Electrocardiographic predictors of successful resynchronization of left bundle branch block by His bundle pacing

Abstract Background His bundle pacing (HBP) is an alternative to biventricular pacing (BVP) for delivering cardiac resynchronization therapy (CRT) in patients with heart failure and left bundle branch block (LBBB). It is not known whether ventricular activation times and patterns achieved by HBP are equivalent to intact conduction systems and not all patients with LBBB are resynchronized by HBP. Objective To compare activation times and patterns of His‐CRT with BVP‐CRT, LBBB and intact conduction systems. Methods In patients with LBBB, noninvasive epicardial mapping (ECG imaging) was performed during BVP and temporary HBP. Intrinsic activation was mapped in all subjects. Left ventricular activation times (LVAT) were measured and epicardial propagation mapping (EPM) was performed, to visualize epicardial wavefronts. Normal activation pattern and a normal LVAT range were determined from normal subjects. Results Forty‐five patients were included, 24 with LBBB and LV impairment, and 21 with normal 12‐lead ECG and LV function. In 87.5% of patients with LBBB, His‐CRT successfully shortened LVAT by ≥10 ms. In 33.3%, His‐CRT resulted in complete ventricular resynchronization, with activation times and patterns indistinguishable from normal subjects. EPM identified propagation discontinuity artifacts in 83% of patients with LBBB. This was the best predictor of whether successful resynchronization was achieved by HBP (logarithmic odds ratio, 2.19; 95% confidence interval, 0.07–4.31; p = .04). Conclusion Noninvasive electrocardiographic mapping appears to identify patients whose LBBB can be resynchronized by HBP. In contrast to BVP, His‐CRT may deliver the maximum potential ventricular resynchronization, returning activation times, and patterns to those seen in normal hearts.

successfully shortened LVAT by ≥10 ms. In 33.3%, His-CRT resulted in complete ventricular resynchronization, with activation times and patterns indistinguishable from normal subjects. EPM identified propagation discontinuity artifacts in 83% of patients with LBBB. This was the best predictor of whether successful resynchronization was achieved by HBP (logarithmic odds ratio, 2.19; 95% confidence interval, 0.07-4.31; p = .04).

Conclusion:
Noninvasive electrocardiographic mapping appears to identify patients whose LBBB can be resynchronized by HBP. In contrast to BVP, His-CRT may deliver the maximum potential ventricular resynchronization, returning activation times, and patterns to those seen in normal hearts.  1 The aim of cardiac resynchronization therapy (CRT) is to correct this abnormality of electrical activation to improve cardiac function in patients with heart failure and LBBB. Biventricular pacing (BVP) is the most widely used method for delivering CRT and has been shown to improve symptoms, clinical outcomes, and mortality. 2 However, many patients treated with BVP continue to experience high symptom burdens and poor prognoses. 3 BVP is thought to produce its beneficial effect by delivering more synchronous ventricular activation and shortening atrioventricular interval. 4 However, BVP does not completely correct the disordered ventricular activation that occurs with LBBB. In fact, it produces only relatively modest reductions in ventricular activation time and results in nonphysiological ventricular activation patterns. 1 As a result, there has been interest in the development of more effective CRT.
His bundle pacing (HBP) has recently been proposed as a method for delivering more effective ventricular resynchronization than BVP. 5 His-cardiac resynchronization therapy (His-CRT) resynchronizes ventricular activation by overcoming proximal conduction system block, thereby activating the ventricles via the native His-Purkinje system. 6,7 His-CRT can produce greater QRS duration (QRSd) shortening and LVAT reduction than BVP, which translate to larger improvements in acute hemodynamic function. 3 However, it does not appear to be possible to shorten ventricular activation time in all patients with 12-lead ECG features of LBBB. 8 HBP is most likely to deliver ventricular resynchronization in patients with proximal conduction system disease. The 12-lead ECG is an imperfect tool for identifying the mechanism of conduction impairment. Upadyay et al. found, using intracardiac septal mapping, that conduction block within the left-sided His fibers or proximal portion of the left bundle branch was present in 64% of patients with a 12-lead ECG appearance of LBBB. These patients have the highest chance of successful resynchronization with HBP. 7 A second challenge is that it is difficult, using the 12-lead ECG, to quantify left ventricular resynchronization and therefore to determine whether the maximum potential resynchronization has been achieved: LVAT cannot be easily identified with the 12-lead ECG.
This can be particularly challenging in the presence of nonselective His bundle capture, which results in a pseudo-delta wave as a result of local right ventricular myocardial capture at the lead tip. 3 Furthermore, it is not known whether the reduction in activation times achieved with HBP is associated with restoration of the normal physiological ventricular activation pattern. 9 Therefore, although His-CRT is a promising alternative to BVP, before proceeding to long-term head-to-head randomized control trials it would be helpful to develop tools for improving patient selection and intra-procedural targets: 1) Noninvasive identification of patients in whom His-CRT is likely to be successful would allow resynchronization strategy to be targeted to underlying conduction disorder.
2) Quantification of ventricular electrical resynchronization, with a normal range, will allow operators to easily establish whether ventricular resynchronization has been delivered and assess whether maximum activation time reduction has been realized.
We addressed these two questions by analysing the activation times and patterns of HBP, BVP, and intrinsic activation during LBBB ARNOLD ET AL. and compared these to intact conduction systems using noninvasive epicardial activation and propagation mapping.

