling and ‘strain’ improvement with CSP and BiVP. Longitudinal analyses showed a significant improvement of LVEF, LVESV, and GLS over time (P < 0.001), but not different between groups (P = 0.31, P = 0.66, and P = 0.55, respectively) (A, B, C).

Similarly, in the CSP group, baseline LVESV was 135.3 mL (95% CI 110.9, 159.7 mL) decreasing to 79.9 mL (95% CI 55.3, 104.5 mL). For BiVP, the change was from baseline 139.2 mL (95% CI 118.1, 160.4 mL) to 100.1 mL (95% CI 78.7, 121.6 mL). No LVESV differences were observed between CSP and BiVP (P = 0.66; Figure 4B). Considering the improvement of LVESV at 1-year follow-up as a percentage, the means were −34.4% (95% CI −44.9 and −23.8%) for CSP and −26.8% (95% CI −33.2 and −17.3%) for BiVP, without differences between groups (P = 0.29).

For both CRT modalities (CSP and BiVP), GLS improved significantly over time (P < 0.001), without differences between groups (P = 0.55; Table 2 and Figure 4C).

Correlation of baseline GLS and ventricular remodelling

Baseline GLS significantly correlated with LVEF (r = 0.54, P < 0.001; Figure 5A) and LVESV (r = −0.51, P < 0.001) at the 12-month follow-up.

Figure 5.

Baseline GLS predicting LVEF improvement and response at 12-month follow-up. (A) Baseline GLS significantly correlated with LVEF at 12 months of follow-up (r = 0.54, P < 0.001). (B) The area under the ROC curve was calculated for GLS; the value of baseline GLS of the ROC curve with the best sensitivity and specificity to discriminate between those patients who will present an echocardiographic response (LVEF > 5%) was GLS −7.1%.

The capacity to predict echocardiographic response was analysed for all dyssynchrony parameters tested in our study and for GLS (Table 3). Percentage of response was 84% (58/69). The variable that better predicted the response was GLS. The area under the ROC curve was calculated for GLS; the value of baseline GLS of the ROC curve with the best sensitivity and specificity to discriminate those patients who will present echocardiographic response (LVEF > 5%) was −7.1% (sensitivity 73%, specificity 78%, positive predictive value 40%, negative predictive value 93%, area under the curve 0.77) and OR 9.11 (95% CI 2.09, 39.81, P < 0.01; Figure 5B). Considering an absolute improvement of GLS > 5% as criteria for response, we found 25.5% of responders, in line with previous publications,²³ without differences between groups (P = 0.82).

Table 3.

Baseline dyssynchrony parameters as predictors for LV remodelling

	AUC (95% CI)	Cut-off	Sensitivity	Specificity	PPV	NPV	Diagnostic OR	P
GLS	0.77 (0.61, 0.93)	−7.1%	73%	78%	40%	93%	9.11 (2.09, 39.81)	<0.01
Mechanical dispersion	0.62 (0.44, 0.80)	106	82%	55%	26%	94%	5.4 (1.07, 27.3)	0.04
LV filling time/RR	0.52 (0.31, 0.72)	54.5	27%	85%	25%	86%	1.64 (0.42, 6.39)	0.48
Difference outflow times	0.57 (0.40, 0.73)	4.5	100%	7%	17%	100%	—	—
Septal rebound (‘strain’)	—	—	73%	23%	16%	80%	0.78 (0.18, 3.41)	0.74
Septal flash	—	—	45%	19%	10%	65%	0.20 (0.05,0.76)	0.02
LBBB ‘strain’ pattern	—	—	55%	25%	13%	72%	0.39 (0.10, 1.49)	0.17

Abbreviations: GLS, global longitudinal ‘strain’; PPV, positive predictive value; NPV, negative predictive value; RR, R–R interval.

In relation to SF, a 1-mm SF correction was the better cut-off to predict response at 12-month follow-up (sensitivity 55%, specificity 72%, positive predictive value 27%, negative predictive value 89%) with an area under the ROC curve of 0.68 (P = 0.02) and OR for response of 3.15 (95% CI 0.84, 11.78; P = 0.09).

Agreement and correlation of ‘strain’ measurements

Inter-observer reproducibility of ‘strain’ measurements between two blind observers was analysed. Septal rebound stretch showed almost perfect agreement with a Cohen’s kappa coefficient of 0.82 (95% CI 0.60, 1.0; P < 0.001). LBBB ‘strain’ pattern showed substantial agreement with a Cohen’s kappa coefficient of 0.67 (95% CI 0.37, 0.97; P < 0.001). GLS showed substantial concordance with a correlation coefficient of 0.95 (95% CI 0.90, 0.98; P < 0.001; see Supplementary data online, Figure B). Last of all, mechanical dispersion showed substantial concordance with a correlation coefficient of 0.97 (95% CI 0.94, 0.99; P < 0.001; Supplementary data online, Figure C).

