Comparative effects of left bundle branch area pacing, His bundle pacing, biventricular pacing in patients requiring cardiac resynchronization therapy: A network meta‐analysis

Abstract Background The comparative effects of different types of cardiac resynchronization therapy (CRT) delivered by biventricular pacing (BVP), His bundle pacing (HBP), and left bundle branch area pacing (LBBAP) remain inconclusive. Hypothesis HBP and LBBAP may be advantageous over BVP for CRT. Methods PubMed, Embase, Web of Science, and the Cochrane Library were systematically searched for studies that reported the effects after BVP, HBP, and LBBAP for CRT. The effects between groups were compared by a frequentist random‐effects network meta‐analysis (NMA), by which the mean differences (MDs) and 95% confidence intervals (CIs) were calculated. Results Six articles involving 389 patients remained for the final meta‐analysis. The mean follow‐up of these studies was 8.03 ± 3.15 months. LBBAP resulted in a greater improvement in LVEF% (MD = 7.17, 95% CI = 4.31 to 10.04), followed by HBP (MD = 4.06, 95% CI = 1.09 to 7.03) compared with BVP. HBP resulted in a narrower QRS duration (MD = 31.58 ms, 95% CI = 12.75 to 50.40), followed by LBBAP (MD = 27.40 ms, 95% CI = 10.81 to 43.99) compared with BVP. No significant differences of changes in LVEF improvement and QRS narrowing were observed between LBBAP and HBP. The pacing threshold of LBBAP was significantly lower than those of BVP and HBP. Conclusion The NMA first found that LBBAP and HBP resulted in a greater LVEF improvement and a narrower QRS duration compared with BVP. Additionally, LBBAP resulted in similar clinical outcomes but with lower pacing thresholds, and may therefore offer advantages than does HBP for CRT.


| INTRODUCTION
Heart failure (HF) remains to be a serious public health concern, with high mortality, morbidity, and poor quality of life. 1 Cardiac resynchronization therapy (CRT) is effective for HF, particularly in those with reduced systolic heart failure and either bundle branch block (BBB) or need for frequent ventricular pacing. 2,3 The clinical benefits of CRT delivered by biventricular pacing (BVP) are remarkable. 4,5 However, approximately 30% of patients do not respond to CRT with BVP. In fact, BVP results in the fusion of two nonphysiological wave fronts and leaves a substantial degree of residual dyssynchrony. 6,7 Moreover, the success rate of BVP for CRT was about 95.9%, and the complication incidence was 7%-10%. 8,9 Computer modeling indicates that there would be more possibilities to improve cardiac functionality when greater ventricular resynchronization could be achieved. 10 His bundle pacing (HBP) has become a possible alternative for CRT with physiological restoration of normal physiologic His-Purkinje conduction and promoted superior electrical ventricular resynchronization than BVP. 6,11,12 Moreover, several other studies confirmed that HBP corrected the left bundle branch block (LBBB) by pacing the distal portions of the His bundle (His-CRT). 13,14 When successful, this contributes to a normality of LV electrical activation and thereby a more "physiological" correction of dyssynchrony.
Nevertheless, HBP has some shortcomings and limits its application, such as low R-wave amplitude, high pacing thresholds, and technical difficulties. 15 The average implant success rate of HBP was 84.8%, and the complication incidence was about 4.7%. 16 In 2017, Huang et al. 17  both the LBB and the local myocardium. 18 It is defined as left ventricular septal pacing (LVSP) if only LV septal myocardium is captured. 18 Left bundle branch area pacing (LBBAP), with the lead implanted slightly distal to the His bundle and screwed deep in the LV septum ideally to capture LBB, which means LBBP or LVSP, without clear evidence for LBB capture. 19 Subsequently, several case reports and observational studies demonstrated the effectiveness and safety of LBBAP in patients requiring CRT during short-and mid-term follow-up. [20][21][22][23] Furthermore, the success rate of LBBAP was reported varied from 90.9% to 97.8%, [24][25][26] and the overall complication incidence of procedure-related and long-term follow-up was about 1.6%-2.8%. 26,27 However, only a few studies compared the feasibility and effects of these different types of CRT delivered by BVP, HBP, and LBBAP, especially direct comparison between HBP and LBBAP. Thus, we aimed to systematically review the studies of BVP, HBP, and LBBAP for CRT to perform a network meta-analysis of existing data.