| Epicardial propagation mapping (EPM)
To determine activation patterns, custom software was used to render movies of wavefront propagation across the epicardium. The entire duration of a single beat's EGM for each virtual epicardial electrode was visually represented on the 3D cardiac model. The (i.e., positive dv/dt) clipped. By displaying the entire EGM, rather than the most negative dv/dt, visual interpretation was used to determine activation wavefronts rather than relying on potentially mis-annotated activations.

| Pacing
Temporary HBP was performed in patients with LBBB and was achieved via either the femoral or subclavian approach. If the femoral route was used, a quadripolar electrophysiology catheter was placed on the bundle of His. If the subclavian route was used, a SelectSecure 3830 lead was delivered via either a C304-His deflectable sheath or C315 fixed curve sheath (leads and delivery system: Medtronic). 14 The lead was not actively fixated unless BVP failed, in which case the SelectSecure 3830 lead was deployed for permanent HBP. Selective and nonselective His bundle capture was determined using standard criteria. 15 BVP was performed using the standard technique and AAI pacing was performed from the right atrial appendage.

| Statistical analysis
Patients were classified by the degree of resynchronization occurring with HBP. Successful resynchronization, referred to as His-CRT, was defined as reduction in LVAT-95 by 10 ms from intrinsic activation. This cut-off was applied because we have previously demonstrated that the variation of LVAT-95 in this setting is 10 ms 3 . Therefore, smaller reductions in LVAT-95 may be due to measurement variation rather than true resynchronization. Failed His-CRT was defined as a <10 ms reduction in LVAT-95 compared to intrinsic activation (as long as 12-lead ECG criteria for His bundle capture were fulfilled).
His-CRT was further subclassified into incomplete His-CRT and complete His-CRT. Complete His-CRT was defined as HBP LVAT-95 within the 99% range (mean ± 2.58 × SD) for intrinsic LVAT-95 in patients with normal QRS. Incomplete His-CRT was defined as patients with mean HBP shortening of LVAT-95 by >10 ms but with mean HBP LVAT-95 above the upper limit of normal defined by the 99% range for LVAT-95 in patients with normal QRS. The unpaired t test was used to compare electrical parameters between groups, with paired t tests for within-patient comparisons.

An ordinal regression model predicting the HBP LVAT (LVAT-95)
and including intrinsic LVAT-95 was used to assess the impact of propagation pattern on the change in LVAT95. To present the data, the results of the ordinal regression model were transformed to produce LVAT-95 times. Analyses were performed using the statistical environment "R," using the package "rms." 16,17 3 | RESULTS Forty-six subjects were recruited, 25 with LBBB and left ventricular impairment, and 21 with narrow QRS and normal ventricular function. Due to technical error with the ECGI system, data was unavailable during HBP in one patient in the LBBB group. Therefore 24 patients with LBBB undergoing CRT device insertion were included alongside 21 subjects with normal ventricular function and narrow QRS. All 24 patients with LBBB underwent HBP and 20 had BVP ECGI recordings (due to unsuitable coronary sinus anatomy in four patients). Baseline demographics for both groups of patients are displayed in Table 1 and the flowchart for inclusion is found in Figure S1.