Clinical endpoints at 12-month follow-up

In relation to mortality or cardiac transplant, treatment-received analysis showed four events (4/40, 10%) in the BiVP arm and no events (0/29, 0%) in the CSP arm (P = 0.08 for superiority). Heart failure hospitalizations were 17.5% (7/40) in the BiVP arm and 0% (0/29) in the CSP arm (P = 0.02 for superiority).

Discussion

Similar dyssynchrony and ‘strain’ correction achieved by CSP and BiVP

The main finding of our study is that the ability to correct mechanical dyssynchrony (AV, intraventricular, and VV) is similar for both CSP and BiVP. Although multiple single-centre observational studies support the use of echocardiography to assess cardiac mechanics as a tool to identify favourable patients for such therapy,^12–14,24 none of the dyssynchrony parameters have demonstrated sufficient robustness in multicentre studies to be defined as selection criteria for candidates to CRT.¹⁵ However, most of these studies have been developed in patients receiving BiVP, and data regarding CRT with CSP are lacking.

Two studies have used ‘strain’-based techniques to evaluate the differences in mechanical dyssynchrony between isolated right ventricular pacing and physiological pacing in patients with indication for pacemaker implantation. Both found greater homogeneity in the contraction of different segments of the left ventricle in the patients receiving physiological pacing.^25,26 In a multicentre—but non-randomized—study, Liu et al.²⁷ compared the reduction in mechanical dispersion and improvement in myocardial function in a cohort of 62 CRT candidates (35 with BiVP and 27 with LBBP), showing that physiological pacing achieved better results compared with BiVP.

Very few randomized studies exist,^6,7,9 and none has compared the effect of both pacing modalities with echocardiographic dyssynchrony parameters. In our sub-study, we observed that the correction of AV, intraventricular, and VV dyssynchrony parameters was similar and evolved in parallel in both groups (Graphical abstract; Figure 3).

SF has been extensively studied in this context,^12,14,18,19 and its value lies in being an easily obtainable measure with high inter- and intra-observer reproducibility.^12,28 In the context of physiological pacing, our group carried out an observational study that included 30 CRT patients with CSP and reported a significant reduction in SF magnitude, which correlated with an increase in LVEF; SF predicted echocardiographic response at 6-month follow-up, with a correction ≥ 1.5 mm as a cut-off point with a larger area under the ROC curve.¹⁹ The current study confirms a cut-off value of SF correction > 1 mm to predict response at 12-month follow-up. The use of other intraventricular dyssynchrony ‘strain’-based parameters (such as septal rebound stretch and typical LBBB contraction pattern) permits a more comprehensive assessment of global LV mechanics (rather than just septal mechanics) and again showed similar improvement with both pacing techniques.

Some studies have analysed the utility of mechanical dispersion as a predictor of ventricular arrhythmias in different clinical scenarios.^29–32 Additionally, its capacity as a predictor of response to CRT has also been proposed.³³ Our analysis of mechanical dispersion, according to the methodology used in previous studies, suggests that there are no differences in the correction of this parameter between both therapeutic strategies, although its capacity as a predictor of response was moderate with an area under the ROC curve of 0.62 (95% CI 0.44, 0.80) for a mechanical dispersion value of 106 ms, which may be explained by the small sample size.

Of note, the parallel improvement of all the assessed parameters highlights the value of each one in assessing mechanical dyssynchrony and monitoring its correction with CSP or BiVP. Many previous studies have focused on finding an optimal parameter to assess mechanical dyssynchrony, while others shed some doubt due to its lack of usefulness to predict response.¹⁵ Our results showed a similar improvement in all the parameters assessed, although no individual parameter was as powerful as GLS to predict response. Interestingly, all parameters showed good negative predictive value, underlining the low probability of echo response in patients without any degree of mechanical dyssynchrony at baseline. However, we should take into account the low percentage of no-response (16%, 11/69) in our study.

GLS predicted ventricular remodelling

Our second major finding is that GLS value is a predictor of response (defined as an increase in LVEF > 5%) at 12-month follow-up. This finding is consistent with observational studies published to date. A recent meta-analysis by Bazoukis et al.³⁴ included 12 studies and demonstrates that baseline GLS value and its increase over time are predictors of reverse LV remodelling and CRT response. However, it should be noted that this meta-analysis included observational studies with heterogeneous definitions of therapy response. In our study, a GLS value of −7.1% is the cut-off that best discriminated patients who responded to CRT, yielding better sensitivity, specificity, and area under the ROC curve. As D’Andrea et al.³⁵ conclude, having better baseline GLS values may indicate less myocardial scar burden estimated by magnetic resonance and thus a larger recruitable myocardium for resynchronization, regardless of the pattern of intraventricular mechanical dyssynchrony, resulting in a higher response rate in this patient group, as was observed in our study. The mean difference in baseline GLS between responders and non-responders was −2.85 (95% CI −4.79, −0.90, P < 0.01), which is consistent with previous data.³⁴

In our study, most of the patients (>75 and even >90% had a ‘strain’ pattern of LBBB) presented echocardiographic parameters consistent with the presence of dyssynchrony (Table 2). Therefore, these parameters did not show discriminatory impact on the prediction of CRT response. Nonetheless, our results highlight the evidence that even though mechanical dyssynchrony must be present, it is not sufficient to ensure response to CRT.

