Patient selection

We evaluated 88 patients undergoing either Boston Lotus (Boston Scientific Corporation, Natick, MA) or Medtronic CoreValve (Medtronic Inc., Minneapolis, Minnesota) TAVR for severe symptomatic aortic stenosis at a single tertiary centre from April 2009 to April 2015. Exclusion criterion included pre-existing QRS prolongation (>120 ms) or contra-indication to CMR scanning. Decision for TAVR was taken by a multi-disciplinary heart team in accordance with international guidance [15]. Trans-femoral was the default approach with other techniques (subclavian and carotid) employed if femoral access was unsuitable.

Matching

24 patients with LBBB-T were identified. These were matched with 24 patients with a nQRS post-procedure for sex, valve type, and baseline CMR variables known to impact on reverse remodeling following TAVR including LV ejection fraction (LVEF), indexed LV mass and indexed LV end diastolic volume (LVEDVi) [16] (Fig. 1).

Electrocardiographic data

12-lead electrocardiogram recordings acquired immediately prior to TAVR and at the time of post-procedure hospital discharge were reviewed by a single author blinded to clinical and procedural data. Heart rhythm, PR interval and QRS duration were recorded. LBBB-T was defined as post-procedural v1-negative QRS complex with a duration of >120 ms and a notched or slurred R wave in at least one of the lateral leads (I, aVL, V₅, V₆) [17].

CMR protocol

Details of the CMR pulse sequence acquisition protocol have been published previously [16]. Briefly, identical CMR scans were obtained at baseline and 6 m following TAVR using a 1.5 T scanner (Intera, Philips Healthcare, Best, Netherlands or Avanto, Siemens Medical Systems, Erlangen, Germany). Multi-slice, multi-phase cine imaging was performed using a standard steady-state free precession pulse sequence in the short axis (10 mm thickness, 0 mm gap, 30 phases, 192 by 192 matrix, typical field of view 340 mm) to cover both ventricles. Standard 2, 3 and 4 chamber cine images were also acquired. Through-plane velocity encoded phase contrast imaging was performed perpendicular to the aortic valve jet at the aortic sinotubular junction (VENC 250–500 cm/s, retrospective gating, slice thickness 6 mm, 40 phases). Late gadolinium enhancement (LGE) imaging (10–12 short axis slices, 10 mm thickness, matrix 240x240, 320–460 mm field of view) was performed with inversion time individually adjusted according to TI scout, 10–15 min after 0.2 mmol/kg of gadoteric acid (Dotarem, Guerbet, Villepinte).

CMR analysis

CMR analysis was performed by a single experienced operator blinded to clinical data using cmr⁴² (Circle Cardiovascular Imaging, Alberta, Canada). Endocardial and epicardial contours were manually contoured at end-diastole and end-systole with papillary muscles and trabeculations excluded to allow the calculation of ventricular volumes (summation of discs methodology) and mass (epicardial volume - endocardial volume multiplied by myocardial density (1.05 g/cm³)). Values were indexed to body surface area. Post-procedural myocardial infarct was determined by direct comparison of pre- and 6 m CMR LGE acquisitions. Fibrosis mass was quantified using a threshold of 5 standard deviations technique [18]. Left atrial volume was calculated using the biplane area-length method [19]. Aortic flow was quantified using cross-sectional phase contrast images with the slice positioned at least 10 mm above the aortic prosthesis and contouring of the aortic lumen to provide a regurgitant fraction (%). Longitudinal right ventricular function was measured at the lateral tricuspid annulus in the 4 chamber cine view as the distance between end systole and end diastole. Feature tracking analysis was performed on cine imaging of the mid ventricular short axis slice at the papillary muscle level to determine time to peak LV and right ventricular radial strain and the 4 chamber cine to measure global longitudinal strain. Interventricular dyssynchrony was the difference between time to peak radial strain of the right ventricular free wall and the lateral LV wall (an average of segments 11 and 12 of the American Heart Association 17 segment model). Intraventricular dyssynchrony was the difference between time to peak radial strain of the LV septal (an average of segments 8 and 9) and lateral LV segments (an average of segments 11 and 12). Segments with LGE indicating previous myocardial infarction were excluded from radial strain analysis. For the assessment of inter-observer variability, two independent investigators analysed LV volume, mass and function, GLS and time to peak radial strain on a random selection of 10 patients. For intra-observer variability a similar dataset from 10 patients was analysed twice by one investigator one month apart. The coefficient of variation was calculated by dividing the standard deviation of the differences between measurements by their mean and expressed as a percentage. Inter-observer variability was 1.4, 4.5, 3.7, 9.2 and 12.6% and intra-observer variability was 6.8, 2.6, 5.0, 2.6%, 6.8 and 9.1% for LVEDV, LV mass, LVEF, GLS and time to peak radial strain respectively.