| Activation times
The activation times in intrinsic rhythm, HBP, and BVP are displayed in Table 2. Figure

| Activation patterns
Visual analysis of ventricular propagation maps identified revealed two distinct patterns of LV activation during intrinsic LBBB. In the majority of patients (20/24, 83.3%), there were appearances of regions of epicardial propagation block, which we term "propagation discontinuities." In most of these patients (17/20, 85%) propagation discontinuities manifested as distinct "lines of block," where wavefront propagation appeared to halt simultaneously along a line. Beyond the line of discontinuity, epicardium appeared to be activated from a different direction than the original propagation wavefront.
This was not necessarily the opposite direction: the wavefronts could be orthogonal resulting in the appearance of "U-shaped" activation.
In the remaining 15%, epicardial propagation discontinuity was instead observed as "zones of slow conduction," where epicardium beyond the line of block appeared to be activated, after a delay, in the same direction that the original wavefront was traveling in. Two patients displayed more than one line of discontinuity (anterior and posterior Note: Values are mean ± SD (range) or n (%). In one patient of patients with left bundle branch block undergoing His bundle pacing, the ejection fraction measured at referral was higher than 35%. One patient in the group of patients with a narrow intrinsic QRS had an ejection fraction of 49%, which is in the mildly impaired category but the assessment clinically was low-normal function. a n = 21.
ARNOLD ET AL.

| DISCUSSION
In this study we demonstrate the novelty and utility of analysing activation patterns and times of His-CRT for LBBB in several ways: (1) EPM is a novel tool for noninvasively assessing ventricular activation patterns.
(2) The activation pattern in LBBB can identify patients in whom HBP is likely to result in ventricular resynchronization.
(3) Noninvasive electrical mapping can quantify ventricular resynchronization and identify mechanisms of incomplete resynchronization.
(4) Successful His-CRT results in normalization of ventricular activation with a more physiological activation pattern than BVP.

| LBBB activation patterns
We used noninvasive mapping to identify two distinct types of the ventricular activation patterns in patients with 12-lead ECG appearances of LBBB: those with the appearance of regions of propagation block, which we term "propagation discontinuity" and those with slow propagation without regions of discontinuity. Our findings are consistent with those from previous studies investigating ventricular activation patterns during LBBB, using both endocardial and epicardial mapping. 18

| Conduction system block versus propagation discontinuity
It is important to conceptually and anatomically distinguish disease Propagation discontinuity is not observed on direct endocardial contact mapping, 18 indicating the specific pattern of slowed or blocked propagation may be an artifact produced by the ECGI system's inverse solution 22 (which is also supported by in silico modeling 23 ) or an artifact arising from epi-endo anisotropy.
ARNOLD ET AL.
| 435 Importantly our study shows that these block artifacts correspond to biological phenomena due to their ability to identify patients amenable to resynchronization. We term the appearance discontinuity, rather than block, to differentiate the appearance we observed, that might be artifactual, from the specific electrophysiological definition of "block."

