Effectiveness of conduction system pacing for cardiac resynchronization therapy: A systematic review and network meta‐analysis

Cardiac resynchronization therapy (CRT) with biventricular pacing (BiV‐CRT) is ineffective in approximately one‐third of patients. CRT with Conduction system pacing (CSP‐CRT) may achieve greater synchronization. We aimed to assess the effectiveness of CRT with His pacing (His‐CRT) or left bundle branch pacing (LBB‐CRT) in lieu of biventricular CRT.

lower pacing thresholds.Future randomized trials are needed to determine CSP-CRT efficacy.

K E Y W O R D S
biventricular pacing, cardiac resynchronization therapy, conduction system pacing, His pacing, left bundle branch pacing, network meta-analysis

| INTRODUCTION
Cardiac resynchronization therapy (CRT) with biventricular pacing (BVP) is an important treatment for patients with wide QRS and heart failure; however, approximately one third of patients who receive biventricular CRT (BiV-CRT) do not demonstrate a favorable response. 1,2New approaches to cardiac pacing which engage the native His-Purkinje system-conduction system pacing (CSP)-may help achieve resynchronization and improve clinical outcomes in candidates of CRT, since CRT with both His-pacing (His-CRT) and left bundle branch (LBB) pacing (LBB-CRT) can theoretically restore electromechanical synchrony. 3 date, studies investigating CRT via CSP (CSP-CRT) have been limited by the small sample sizes, and heterogeneity in their findings, [4][5][6][7][8] and the clinical evidence has lagged behind the enthusiasm for implementing these novel techniques for CRT.Herein, we report a systematic review and network meta-analysis of studies that report outcomes in patients undergoing CSP-CRT with either His-CRT or LBB-CRT.Since His-CRT and LBB-CRT are distinct approaches, a network meta-analysis is the optimal approach to compare outcomes between these novel approaches and against the conventional BiV-CRT control group.

| Design, search, and study selection
The review protocol is available on PROSPERO (CRD42022328042).
The reporting conforms to the Preferred Reporting Items for Systematic Reviews (PRISMA). 9,10Since this study used data from other publications, it was exempted from an additional institutional review board and ethics committee approval.
A systematic search was conducted in PubMed, Embase, Web of Science, Scopus, and the Cochrane Library from database inception until August 19, 2023.Details of search keywords are presented in Supporting Information: Supplemental Methods.Detection, screening, and removal of duplicate records, and then title/abstract screening was performed with the Rayyan web application (Rayyan Systems, Inc.). 11 each stage of review, original studies-randomized trials or observational-were selected if they had the following eligibility criteria: (1) focused on a population of patients with reduced left ventricular ejection fraction (LVEF) who had indications for CRT; (2)   investigated the intervention of CSP in the form of his or LBB pacing, whether in comparison to BiV-CRT or as a single group; and (3)   reported at least one outcome of interest in the CSP group, including CSP success rate, LVEF, QRS duration (QRSd), New York Heart Association (NYHA) classification, death, or hospitalization.Non-English publications were excluded.

| Review and data extraction
Two reviewers (HT and SK) independently evaluated the retrieved full texts for eligibility, extracted data, and assessed risk-of-bias.Discrepancies were resolved by discussion with a third author (AB).The study publication year, country, design, number, age, sex, reported study eligibility criteria, and study indications for CRT were recorded.
Characteristics including baseline QRS morphology and rhythm, QRSd, LVEF, NYHA, left ventricular end-systolic volume (LVESV), pacing thresholds after implantation and at follow-up, clinical events, etiology of cardiomyopathy, and device type were extracted.For continuous data, the number, mean, and standard deviation for data at baseline and followup/post-implant (for QRSd, and NYHA classification, thresholds), or for changes from baseline (follow-up minus baseline or Δ, for LVEF and LVESV) were entered into data sheets.Extracted binary data included CSP success rate, clinical response (defined based on NYHA class), death, and hospitalization.

| Risk of bias
For randomized studies, the second version of the Cochrane risk-ofbias tool for randomized trials (RoB-2) was used, whereas the Methodological Index for Non-Randomized Studies (MINORS) was considered for observational studies. 12,13The RoB-2 evaluates studies in five domains and grades each as "low risk," "some concerns," or "high risk."The MINORS includes 12 items graded as 0 (not reported), 1 (reported, but inadequately), or 2 (reported and adequate).Four items are specific to comparative studies; therefore, the maximum scores for comparative and non-comparative studies would be 24 and 16, respectively.

