The role of guidance in delivering cardiac resynchronization therapy: A systematic review and network meta-analysis

Background Positioning the left ventricular lead at the optimal myocardial segment has been proposed to improve cardiac resynchronization therapy (CRT) response. Objectives We performed a systematic review and network meta-analysis evaluating echocardiographic and clinical response delivered with different guidance modalities compared to conventional fluoroscopic positioning. Methods Randomized trials with ≥6 months follow-up comparing any combination of imaging, electrical, hemodynamic, or fluoroscopic guidance were included. Imaging modalities were split whether one modality was used: cardiac magnetic resonance (CMR), speckle-tracking echocardiography (STE), single-photon emission computed tomography, cardiac computed tomography (CT), or a combination of these, defined as “multimodality imaging.” Results Twelve studies were included (n = 1864). Pair-wise meta-analysis resulted in significant odds of reduction in left ventricular end-systolic volume (LVESV) >15% (odds ratio [OR] 1.50, 95% confidence interval [CI] 1.05–2.13, P = .025) and absolute reduction in LVESV (standardized mean difference [SMD] -0.25, 95% CI -0.43 to -0.08, P = .005) with guidance. CMR (OR 55.3, 95% CI 4.7–656.9, P = .002), electrical (OR 17.0, 95% CI 2.9–100, P = .002), multimodality imaging (OR 4.47, 95% CI 1.36–14.7, P = .014), and hemodynamic guidance (OR 1.29–28.0, P = .02) were significant in reducing LVESV >15%. Only STE demonstrated a significant reduction in absolute LVESV (SMD -0.38, 95% CI -0.68 to -0.09, P = .011]. CMR had the highest probability of improving clinical response (OR 17.9, 95% CI 5.14–62.5, P < .001). Conclusion Overall, guidance improves CRT outcomes. STE and multimodality imaging provided the most reliable evidence of efficacy. Wide CIs observed for results of CMR guidance suggest more powered studies are required before a clear ranking is possible.


Introduction
Cardiac resynchronization therapy (CRT) is an effective treatment for patients with heart failure and electrical dyssynchrony characterized by left bundle branch block; however between 30% and 50% fail to derive benefit. CRT nonresponse is multifactorial; however, placement of the left ventricle (LV) pacing lead in the optimal position is considered integral to this. 1 Although it is established that an apical position of the LV lead should be avoided, the best location across the LV axis is less certain. 2 Evidence has demonstrated that a position away from scar 3 at the point of latest mechanical activation (LMA) may significantly determine response. 4 The targeting of the optimal LV lead position through a variety of guidance techniques has been investigated; however, with the increasing use of image-fusion technology 5 and number of randomized trials being undertaken with more advanced imaging techniques, 6 the need for an updated evaluation of the data is needed. A recently published meta-analysis by Hu and colleagues 7 was limited to evaluating only image guidance and included both randomized and nonrandomized studies. This was limited by heterogenous study designs and excluded non-image-based guidance techniques. 7 A network meta-analysis (NMA) was deemed appropriate to allow comparative assessment of different guidance modalities and potentially direct clinicians to which guidance modalities derived most benefit for patients.

Literature search
The systematic review and meta-analysis were conducted in accordance with the PRISMA guidelines (Supplemental Table S1). 8 We systematically reviewed the relevant literature by searching EMBASE, CENTRAL, and MEDLINE databases from inception to June 2021 without language restriction. The Quality of Reporting of Meta-Analyses statement 9 and the empiric study by McAuley and colleagues 10 indicate the exclusion of unpublished studies produces a systematic positive bias, and therefore "gray literature" in the form of poster presentations, unpublished data from Cochrane Reviews or other meta-analyses, conference abstracts, and preprints were included. In addition, references of relevant literature were searched. The following keywords were used for search: (("guide") OR ("guided") OR ("guidance")) AND (("cardiac resynchronisation therapy") OR ("cardiac resynchronization therapy") OR ("LV lead") OR ("left ventricular lead")).

Selection criteria
We included all eligible randomized studies that met the following inclusion criteria: (1) CRT-pacemaker/defibrillator (CRT-P/D), (2) heart failure with a QRSd .120 ms and LV ejection fraction ,35%, (3) human studies only, and (4) minimum of 6 months mean follow-up. For each study, the following efficacy endpoints were evaluated: (1) echocardiographic volumetric response (change in left ventricular endsystolic volume [LVESV]) and (2) symptomatic response (as heterogenous reporting of this was expected, a change in Clinical Composite Score 1 or NYHA class 1 represented response). Change in LV ejection fraction was not included as an efficacy endpoint, as variable reporting of this outcome was anticipated.