| Literature review and search strategy
All search results were assessed in accordance with the PRISMA guidelines. 28 A systematic literature search of PubMed, Embase, Web of Science, and the Cochrane Library was conducted to compare the following outcomes: changes in left ventricular ejection fraction (LVEF) between BVP, HBP, and LBBAP for CRT. Two investigators (Juan Hua and Chenxi Wang) conducted a systematic literature review independently. The search was performed with keywords as follows: "Cardiac resynchronization therapy," "Biventricular pacing," "His bundle pacing," and "Left bundle branch pacing" or "Left ventricular septal pacing" or "Left bundle branch area pacing," alone and in combination. The search strategies were shown in Table S1. screened the selected articles independently based on the title and abstract. Any discordance was settled through discussion between the reviewers. The pacing threshold was the His lead for HBP and LBBAP and LV lead for BVP at implantation or a week after implantation at 0.4, 0.5, or 1.0 ms.

| Data abstraction and quality assessment
Data from each enrolled article were independently extracted by two reviewers (Qijun Wang and Ziyi Xiong). Background information such as authors, years, the region of trial, indication, intervention, duration of follow-up, and outcomes were extracted from each article. All conflicts were resolved through discussion between the reviewers.
The Cochrane Handbook for Systematic Reviews of Interventions (version 5.4.0) was used to evaluate the quality of the selected RCTs. 29 Observational studies were evaluated using the Newcastle-Ottawa Scale. 30 Studies with six or more points were regarded as having a high quality.

| Statistical analyses
Network meta-analysis using a network analysis tool that combined direct and indirect evidence in a mixed-intervention model was performed. 31 For each interest outcome, the effect measurement estimated chosen for the continuous variables were the mean differences (MDs) and their corresponding 95% confidence intervals (CIs). The surface under the cumulative ranking area (SUCRA) probabilities were selected to calculate the ranking and hierarchy of the different treatments. 31 The larger SUCRA indicates the greater probability of becoming the best intervention. The network metaanalysis was conducted using the frequentist methods with restricted maximum likelihood estimation to quantify network heterogeneity and to assume a common heterogeneity estimate within a network.
In addition, the local inconsistencies across studies in each closed loop were evaluated using the node splitting approaches. 32 Publication bias was evaluated using the funnel plots. These analyses were conducted using the Stata version 15.0 (Stata Corp).

| Study selection and quality assessment
The literature search yielded 1471 articles (804 from PubMed, 478 from Web of Science, 115 from Embase, and 74 from the Cochrane Library), which were considered as potential studies. After the removal of duplicates, 679 publications remained. Afterward, 650 articles were excluded by preliminary screening of titles and abstracts, and a total of 29 were further evaluated comprehensively. Then, 22 records were excluded because of various reasons. Seven articles were found to be eligible for the present meta-analysis after a fulltext review. 11,21,23,[33][34][35][36] However, two included studies were from the same center, the patients were included from December 2012 to December 2018, 35 and January 2012 to June 2017, 11 respectively. It seems that some patients included were overlapped, so we determined to delete the study of Huang et al, which included relatively few patients. 11 Finally, six studies were selected for the present analysis. The selection process for the literature included in the analysis was shown in Figure 1. The basic characteristics, demographics of the study participants, and the quality evaluation of observational studies were shown in Table 1. The mean follow-up of these studies was 8.03 ± 3.15 months. RCTs were evaluated by Cochrane Handbook for Systematic Reviews of Interventions and shown in Figure S1. The network plots for comparisons of outcomes were shown in Figure S2.