| Data synthesis
All analyses were conducted using R version 4.1.3,and the packages "meta," "metafor," and "netmeta."For continuous effect measures, | 2343 the number of cases, mean, and standard deviations were extracted.
For studies that did not report standard deviations for change from baseline (for ΔLVEF and ΔLVESV), these values were estimated using the methods from Cochrane Handbook for Systematic Reviews of Interventions, version 6.3. 14The means were pooled in the CSP and BVP groups to acquire estimates for each outcome measure.
Between-group mean difference (MD) was used to compare LVEF, LVESV, and QRSd between CSP and BVP.Hedges' g Standardized MD (SMD) was used for the comparison of pacing thresholds (due to differences in measurements and slightly variable pulse widths) and the NYHA classification (due to subjectivity of its assessment and expected variation in measurements).The proportions of successful CSP implantations were pooled using generalized linear mixedeffects model with logit transformations.For binary outcomes (Clinical response, death, and hospitalization), the relative benefit/ relative risk were calculated and pooled using the Mantel-Haenszel method.The Higgins and Thompson's I 2 -statistic and the betweenstudy variance in random-effects models (τ 2 ) were used to measure statistical heterogeneity.To calculate the heterogeneity variance τ 2 , the restricted maximum likelihood estimator and the Sidik-Jonkman estimator were used for continuous and binary effect sizes, respectively. 15,16Since considerable between-study heterogeneity was anticipated, all analyses were conducted with a random-effects model.To compare the outcomes of His-CRT or LBB-CRT with the control group of BiV-CRT, the frequentist network meta-analysis was used with the netmeta function.Consistency in the networks was evaluated by node splitting, looking for agreement between direct and indirect evidence. 17Comparison-adjusted network funnel plots were visually inspected for symmetry and the Egger's test was applied to investigate the risk of publication bias.

| Search results and study characteristics
The systematic search identified 37 eligible publications, [4][5][6][7][8] including 20 comparative studies-among which four were randomized controlled trials-and 17 single-arm investigations reporting outcomes in patients undergoing CSP for CRT (details shown in Figure 1). The corative studies enabled network meta-analyses of the three interventions-His-CRT, LBB-CRT, and BiV-CRT.The single-arm data were used to pool outcome measures in patients undergoing CSP.Eight studies presented patient-level data for outcomes.25,26,29,32,40,45,47,48 Durations of follow-up ranged between 5 and 31 months, with most studies (25/37) following patients between 6 and 12 months.Notably, the most common etiology for cardiomyopathy was nonischemic (64.5% of study populations). 42,48

| Network consistency
Node splitting analysis of the networks did not reveal evidence of statistical inconsistency (Supporting Infomation: Table S2), except for one comparison between LBB-CRT and BiV-CRT for QRSd (p = .007).
This was caused by one three-arm study; 39 thus, the analysis of QRSd was repeated after its exclusion, which produced similar findings

| Evidence from randomized trials
Four randomized controlled trials were included. 4,5,7,8The main limitations of these trials were small samples, and the high rate of cross-overs.Using the intention-to-treat data from these studies, we ran the analyses for each outcome, whenever such data was available.LVEF improvement was statistically higher (MD: +2.5%

| Publication bias
We found evidence of publication bias for LVEF outcome (Egger's test p = .002).There was no evidence for publication bias for other T A B L E 2 (Continued) outcomes after inspection of funnel plots and applying Egger's test (Supporting Infomation: Figures S31-S36).