Data extraction
All data from included studies were independently extracted and assessed for further analysis by 2 reviewers (N.W. and V.S.M.). Any discrepancies were resolved through a third reviewer (M.K.E.). From each study, relevant information was extracted and tabulated. Study characteristics are reported. Modalities were split by whether only 1 imaging modality was used in the form of cardiac magnetic resonance (CMR), speckle-tracking echocardiography (STE), singlephoton emission computed tomography (SPECT), cardiac computed tomography, or a combination of these, which was defined as "multimodality imaging." Additional modalities evaluated included acute hemodynamic response (AHR) and electrophysiological guidance.

Pairwise meta-analysis
A pairwise meta-analysis was performed in studies comparing an advanced guidance technique to fluoroscopic guidance (standard of care). Intention-to-treat data were used for evaluating endpoints from included studies whenever possible. The 95% confidence intervals (CIs), odds ratios (OR), and standardized mean difference (SMD) were computed for categorical and continuous variables, respectively. To estimate prespecified efficacy endpoints of continuous data, only those publications that contained both baseline and follow-up means 6 standard deviations were used. A random-effects meta-analytical approach was applied to all analyses. Heterogeneity was considered low, moderate, and high for I 2 values of ,30%, 30%-60%, and .60%, respectively. 11 Subgroup analyses were performed according to the type of guidance method. All analyses were performed using R version 1.3.1093 with the "metafor" package. 12 Network meta-analysis The NMA was performed using the generic inverse variance method with the "netmeta" statistical package in R version 1.3.1093. 13 Random-effects meta-analysis was reported, and inconsistency was evaluated with Cochran's Q. 12 Inconsistency between direct and indirect estimates was checked using "node splitting." 14 A significance level of .05 and CIs were used for testing, and all testing was 2-tailed. Rank scores with probability ranks of different treatment groups were calculated. 15 These statistics are used to measure the extent of certainty that a treatment is better than another treatment (ranks closest to 1, being best), averaged over all competing treatments. Risk of bias for the individual studies was performed with the Cochrane Risk of Bias 2 tool. 16 Publication bias was assessed by funnel plot and Egger's test for each network analysis where 10 studies were included. The robustness of the inclusion of different patient subgroups and "grey literature" were tested by a sensitivity analysis.

Results
A total of 1458 unique records were identified through the searches. After screening, 12 met inclusion criteria (Supplemental Figure S1). In total, 1977 patients were enrolled in the studies, with 1864 patients included in the final follow-up analyses (guidance arm, n 5 1096;

KEY FINDINGS
-Overall guidance and accurately placing the left ventricular lead in the optimal myocardial segment results in improved response to cardiac resynchronization therapy.
-Speckle-tracking echocardiography and multimodality imaging provided the most reliable evidence of efficacy in improving cardiac resynchronization therapy response. Cardiovascular magnetic resonance guidance was the most efficacious; however, wide confidence intervals and indirect evidence suggest approaching this with caution.
-Ranking superiority of guidance modalities remains difficult, and more appropriately powered studies are required.    Mean follow-up ranged from 6 to 24 months. Graphical risk-of-bias assessment is reported (Supplemental Figure S2). Relevant funnel plot assessing for publication bias in shown in Supplemental Figure S3.

Study characteristics
Baseline demographics of included studies are summarized in Table 1. Three studies 17,18 employed STE to identify the LMA and surrogates from circumferential strain imaging of viable myocardium in a mixed etiology group. [17][18][19] All multimodality imaging studies used STE to identify LMA alongside additional imaging modalities to identify scar, with 3 studies using computed tomography to identify coronary sinus (CS) anatomy. [20][21][22] Stephansen and colleagues 22 compared multimodality imaging to local LV electrical delay by measuring intrinsic left ventricular (LV) electric delay (QLV) in the basal, mid, and apical positions to identify the optimal LV lead position. Cannizzaro and colleagues 23 used STE to identify LMA and scar transmurality derived from CMR imaging in an ischemic population. Ko ckov a and colleagues 24 used CMR to identify LMA using Tagging FLASH sequences 25 outside of .50% scar and compared this to optimal site by QLV derivation. Zou and colleagues 26 used SPECT to measure dyssynchrony using phase polar maps and myocardial perfusion uptake as a surrogate of viable myocardium. The optimal position was displayed using 3D LV surfaces alongside fluoroscopic images intraprocedurally. Singh and colleagues 27 used QLV to guide LV lead placement in patients with right bundle branch block only. Sipal and colleagues 28 aimed to implant the LV lead at the site with the narrowest biventricular-paced QRSd interprocedurally using surface electrocardiography. Finally, Sohal and colleagues 29 identified the CS branch with the greatest AHR as derived by LV dP/dT max as the optimal lead position. Baseline characteristics of study participants are reported in Table 2.