Statistical analysis

Statistical analysis was performed using SPSS version 22 (IBM, Armonk, New York). Categorical data were presented as numbers (percentages) and compared using the Pearson Chi squared test. Continuous variables were expressed as mean ± SD and were tested for normality using the Shapiro-Wilks test. Data were compared using Students t Test (for normally distributed data) and the Mann-Whitney or Wilcoxen signed rank test (for non-normally distributed data). Linear regression analysis (Enter model) was performed to establish univariate and multivariate predictors of change in LVEF and GLS post-procedure. Univariate predictors with P < 0.05 were included in the multivariate analysis. P < 0.05 was considered statistically significant. Using CMR, in order to detect a 3% difference in LVEF with a 90% power and an α error of 0.05 a sample size of at least 12 patients in each arm was required [13].

Electrocardiographic characteristics

Mean heart rate at baseline was 67 ± 11 and at 6 m was 68 ± 13 bpm. 7 patients (15%) (nQRS n = 2, LBBB-T n = 5) had atrial fibrillation at baseline. There were no new cases of post-procedural atrial fibrillation. For those in sinus rhythm, mean PR interval remained similar pre and post procedure in both the nQRS group (179 ± 33 to 191 ± 39 ms, p = 0.053) and the LBBB-T group (181 ± 30 to 192 ± 37 ms, p = 0.171). In the nQRS group there was no change in QRS duration (93 ± 17 to 96 ± 11 ms, p = 0.098). In the LBBB-T group, QRS duration increased from 96 ± 14 to 151 ± 12 ms (p < 0.001).

Reverse remodeling according to post-procedure QRS duration

Change in LVEF and GLS was significantly different between the two groups (LVEF: nQRS +4.6 ± 7.8 vs LBBB-T -2.1 ± 6.9%, p = 0.002 and GLS: nQRS -2.1 ± 3.6 vs. LBBB-T +0.2 ± 3.2%, p = 0.024) (Fig. 2). The change in LVEF was driven by a reduction in indexed LV end systolic volume (LVESVi) in the nQRS group not seen in the LBBB-T group (nQRS -7.9 ± 14.0 vs. LBBB-T -0.6 ± 10.2 ml/m², p = 0.02). Pre and post-procedure values for all CMR characteristics can be seen in Table 2. Change in indexed LV mass was similar between the two groups (nQRS -15.9 ± 10.4 vs LBBB-T -13.3 ± 9.6 g/m², p = 0.367) as was change in LVEDVi (nQRS -7.3 ± 17.4 vs LBBB-T -3.2 ± 14.5 ml/m², p = 0.373). Neither group experienced any change in right ventricular longitudinal function (nQRS 21.7 ± 7.0 to 21.5 ± 6.2 mm, p = 0.817, LBBB-T 18.9 ± 5.8 to 18.6 ± 5.8 mm, p = 0.773). Post-procedure aortic regurgitant fraction was similar between groups (nQRS 5.4 ± 5.7 vs LBBB-T 5.5 ± 3.3%, p = 0.948). There was a significant correlation between change in QRS and change in LVEF (r = -0.434, p = 0.002) and between change in QRS and change in GLS (r = 0.462, p = 0.001). The relationship between post-procedure QRS duration and change in LVESVi, LVEF and GLS can be seen in Fig. 2.

	nQRS (n = 24)	LBBB-T (n = 24)	P Value
Left ventricular ejection fraction
Pre-procedure	54.1 ± 11.5	56.6 ± 10.5	0.386
Post-procedure	58.7 ± 9.0	54.4 ± 9.3	0.070
P Value	0.010	0.092
Global longitudinal strain, %
Pre-procedure	−15.6 ± 3.9	−16.2 ± 4.2	0.638
Post-procedure	−17.7 ± 2.7	−15.9 ± 3.4	0.053
P Value	0.009	0.771
Indexed left ventricular mass, g/m2
Pre-intervention	74.3 ± 14.7	73.3 ± 17.4	0.650
Post-intervention	58.4 ± 12.6	60.0 ± 13.7	0.665
P Value	<0.001	<0.001
Indexed left ventricular end diastolic volume, ml/m2
Pre-intervention	97.8 ± 22.8	93.4 ± 22.1	0.500
Post-intervention	90.5 ± 21.0	90.3 ± 21.0	0.968
P Value	0.051	0.298
Indexed left ventricular end systolic volume, ml/m2
Pre-intervention	46.6 ± 20.4	41.8 ± 17.7	0.458
Post-intervention	38.7 ± 16.2	42.4 ± 17.8	0.523
P Value	0.011	0.886
Indexed left ventricular stroke volume, ml/m2
Pre-intervention	51.2 ± 10.3	51.4 ± 10.5	0.945
Post-intervention	51.8 ± 8.7	47.9 ± 8.5	0.122
P Value	0.742	0.035
Indexed left atrial volume, ml/m2
Pre-intervention	67.9 ± 19.2	72.9 ± 23.3	0.232
Post-intervention	60.0 ± 18.2	67.9 ± 23.8	0.199
P Value	0.002	0.180