| Clinical utility of ECGI derived measures for His-CRT
Regardless of their pathophysiological basis, ECGI-derived activation patterns in LBBB allow accurate, noninvasive prediction of patients whose LVAT-95 is shortened by HBP. We have previously demonstrated that LVAT-95 shortening is a key mechanism through which successful His-CRT improves cardiac output, with incremental activation time shortening correlating with hemodynamic improvement. 3 Although left-sided septal conduction system mapping also appears to be a powerful predictor of HBP response, 7 the invasive nature of the technique precludes routine clinical use. The baseline 12-lead QRSd is less reliable at predicting HBP resynchronization and EPM provides superior predictive ability to 12-lead QRS morphology analysis: the Strauss criteria's predictive ability trended toward significance in this study but the degree of resynchronization predicted with this was much lower than with ECGI appearance of propagation block. Our study provides evidence that it is technically feasible to improve patient selection for His-CRT using noninvasive propagation mapping. We have also shown that the predictive feature of block artifact can be observed even when a different patient's anatomy is used to analyse propagation. Therefore it is likely that simpler noninvasive mapping methods which require fewer electrodes and eliminating the need for CT imaging could be develop, facilitating widespread use in clinical practice. 24,25 LVAT-95 also provides an intra-procedural target with complete His-CRT representing activation normalization that is not affected by selectivity of His bundle capture and with superior signal-to-noise ratio than QRSd. QRSd is affected by capture selectivity ( Figure S2)  12 This is not surprising as the activation pattern of "true" LBBB with posterolateral late activation, due to anterior propagation block, will also be more amenable to coronary sinus leads.
The HIS-SYNC randomized pilot evaluation of His-CRT vs BVP for heart failure with LBBB did not demonstrate the superiority of His-CRT over BVP. 8 The authors suggest that a 50% crossover from the His-CRT arm to the BVP arm was due to two factors: a current lack of reliably successful HBP tools and, crucially, patient selection.
It was felt that many patients enrolled who did not successfully resynchronize with HBP actually had a form of nonspecific intraventricular conduction delay (IVCD) rather than LBBB and even in those with ECG-diagnosed LBBB, conduction system block may not have been the mechanism of LBBB as this could not be elucidated from the 12-lead ECG. These patients crossed over the BVP arm as they could not be resynchronized.

| Partial and failed correction of LBBB
In three patients HBP did not improve LVAT-95 by at least 10 ms "failed His-CRT" and 13 patients reduced LVAT by more than 10 ms but did not achieve LVAT-95 within the normal range defined in this study. We have previously demonstrated that even when no resynchronization occurs there is still scope for hemodynamic improvement through AV optimization. 26 In patients with partial HBP correction of LBBB (incomplete His-CRT), the within-patient incremental LVAT-95 improvement with HBP over BVP was 17 ms, suggesting that even incomplete His-CRT, without activation normalization, produces superior electrical resynchronization to BVP. Propagation mapping also allows us to postulate mechanisms incompleteness of His-CRT in this group. Where lines of discontinuity in regions, whose conduction system supply is the anterior fascicle, resolve with HBP but posterior portions are left intact we can infer that selective recruitment of the anterior fascicle has occurred. In such cases, propagation mapping can guide the operator to attempt pacing more distally to attempt posterior fascicle capture either in the distal His bundle or the left bundle area.

| Epicardial propagation mapping
This study introduces EPM as a novel method for noninvasively assessing activation patterns. Conventional activation maps that annotate an activation time at most negative dv/dt are prone to potential mis-annotation of activation when electrodes' EGMs show multiple waves. This occurs in bundle branch block but EPM visualizes the entire EGM so that human analysis can determine the true wavefront. This provides a powerful tool for accurately depicting patterns of activation that makes use of all electrical information acquired by ECGI systems.

| Limitations
ECGI-derived measures provide information on epicardial activation; the septum, where the conduction system lesion is likely to be located, is thus not analysed. However, LVAT-95 measures the downstream effect of conduction system behavior in the septum on the LV myocardium overall. Furthermore, our findings are consistent with those produced by septal mapping of the conduction system and LVAT-95 correlates with hemodynamic outcomes. 3 Nevertheless, there are other limitations of the ECGI methodology including the assumption of static geometry and validity of particular solution to the inverse problem employed by ECGI. The hemodynamic changes predicted by LVAT-95 are acute and longitudinal evidence is required to assess the long-term clinical correlates of LVAT-95. As discussed above, it has been suggested that ECGI appearances of lines of discontinuity may in fact be artifactual due to the methodology of ECGI. 27 However, the elimination of lines of discontinuity correlate with shorter activation time suggesting biological electrophysiological phenomena. Although the appearance of lines of discontinuity are stark, they currently require human visual interpretation and thus may be subject to a degree of inter-and intra-rater irreproducibility. Operator experience can affect the success of His-CRT to confound prediction of successful His-CRT, however, by using very high output temporary HBP (up to 25 mA) we maximized the chance of correcting LBBB even if the position was not optimal.

| CONCLUSION
EPM derived from ECGI allows accurate noninvasive discrimination of LBBB with regions of propagation discontinuity that is potentially amenable to resynchronization by HBP from diffuse slow conduction that cannot be corrected. When HBP normalizes LVAT, the activation pattern produced is physiological and indistinguishable from normal activation with intact conduction system.