Pairwise meta-analysis
Pairwise meta-analysis ( Figure 1) evaluated studies comparing advanced guidance to fluoroscopic positioning. There was no significant improvement using STE with respect to LVESV .15%

Network meta-analysis
Change in LVESV .15% The evidence network is illustrated in Figure 2A and p-score ranking in Table 3 were similarly on a direction favoring guidance ( Figure 3A). Overall heterogeneity was high (I 2 5 82%).

Change in absolute LVESV
The evidence network is illustrated in Figure 2B and p-score ranking in Table 3. Compared to fluoroscopic guidance, only   Figure S4A.

Clinical response
The evidence network is illustrated in Figure 1C and p-score ranking in Table 3. Compared to fluoroscopic guidance, CMR (OR 17.9, 95% CI 5.14-62. 5 Figure S4B.

Sensitivity analyses
Sensitivity analyses (Supplemental Figures S5 and S6) were performed excluding studies that compared specific and mixed etiology of heart failure patient populations; these were (1) non-left bundle branch block QRS morphology and (2) ischemic etiology only. Excluding the abstracts by Cannizzaro and colleagues and by Glikson and colleagues resulted in no differences in the odds ratios and significance of results with regard to the prespecified endpoints. As there was only 1 relatively small study evaluating CMR only, 24 a sensitivity analysis was performed to test the stability of the results excluding this CMR-only study. The full results of the sensitivity analysis (Supplemental "sensitivity analysis") support the inclusion of these specific groups in the meta-analysis.

Discussion
Our main findings are as follows: (1) Overall guidance and accurately placing the LV lead in the optimal myocardial segment results in improved response to CRT. (2) STE and multimodality imaging provided the most reliable evidence of efficacy in improving CRT response. CMR guidance was the most efficacious; however, wide confidence intervals and indirect evidence suggest approaching this evidence with

Type of guidance modality
One of the major determinants of CRT response is LV lead position with respect to myocardial scar distribution. 30 This has the added advantage of potential reduced arrhythmogenicity. 31 Another major determinant of response is placement of the LV lead at the latest electrically activated area-a surrogate of which is LMA.
The NMA identified that all guidance modalities were more efficacious than fluoroscopic guidance, with CMR guidance being most efficacious; however, the CMR evidence should be interpreted with caution, as it involved 1 study of 99 patients. 32 AHR guidance also significantly improved response; however, this was from a single study, 29 albeit with a higher number of patients (n 5 281), and so can be interpreted with more certainty. Multimodality imaging significantly improved response and included 4 studies of 487 patients with narrower CIs, suggesting a greater certainty of the validity of these results.
Electrical guidance was not effective in the clinical response marker; however, this may be because most participants in these studies were from the ENHANCE-CRT study, 27 which only included patients with right bundle branch block. STE was the only modality that was able to detect a significant absolute reduction in LVESV, which may be representative of the largest number of participants in the studies using this guidance modality (n 5 322). In summary, the strongest evidence for guidance came from STE and multimodality imaging.
Overall, use of the NMA technique allowed to obtain mixed estimates. As consistency was found between direct and mixed estimates results, our guidance modalities' effect size assessment is robust (Supplemental Figure S4).

Ischemic vs nonischemic etiology
Cannizzaro and colleagues 23 and Glikson and colleagues 19 evaluated the role of guidance in an exclusively ischemic population, Cannizzaro finding a significant improvement in CRT response as defined by a reduction in LVESV .15% (73% vs 52%, P 5 .045), whereas Glikson identified no significant difference (48% vs 53%, P 5 NS). In comparison, the RADI-CRT study demonstrated greater reverse remodeling in an ischemic subgroup in the pressure-wire guided arm (69% vs 49%, P 5 .02), but not in the nonischemic subgroup (81% vs 71%, P 5 .19). The TARGET study did not demonstrate significant reverse remodeling on multivariable analysis in an ischemic population (OR 1.54, 95% CI 0.69-3.43, P 5 .293); however, absence of scar at the LV lead pacing site did increase likelihood of response (OR 3.06, 95% CI 1.01-9.26, P 5 .048). 18 This suggests an ischemic population may have the most benefit from avoiding scar and therefore from guidance use. The sensitivity analysis whereby the ischemic-only studies were excluded (Supplemental Figure S6B) suggests that ischemic patients presented a disproportionately greater challenge to improved outcomes.