nQRS narrow QRS post-procedure, LBBB-T new LBBB post-procedure

Inter- and intra-ventricular dyssynchrony

As a group as a whole baseline inter- and intra-ventricular dyssynchrony was 68 ± 62 and 54 ± 83 ms respectively. Those that subsequently developed LBBB demonstrated more interventricular dyssynchrony at baseline but had similar baseline intraventricular dyssynchrony (Inter: LBBB-T 88 ± 61 vs. nQRS 48 ± 57 ms, p = 0.021, Intra: LBBB-T 74 ± 90 vs. nQRS 35 ± 73 ms, p = 0.108). In the nQRS group, TAVR was not associated with an improvement in dyssynchrony (Inter: pre-TAVR 47 ± 57 vs. post-TAVR 23 ± 86 ms, p = 0.174, Intra: pre-TAVR 35 ± 73 vs. post-TAVR 7 ± 102 ms, p = 0.207). There was evidence of significant inter- and intra-ventricular dyssynchrony in the LBBB-T group at 6 m compared with the nQRS population (Inter: LBBB-T 130 ± 73 vs. nQRS 23 ± 86 ms, p < 0.001, intra: LBBB-T 118 ± 103 vs. nQRS 13 ± 13 ms, p = 0.001). There was a significant correlation between post-procedure QRS duration and inter- and intra-ventricular dyssynchrony (r = 0.57, p < 0.001 and r = 0.49, p = <0.001 respectively). A typical LV contraction pattern in nQRS and LBBB-T can be seen in Fig. 3.

Predictors of change in LVEF and change in GLS

Impact of TAVR-induced LBBB on cardiac reverse remodeling

TAVR-induced LBBB is common, occurring in 16–65% patients depending on valve type [2]. Although predictors of LBBB-T have been extensively studied [2], the impact of LBBB-T on cardiac reverse remodeling is less well described, with studies limited to echocardiographic evaluation and containing a heterogeneous mix of patients. A PARTNER echocardiographic sub-study reported a lower LVEF at 12 months in patients with LBBB on discharge electrocardiogram compared to those with a narrow QRS, however, there was an increased number of those undergoing trans-apical TAVR in the LBBB-T group [6], findings which were replicated in another similar study, again with more undergoing trans-apical TAVR in the LBBB-T group [20]. Tzikas et al. [10] reported an 8% difference in LVEF between those with LBBB-T and nQRS prior to and 6 days following self-expanding TAVR. Longitudinal strain was also non-significantly reduced in those with new conduction abnormalities. Hoffman et al. [11] investigated 90 patients using 2D and speckle tracking trans-thoracic echocardiography prior to and at 1 and 12 months following TAVR. Patients with new conduction defects had a significantly larger LVESVi at 12 months compared with those with a narrow QRS, with less difference in LVEDVi, mirroring the findings in our study. New conduction defects and baseline LVEF were independent predictors of reduction in LVEF at 12 months. The inclusion of patients with trans-apical access in the majority of these studies [6, 11, 20] and those with post-procedural pacemaker insertion [6, 10, 11, 20] is a significant confounder, however, given that trans-apical access has been linked to reduced LVEF in a number of studies [20, 21] and pacing induced LBBB has been shown to cause different patterns of strain to those with idiopathic LBBB [22].