| Exploration of inconsistency and publication bias
No inconsistency was found to be in both global ( Figure S3) and local tests (Tables S2-S4) in this NMA. No significant publication bias but changes in QRS duration was found in the funnel plot ( Figure S4).   The His-Purkinje system pacing is currently considered to be the optimal physiologic pacing technique, with the pacing lead directly implanted in the conduction system to narrow the QRS complex and improve cardiac function by selective or nonselective HBP. 11,42 Nevertheless, there are several limitations with HBP, which may restrict its wide clinical application, such as high corrective thresholds and late threshold increases. LBBAP, as a novel pacing technique, aims to correct the LBB conduction system desynchrony therefore deliver satisfactory LV synchrony and immediate hemodynamic benefits. 43 Meanwhile, it has a lower and stable pacing threshold and a physiological pacing site to prevent the occurrence of conduction disorders. 26 Furthermore, LBBAP is associated with high success rate and low complication incidence. 44 The QRS duration has been identified as a powerful prognostic marker, and its significance is well known in patients with heart failure. 45 A QRS complex ≥120 ms results in a more advanced myocardial disease, worse prognosis, and higher all-cause mortality. 46 The QRS duration is an established predictor of response to CRT, 47 and its changes from before to after pacing are also considered predictors of response to CRT. 48 The narrower the QRS duration, the higher is the degree of ventricular synchronization that can be obtained after pacing. In this study, we observed that HBP and LBBAP delivered a significantly narrower QRS duration compared with BVP. This may be explained by that HBP had the potential to capture the His-bundles and contribute to the most effective ventricular resynchronization. 13 Theoretically, LBBAP corrects the left bundle branch and leaves right bundle dyssynchronization, and may therefore have a longer QRS duration compared with HBP. However, our meta-analysis found that there were no differences between the HBP and LBBAP groups. This may be explained by the fact that the conduction velocity in the Purkinje fibers is so rapid that there is almost no difference in the QRS width after HBP and LBBAP. 49 Moreover, if the proximal LBBAP is performed, the paced QRS duration maybe not significantly longer compared with HBP. And the paced QRS duration can be further shortened during LBBAP by adjusting the AV delay or bipolar pacing to eliminate right bundle branch block pattern, which results in a nearly normal QRS complex. In contrast, BVP simply confers a mechanical synchronization rather than a physiologic synchronization;

| DISCUSSION
thus, it may not contribute to the full potential of CRT, 10 which is why the QRS duration was longer than that of LBBAP. To the best of our knowledge, the long-term results of the MADIT-CRT study (7 years) highlighted the lack of benefit of CRT in nonspecific intraventricular conduction delay (NICD) patients compared with patients with LBBB. 50 In the one included study of Upadhyay et al., 33 however, the crossover rates accounted for 50% from His-CRT to BVP-CRT due to NICD, which may affect the feasibility and outcomes between the HBP and BVP.
In theory, HBP and LBBAP confer physiologic pacing with a highly ventricular resynchronization; thus, there should be more LVEF improvement than BVP. We observed that all these re- In addition to the clinical benefits and electrical synchrony, pacing parameters were also important in pacing treatments, such as pacing threshold and impedance. The early studies found that the pacing threshold of LBBAP was significantly lower than that of HBP, 35,54 which was even up to 2.75 V/1.0 ms in some cases. 33 Our results were consistent with those of the previous studies that reported that HBP had a higher pacing threshold. The following reasons may explain these results. First, HB is covered with a fibrous sheath that is electrically nonconducting. Second, the HB is in a nondependent position, and orientation of the active fixation lead may influence the pacing thresholds. 55 Furthermore, myocardial fibrosis and degeneration occur in the pacing area, and the pacing threshold increases after HBP implantation. 56 Conversely, the LBBAP lead offers very low capture thresholds. This could be the result of a combination of factors. First, the LBB goes beneath the endocardium of the ventricular septum with a relatively large dimension and is surrounded by myocardium, thereby making it easier to capture. 49 Second, the LBB lead targets the precise area just beyond the site of the conduction block. 17,26 Collectively, our results showed that LBBAP can achieve a comparable LV electrical and mechanical synchrony to HBP but with a lower pacing threshold; therefore, it might be superior to HBP for CRT.

| Limitations
This meta-analysis has several potential limitations. First, the small sample size may therefore possibly contribute to an underestimation of the accuracy of this study. Second, the lack of uniform criteria for LBBAP may influence its actual effects.

CONFLICT OF INTERESTS
The authors declare no conflict of interest.