Risk of ventricular arrhythmia in cardiac resynchronization therapy responders and super-responders: a systematic review and meta-analysis

Abstract

Aims

Response to cardiac resynchronization therapy (CRT) is associated with improved survival, and reduction in heart failure hospitalization, and ventricular arrhythmia (VA) risk. However, the impact of CRT super-response [CRT-SR, increase in left ventricular ejection fraction (LVEF) to ≥ 50%] on VA remains unclear.

Methods and results

We undertook a meta-analysis aimed at determining the impact of CRT response and CRT-SR on risk of VA and all-cause mortality. Systematic search of PubMed, EMBASE, and Cochrane databases, identifying all relevant English articles published until 31 December 2019. A total of 34 studies (7605 patients for VA and 5874 patients for all-cause mortality) were retained for the meta-analysis. The pooled cumulative incidence of appropriate implantable cardioverter-defibrillator therapy for VA was significantly lower at 13.0% (4.5% per annum) in CRT-responders, vs. 29.0% (annualized rate of 10.0%) in CRT non-responders, relative risk (RR) 0.47 [95% confidence interval (CI) 0.39–0.56, P < 0.0001]; all-cause mortality 3.5% vs. 9.1% per annum, RR of 0.38 (95% CI 0.30–0.49, P < 0.0001). The pooled incidence of VA was significantly lower in CRT-SR compared with CRT non-super-responders (non-responders + responders) at 0.9% vs. 3.8% per annum, respectively, RR 0.22 (95% CI 0.12–0.40, P < 0.0001); as well as all-cause mortality at 2.0% vs. 4.3%, respectively, RR 0.47 (95% CI 0.33–0.66, P < 0.0001).

Conclusions

Cardiac resynchronization therapy super-responders have low absolute risk of VA and all-cause mortality. However, there remains a non-trivial residual absolute risk of these adverse outcomes in CRT responders. These findings suggest that among CRT responders, there may be a continued clinical benefit of defibrillators.

Introduction

Cardiac resynchronization therapy (CRT) whether pacemaker alone (CRTP) or with defibrillator (CRTD) is associated with recovery of left ventricular systolic function through positive remodelling, reduction in all-cause mortality, heart failure hospitalization, as well as improvements in quality of life, and functional capacity.¹ Traditionally, left ventricular systolic function recovery after CRT implantation has been categorized into no recovery or CRT non-response (LVEF ≤ 35%), partial recovery or CRT response (LVEF 36–49%), and full recovery or CRT super-response (CRT-SR; LVEF ≥ 50%).^2–7 Left ventricular end-systolic volume (LVESV) has also been used to categorize CRT response into super-responders (decrease in LVESV ≥30%), responders (decrease in LVESV 15–29%), non-responders (decrease in LVESV 0–14%), and negative responders (increase in LVESV), after 6 months following CRT implantation;⁸ other investigators have added a category of non-progressors.⁹ Finally, other left ventricular indices and functional criteria have been used in some studies to define CRT response.^10–14

Despite early conflicting findings and uncertainties, prior meta-analyses have indicated a significant reduction in appropriate implantable cardioverter-defibrillator (ICD) therapy, a surrogate of clinically relevant ventricular arrhythmias (VA), in CRT responders compared with non-responders.^15–17 However, hitherto, no meta-analysis has assessed the impact of CRT-SR (full left ventricular systolic function recovery of ≥ 50%) on VA, and the magnitude of VA risk in this group is not fully understood. In addition, data on the impact of CRT response and super-response on all-cause mortality have not been adequately synthesized. There is on-going debate regarding the optimal management of patients with fully recovered LVEF who are undergoing generator change, and strategies such as risk re-stratification with consideration of downgrade from CRTD to CRTP have been suggested.^2,3 Furthermore, since the publication of the prior meta-analyses, a significant number of additional studies has accrued in this area. We therefore undertook an updated systematic review and meta-analysis with the addition of more data from recent studies including a previously unexamined category of fully recovered LVEF or CRT-SR, to determine the impact of CRT response and super-response on VA requiring appropriate ICD therapy, inappropriate ICD therapy, and all-cause mortality.

Methods

This study was registered prospectively with the international prospective register of systematic reviews, PROSPERO (registration number CRD42020193128).

Search strategy

We systematically searched three databases (PubMed/MEDLINE, EMBASE, and Cochrane Library) to identify all relevant English language studies published until 31 December 2019 and restricted to adults and humans. The combined search terms used were as follows: ‘(cardiac resynchronization therapy OR cardiac resynchronization therapy defibrillator OR cardiac resynchronization therapy pacemaker OR implantable cardioverter-defibrillator OR CRT OR CRTD OR CRTP OR ICD) AND (super-responders OR responders OR ventricular function recovery OR ventricular function improvement OR ventricular function normalisation) AND (ventricular arrhythmias OR shocks OR therapies OR anti-tachycardia pacing OR mortality OR survival OR death)’. Manual searches of bibliography of published articles were also undertaken.