Our study adds further insight into the impact of LBBB-T on cardiac reverse remodeling. The accuracy and reproducibility of CMR means that important differences can be determined using studies 87% smaller than echocardiographically based studies, with only 11 patients per group required to detect a 3% difference in LVEF [13]. Our study groups were matched for clinical and baseline CMR characteristics, all parameters which have been found to strongly influence reverse remodeling following valve intervention [16]. None of the patients in our study received trans-apical TAVR or permanent pacemaker insertion and the unique ability of CMR LGE imaging allowed us to identify and exclude any patients who had a post-procedural myocardial infarction, another factor that may have confounded the earlier echocardiographically based studies. Finally, the two groups experienced similar amounts of post-procedural aortic regurgitation, which is an important modulator of post-TAVR reverse remodeling [23, 24], and which was not reported in most of the echocardiographic studies [6, 10, 11].

Inter and intra-ventricular dyssynchrony

The novel use of CMR feature tracking allows us to report values for intra- and inter-ventricular dyssynchrony. In LBBB, the normally functioning right bundle conducts the electrical impulse to the right ventricle prompting early right ventricular contraction followed by activation of the interventricular septum and finally lateral wall contraction resulting in inter- and intra-ventricular dyssynchrony. This dyssynchrony leads to the classical abnormal septal motion pattern of contraction seen in LBBB which is felt to impair LV filling and ejection in its own right (Fig. 3). This dysynchronous contraction leads to an increase in LVESVi, as seen in our LBBB-T group and it is this, rather than a change in LVEDVi that is the largest driver of reduction in LVEF. We have also shown that LBBB-T impacted on change in GLS, with no improvement in this group compared to a significant improvement in the nQRS group. Although GLS may be affected by dyssynchrony [25], this, coupled with the reduction in left atrial volume in the nQRS but not the LBBB-T group, and the reduction in LV stroke volume in those with LBBB-T, suggests that the effects of LBBB-T may go beyond that of simple mechanical dyssynchrony.

Conduction system damage during TAVR

It is well established that TAVR leads to conduction abnormalities [2]. Trauma can occur at multiple time-points during the TAVR procedure; from guidewire manipulation, to during balloon valvuloplasty, device manipulation and valve deployment. It is likely that the different valve designs can cause differing degrees of compression to the conduction system; with the self-expanding Medtronic CoreValve felt to cause more compression of the LV outflow tract than the balloon expandable Edwards Sapien device [26]. The unique design of the mechanically expandable and repositionable Boston Lotus valve with its adaptive seal, may also be associated with more conduction system trauma, although reports to date are limited [27]. Global ischaemia during rapid pacing required for valve deployment may exacerbate the issue [2]. Other procedure-related factors felt to be implicated include pre-implant valvuloplasty, deep implant, low ratio of the annulus:balloon or annulus:prosthesis and operator experience [28].

Clinical implications

The impact of TAVR-induced LBBB on mortality is a subject of debate, however, it has been shown in many studies to be a predictor of mortality [3–5, 29] and has been associated with increased hospitalisation [6]. Other studies have failed to demonstrate a link [12, 20, 30]. Nonetheless, LVEF is a strong independent predictor of long term survival [31]. Our study has shown that TAVR-induced LBBB is associated with reduced global longitudinal and radial systolic function compared with those with a narrow post-procedure QRS, which could partially explain the link with mortality. Given the adverse effect of TAVR-induced LBBB on cardiac reverse remodeling, restoring inter- and intra-ventricular dyssynchrony using cardiac resynchronisation therapy, could be considered, especially if another conventional indication for device therapy exists. Furthermore, every effort should be made by the operator to reduce the risk of TAVR-induced LBBB given the adverse effects on ventricular remodeling seen. As newer devices are being developed, designs should be focused on minimising damage to the electrical conducting system in order to prevent the deleterious effects on the LV that this entails.

Study limitations

Although patients were recruited in a prospective manner, they were matched retrospectively and hence the study is prone to the selection bias of this type of study. Patients with LBBB-T were matched according to those factors known to influence cardiac reverse remodeling [16] but other factors may be unaccounted for. Specifically, patients with coronary artery disease and previous myocardial infarction were included in the study, however, numbers in each group were similar and infarct pattern LGE at baseline was not a univariate predictor of change in LVEF or GLS. Group allocation was based on the discharge electrocardiogram and not re-confirmed at 6 months, however, there are evidence to suggest that virtually all those with LBBB at discharge have persistent LBBB at 30 days following self-expanding TAVR [30]. Furthermore, the demonstration of ongoing dyssynchrony at 6 m in the LBBB-T group suggests that the conduction abnormality was persistent. Finally, although adequately powered to detect a difference in reverse remodeling using the accurate technique of CMR, the study is small with a relatively short follow-up period and a larger study with a longer follow-up interval may be helpful to further investigate the impact of TAVR-induced LBBB on cardiac reverse remodeling.