Clinical implications

To date, very limited randomized data regarding resynchronization with HBP^6,7,36 and LBBP^8–10 have been available. Non-inferiority of CSP (including HBP or LBBP) compared with BiVP in achieving cardiac resynchronization has been previously demonstrated.⁸ The LEVEL-AT trial showed that CSP provides a degree of electrical resynchronization similar to BiVP in patients with CRT indication.⁸ In addition, as a secondary endpoint, the same trial showed non-inferiority of ventricular remodelling and clinical outcomes. The LBBP-RESYNC trial⁹ demonstrated greater LVEF improvement than BiVP in heart failure patients with non-ischaemic cardiomyopathy and LBBB. Both studies assessing LBBP-LEVEL-AT and LBBP-RESYNC published results at 6 months of follow-up.

The results provided by our study offer a longitudinal treatment-received analysis with 12 months of follow-up. The usefulness of CSP to correct mechanical dyssynchrony was demonstrated, achieving a degree of reverse LV remodelling as good as the gold standard therapy (BiVP). Our results support the use of CSP as an alternative method to BiVP.

It is important to emphasize the use of the FOI method in all the patients receiving BiVP in the LEVEL-AT trial:^16,17 AV and VV intervals were optimized according to the narrowest QRS using fusion with intrinsic conduction. In a previous randomized trial,^17,37 the FOI method demonstrated its ability to narrow the QRS, achieving greater LV remodelling compared with nominal settings. Other published studies comparing CSP with BiVP had not optimized the BiVP patients,^6,38 which may explain their superior results with CSP. The relatively high response rate in our study could be related with the evolution in the heart failure medical treatment,³⁹ advances of CRT device implants, and FOI optimization.¹⁷

Larger randomized trials are warranted to confirm these results. Superiority of CSP could not be demonstrated for all the CRT population; however, the most appropriate approach for CRT would be to find the best niche for each pacing method. It may be less about demonstrating the superiority of CSP and more about personalizing the therapy for each particular patient.⁴⁰

Limitations

Due to the crossover percentage in the LEVEL-AT trial (23% from CSP to BiVP, and 5.7% from CSP to BiVP), the percentage of patients that received CSP was 42% (29/69) rather than 50%. Strocchi et al.⁴¹ showed that CSP was ineffective in patients with severe LV His–Purkinje conduction disease or septal scar. Selection of the best CRT modality before implant could result in a lower crossover ratio.

In this study, we suggest that patients with GLS worse than −7.1% will present nine-fold times greater risk for no-response with either CSP or BiVP. Large multicentre studies are needed to establish predictors for no-response to each type of therapy. Given the still-considerable percentage of crossovers in CSP, personalizing the therapy according to the baseline characteristics of the patient may achieve better results.

The small sample cannot demonstrate the difference or superiority of LBBP compared with HBP. Clinical endpoints at 12-month follow-up (mortality and heart failure hospitalizations) should be interpreted with caution because of the sample size. Finally, our relatively small sample sizes for the longitudinal models with interaction (time × treatment) and consequent lack of statistical power may have limited the capacity to detect between-group differences in outcomes. Maybe our analysis is not capable to detect some important clinical differences. Larger randomized studies are warranted to address this point.

Myocardial work could not be determined because blood pressure was not assessed at the time of the echocardiography.

Conclusion

CSP and BiVP provided similar dyssynchrony and ‘strain’ correction over time in patients who successfully received the device (treatment-received analysis). Baseline GLS predicted ventricular remodelling at 12-month follow-up.

Supplementary data

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

Funding

M.P.-L. received the Josep Font contract 2019–2022 from Hospital Clínic Barcelona (Catalonia, Spain), the Catalan Society of Cardiology research grant 2020 (Catalonia, Spain), and the Research Grant 2020 from Asociación del Ritmo Cardiaco (Spanish Society of Cardiology) and is currently funded (2023–2025) through a Río Hortega contract CM22/00107 [Instituto de Salud Carlos III (ISCIII); Fondo Social Europeo (FSE)]. J.M.T., M.A.C., M.P.-L., and R.J.-A. received funding for Project in Conduction System Pacing (FIS PI21/00615) from Instituto de Salud Carlos III (Madrid, Spain). L.M. was funded by a 2021 Support for Research Groups grant from the Catalan government (2021_SGR_01350, SGR21/GENCAT), Catalonia, Spain. J.B.G. is currently funded through a ‘Contractes d’Investigació Avançada Fundació BBVA—Hospital Clínic Barcelona Joan Rodés—Josep Baselga 2022’ contract (HCB_BIO_001/2).

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.