Inclusion criteria

The included studies consisted of investigations reporting on CRT response or super-response vs. CRT non-response during follow-up. The first study group consisted of CRT responders and the first control group consisted of CRT non-responders. These groups were defined in the respective studies using echocardiographic parameters, which included any of the following: changes in LVEF, LVESV, left ventricular end-systolic volume index (LVESVI), left ventricular end-systolic diameter (LVESD), or left ventricular end-systolic diameter index (LVESDI). The second study group consisted of super-responders (LVEF ≥ 50%) and the second control group consisted of CRT non-super-responders (CRT-NSR; non-responders + responders) with LVEF < 50%. Only studies which employed echocardiographic criteria to define CRT response were included in meta-analysis.

Exclusion criteria

Studies were excluded if only published as an abstract without a full article, non-English language studies without English translated versions, if they lacked clear echocardiographic assessment during follow-up to check for CRT response, no clear follow-up after assessing for CRT response, lacked control group, or did not report on any of the outcomes or reported outcomes in combinations that did not satisfy the objectives of this meta-analysis. We excluded editorials, commentaries, notes, conference abstracts without full published articles, and reviews (Figure 1).

Outcomes

The study outcomes were pooled incidence (first occurrence) of VA episodes requiring appropriate ICD therapies, inappropriate shocks, and all-cause mortality during follow-up in the study group vs. control group. There was some variability in device programming across studies which reported ICD settings and we accepted VA as defined by each study.

Quality assessment of studies

As most of the included studies were non-randomized and mainly cohort observational studies, the Newcastle–Ottawa Scale (NOS) for assessing the quality of non-randomized studies in meta-analyses was used for quality assessment.¹⁸

Data extraction

Data were independently extracted by two authors (M.Y. and S.E.) using standardized forms. Discrepancies were resolved by consensus. Data extraction was done according to pre-defined data elements that included year of publication, study sample size, demographics, baseline covariates, measures of CRT response, duration of follow-up after assessment for CRT response, and outcomes.

Statistical analysis

All analyses were performed using the STATA software package (Stata Corp., TX, USA). The method for pooling study specific estimates was a priori determined to be random-effects model (DerSimonian–Laird) as some degree of heterogeneity was anticipated. Results derived from fixed-effects model (Mantel–Haenszel) were also included in the forest plot for comparison. Incidence (%) and relative risk (RR) estimates with 95% confidence interval (CI) of developing VA or receiving inappropriate shocks or all-cause mortality in the study vs. control groups are presented. The statistical significance of the pooled RR was examined by the Z-test (statistical test of the null hypothesis).

The magnitude of heterogeneity across studies was assessed using the I² statistic, where I² = ([Q − df]/Q) × 100%, with Q being the Cochran’s heterogeneity statistic and df its degrees of freedom.¹⁹ The I² statistic, describes the percentage variability in effect estimates that is due to true between study heterogeneity (difference) rather than sampling error (chance). When I² was < 25%, heterogeneity was considered absent; when I² was 25–50%, heterogeneity was considered low; when I² was 50–75%, heterogeneity was considered moderate; and when I² was > 75%, heterogeneity was considered high.¹⁹ Publication bias was assessed by visual scrutiny of a funnel plot of study-specific estimates by the study standard errors. When funnel plot asymmetry was observed, a contour-enhanced funnel plot was fitted to determine whether it was attributed to publication bias.²⁰

Results

The initial searches of the three databases and the references lists of published articles revealed 4917 citations (Figure 1). After applying inclusion, exclusion, and quality assessment criteria, 34 studies^{2,3,5–8,10,11,13,14,21–44} were retained for meta-analysis (Figure 1 and Table 1). These studies were mostly non-registry-based cohort studies (n = 25), seven of which were prospective; registry-based cohort studies (n = 6); and post hoc randomized controlled trials (RCTs) (n = 3). Supplementary material online, Table S1 shows the quality assessment of included studies.

Baseline characteristics

Table 1 depicts the baseline characteristics of participants at the time of original implantation of CRT device. The mean age of patients ranged from 59 to 71 years and more than 70% were men. Baseline LVEF ranged from 20% to 29%, and apart from one study that included only non-ischaemic cardiomyopathy patients,³⁰ the prevalence of ischaemic cardiomyopathy was 30–75% across studies. Common comorbidities were hypertension, diabetes, and atrial fibrillation. Overall, the use of guidelines directed medical therapy was fairly high with 66–98% of patients on renin angiotensin aldosterone system inhibitors and 60–98% on β-blockers (Supplementary material online, Table S2).