| DISCUSSION
This study demonstrates that CSP-CRT can be implemented with a high success rate, and may achieve superior improvements in LV The statistically significant benefit of improvement in LVEF and reduction in LVESV with both His-CRT and LBB-CRT should be interpreted with the magnitude of changes in mind.It can be argued that a difference in ΔLVEF of 5% or less compared to BiV-CRT may not be clinically meaningful; however, it should be noted that these improvements were observed with relatively short follow-ups, and the benefits may increase with time.Such a trend has been observed in CSP-CRT studies, 5,23,49 as well as in the seminal trials of BiV-CRT versus medical therapy, 50,51 where LVEF improvements became more prominent with longer follow-ups.In other words, it may not be the conduction system capture itself, but rather narrow-paced QRS, whether it is achieved by CSP-or BiV-CRT, that results in improved LVEF and clinical outcomes. 53It is crucial to consider that measurements of QRSd may be subjective and lack reliability and reproducibility.In addition, heterogeneity exists in the methods of measuring QRSd used by each study.Therefore, caution is advised in interpretation of QRSd results.
The reported procedural success rates of 91% for LBB-CRT and 73% for His-CRT are promising; nevertheless, the success rates showed significant heterogeneity among studies.Experience of the operators is perhaps an important determinant of implantation success, as CSP is shown to have a steep learning curve. 54,55I G U R E 3 Pooled mean differences in network meta-analyses with corresponding network graphs for (A) left ventricular ejection fraction (LVEF), (B) left ventricular end-systolic volume (LVESV), and (C) paced QRS duration (QRSd).
Registry data from European centers shows a success rate of 90% for left bundle branch area pacing; however, implantation for heart failure indications was associated with lower success rates of about 82%. 54On the other hand, the implantation success of BiV-CRT in clinical trial setting has been about 95%. 50,51,56In patients with heart failure, it is hypothesized that enlarged cardiac chambers or septal fibrosis may contribute to higher rates of LBB-CRT lead implantation failures. 54The lower success rates with His-CRT may be attributable to higher proportion of unacceptable pacing thresholds.Moreover, the probability of success for His-CRT to correct left bundle branch block is significantly lower when the block is more distal, while His-CRT, and maybe even LBB-CRT, fail to achieve resynchronization in patients with conduction defects due to intraventricular or intramyocardial disease. 3This latter group of conduction abnormalities may be more prevalent among patients with heart failure indicated for CRT.Notably, due to heterogenous reporting of studies, we could not investigate the rates of selective and nonselective Hisor LBB-pacing.In several studies of LBB-CRT, left bundle branch area pacing or left septal pacing could be considered a success.While implantation success rates are acceptable, the question of CSP-CRT durability is still unresolved.In this study, the numerically higher pacing thresholds with His-CRT may result in lower generator longevity.
In addition to His-and LBB-CRT, His-optimized and LBBoptimized pacing, which use a combination of previous methods to optimize resynchronization, are also available as CRT options, albeit with lower number of studies. 57,58Considering the diversity of clinical features in patients indicated for CRT, an ideal scenario would be to individualize CRT options in the future.Our study could not provide data about the effectiveness of CSP-CRT with regard to patients' characteristics.A meta-regression considering baseline LVEF, QRSd, and QRS morphology was considered but was not feasible due to the low number of studies that report outcomes in different subgroups.Notably, most participants in the CSP-CRT studies had nonischemic causes of heart failure.This may be due to a higher proportion of patients with ischemic cardiomyopathy having distal intramyocardial conduction disease, and would potentially derive more benefit from BiV-CRT rather than CSP-CRT.Another notable feature in our study was the higher number of publications investigating LBB-CRT compared to His-CRT, which marks a shift of interest towards the more novel LBB-CRT.Ongoing clinical trials of CSP-CRT will provide a better understanding of the efficacy, as well as tailored indications of these novel approaches. 59,60

| Limitations
First, most of the included studies use observational designs, which increases the risk of selection bias and unmeasured confounding.
Second, we observed a high level of heterogeneity in our metaanalyses, that could not be attributed to a heterogenous design in studies.Third, we could not report outcomes among different subgroups of patients since such data was not available from the included studies.

| CONCLUSION
The currently available evidence favors CSP-CRT as a feasible and effective treatment that achieves greater improvements in LV function, QRSd, and heart failure symptoms.Notably, LBB-CRT showed a higher clinical response rate and lower pacing thresholds than His-CRT.While this study has synthesized evidence supporting the effectiveness of CSP-CRT, the observational designs and relatively short follow-up durations of the included studies limit the robustness of conclusions.Future data from randomized controlled trials is needed to confirm or refute these findings.

F I G U R E 1
The PRISMA flow diagram.T A B L E 1 Study characteristics.
function and dimensions compared to conventional BiV-CRT.The degree of QRS narrowing, as a measure of electrical synchronization, was more favorable after CSP-CRT.Furthermore, patients undergoing CSP demonstrated better clinical response based on NYHA classification, and there was a signal for reduced hospitalizations in individuals receiving CSP-CRT compared to BiV-CRT.This systematic review has generated the largest data set to date for evaluating the effectiveness of CSP as a strategy of CRT.While clinical trials of CSP-CRT are awaited, this investigation highlights notable implications for clinical practice and future research.
Furthermore, observed reductions in LVESV indicate reverse LV remodeling, which supports the notion that electrical and mechanical resynchronization with CSP-CRT has been superior to BiV-CRT in these studies.Results of symptomatic improvements and clinical response, which were based F I G U R E 2 Forest plots for success rates of cardiac resynchronization therapy with His pacing (His-CRT), left bundle branch pacing (LBB-CRT), and biventricular pacing (BiV-CRT) with reported causes of failure.onNYHA classification, should be interpreted with caution due to lack of blinding in studies, the subjective nature of such outcomes, and the ambiguity of the minimal clinically important difference of these subjective changes.52Reports of other more objective measures of symptomatic burden, such as the 6-min walk test, were infrequent among included studies.Despite some inherent limitations, evidence of significant improvements in LVEF and NYHA in meta-analysis of intention-to-treat results of included randomized trials, corroborates the hypothesis of higher efficacy with CSP-CRT in select cases.The significantly greater QRS narrowing with CSP-CRT could have been crucial in achieving echocardiographic and clinical response, since QRSd, as a simple and routinely used measure of electrical resynchronization, determines both the indication and success of CRT.

F I G U R E 4
Pooled Hedges standardized mean differences in network meta-analyses with corresponding network graphs for (A) baseline post-implant pacing thresholds, (B) follow-up pacing thresholds, and (C) New York Heart Association (NYHA) class.

Table 2 .
Baseline characteristics of included patients.