Accuracy of final LV lead location
Borgquist and colleagues 20 evaluated outcomes based on whether the final LV lead was concordant to the optimal cardiac segment or in the adjacent segment and compared this to those with an LV lead in a distant segment. It was found that death or heart failure hospitalization was more likely in the distant LV lead group (P 5 .008). 20 Saba and colleagues 17 observed that pacing at optimal LV sites conferred significantly better CRT-D therapy-free survival rate compared to pacing remotely to the optimal site (HR 5 0.51, 95% CI 0.28-0.90, P 5 .018). Khan

Clinical and future perspective
These clinical trials have demonstrated that integrating guidance into CRT implantation is feasible; however, there continues to be reluctance to integrate guidance routuinely. 1 Potential reasons include the additional cost, investigations, expertise, and equipment required. A proportion of patients may have poor image quality, or may not have the required coronary venous anatomy to reach the desired segment. [33][34][35] This NMA has demonstrated that integrating STE is effective in detecting a significant absolute reduction in LVESV. As echocardiography is part of the minimum dataset required pre-CRT, 33 this could be fundamental to image guidance without burdensome additional equipment.
It must be noted that final LV lead position identification was based on fluoroscopy images in the studies, which is known to be poorly reproducible. 36 More advanced fusion image-based guidance systems are currently used in an investigational setting and have the potential to identify the target LV segment more precisely in real time during implantation. 5 Such increased precision may derive further benefit from guidance. Large, ongoing randomized trials will provide greater insights into this technology's effectiveness. 37

Limitations
The current NMA has important limitations. There were multiple different measures of CRT outcome and follow-up duration was not consistent across studies; therefore, the most common outcome markers were evaluated. Patients and outcome assessors were not uniformly masked to whether they were in the intervention group, which may introduce treatment and observer bias. Not all studies were 1:1 randomized. In some of the network arms there was 1 study with smaller numbers of patients included, notably evaluating CMR only.
This accounts for the wide confidence intervals observed and suggests further clinical trials will increase confidence of their efficacy. Specific patient populations were only recruited in some of the studies, which may reduce the reproducibility of these results in a general dyssynchronous heart failure population. Two studies were not fully published results; however, the inclusion of "gray literature" was to avoid selection and publication bias. Sensitivity analyses were performed to mitigate this risk, and these results demonstrate the robustness of the inclusion of these studies. Only 4 studies specified what proportion of quadripolar or bipolar lead was implanted; 3 used both, with no significant differences between the guidance and fluoroscopic group. 20,22,29 In addition, only 5 studies identified type of programming specified postimplant and whether the device was optimized. The lack of consistency in programming and lead technology may affect the interpretation of these results. Finally, the data used were derived data published by the study authors, and therefore patient-level data were not used for the meta-analysis.

Conclusion
This comprehensive analysis suggests that overall, guiding the LV led to the optimal myocardial segment results in better CRT response. Further evidence in the form of large, randomized studies will allow a more nuanced evaluation of which modality is best placed to guide optimal LV lead delivery, particularly in advanced imaging modalities such as CMR. Easily accessible, reproducible, and interpretable techniques are essential for widespread integration of guidance into routine clinical care.
Funding Sources: The study was supported by the WellcomeWellcome/ EPSRC Centre for Medical Engineering (WT203148/Z/16/Z).
Disclosures: The department is supported by the Wellcome/EPSRC Centre for Medical Engineering (WT203148/Z/16/Z). M.K.E., V.S.M., and J.G. have received fellowship funding from Abbott. C.A.R. receives research funding and/or consultation fees from Abbott, Medtronic, Boston Scientific, Spectranetics, and MicroPort outside of the submitted work. S.A. was funded/supported by the NIHR Biomedical Research Centre based at GSTT and KCL. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health.
Authorship: All authors attest they meet the current ICMJE criteria for authorship.
Ethics Statement: The database collection and analysis were approved by the Institutional Review Board of Guy's and St Thomas' Hospital. The research in this study was conducted in accordance with the Declaration of Helsinki.