Cardiac resynchronization therapy response

The time from CRT device implantation to assessment of CRT-response varied across studies with a median [interquartile range (IQR)] of 7.9 (6–18) months. Included were only studies which employed echocardiographic criteria to define CRT response. There was observed heterogeneity on definition of CRT response across studies with some using ≥ 10%, ≥15%, ≥30% reduction in LVESV; others using increased in LVEF of ≥5%, ≥ 10%, ≥25%; increased in LVEF × 2 from implant; LVEF >35%, LVEF > 40%, LVEF >45%, and LVEF ≥50%. The commonest definitions for CRT response were LVESV decrease of ≥15% or LVEF > 35%; and CRT-SR was LVESV reduction of > 30% or LVEF ≥50% (Table 1). In our analysis, we used LVEF ≥50% cut-off for definition of super-response. The duration of follow-up after initial assessment for CRT response varied from 1 to 6 years.

Risk of ventricular arrhythmia in cardiac resynchronization therapy responders compared with non-responders

A total of 28 studies^{2,3,6,7,11,13,14,21,22,24–30,33–44} with a total of 7605 patients were included in the meta-analysis assessing the risk of VA in CRT responders compared with CRT non-responders. The median duration of follow-up across studies after assessment for CRT response was 2.9 years (IQR 2.2–4.0 years). The pooled cumulative incidence of VA leading to appropriate ICD therapies was 22% (7.6% per annum) for all the participants. The incidence rate was significantly lower at 13% (4.5% per annum) in CRT responders, vs. 29% (10.0% per annum) in CRT non-responders, with a RR of 0.47 (95% CI 0.39–0.56, P < 0.0001) in favour of CRT responders (Figure 2). When analyses were limited to the 23 pure CRT studies only (after excluding five studies where subsets of participants had only ICD implanted), the crude incidence of VA remained significantly lower in CRT responders at 14.0% (5.4% per annum) vs. 29.0% (11.2% per annum), in non-responders; with a RR of 0.46 (95% CI 0.37–0.57, P < 0.0001) in favour of CRT-responders (Supplementary material online, Figure S1).

Inappropriate implantable cardioverter-defibrillator therapies risk in cardiac resynchronization therapy responders compared with non-responders

Data on the risk of inappropriate ICD shocks were available from eight studies^{2,26,27,29,30,34,35,39} with pure CRTD patients only, as shown in Figure 3. During a median follow-up of 2.8 years (IQR 2.3–4.1), the pooled crude incidence of inappropriate ICD therapies in CRT responders was 9.0% (annualized rate of 3.2%) vs. 8.0% (2.9% per annum) in non-responders, with a RR of 1.00 (95% CI 0.52–1.93, P = 0.955), indicating no difference between the two groups.

All-cause mortality in cardiac resynchronization therapy responders compared with non-responders

Fifteen studies^{2,5,7,8,10,23,27,30–32,35,39–42} involving a total of 5874 patients were included in the meta-analysis of all-cause mortality outcome. During the median follow-up of 3.4 years (IQR 3.0–5.5), 1326 of the 5874 patients died, yielding a pooled mortality rate of 23.0% (annualized rate of 6.8%). All-cause mortality rates were significantly lower in CRT responders at 12.0% (3.5% per annum), compared with non-responders 31.0% (9.1% per annum); RR 0.38 (95% CI 0.30–0.48, P < 0.0001, Figure 4). When the meta-analysis was restricted to the 14 pure CRT studies (excluding the one study with <100% CRT), the findings were similar to those reported above, with the pooled all-cause mortality rate being 12.0% (annualized rate of 4.0%) in responders, vs. 31.0% (8.6% per annum) in non-responders; RR 0.38 (95% CI 0.30–0.49, P < 0.0001).

Pure primary prevention cardiac resynchronization therapy with defibrillator studies

Figure 5 depicts forest plots of pure primary prevention CRTD studies showing significantly reduced risk of VA and all-cause mortality in CRT responders, compared with non-responders.

Ventricular arrhythmias and mortality in cardiac resynchronization therapy super-responders (left ventricular ejection fraction ≥ 50%) compared with non-super-responders (left ventricular ejection fraction <50%)

Five studies^2,6,7,40,41 included VA incidence and risk in patients with CRT-SR defined as LVEF≥ 50% vs. those with LVEF < 50%. Only one study³⁵ used a decrease in LVESV ≥30% to define CRT-SR, and was not included in this sub-analysis. Over a median follow-up of 5.5 years, the pooled crude incidence of VA was significantly lower in super-responders at 5.0% (0.9% per annum), compared with 21.0% (3.8% per annum) in those with LVEF <50%; RR of 0.22 (95% CI 0.12–0.40, P < 0.0001) using random-effects model. (Figure 6A). Four studies^2,5,7,41 assessed mortality outcomes in CRT-SR with fully recovered LVEF of ≥ 50%; and the pooled rate of all-cause mortality was significantly lower in super responders at 11.0% (2.0% per annum) compared with those with LVEF of <50% at 24.0% (4.3% per annum); RR 0.47 (95% CI 0.33–0.66, P < 0.0001, Figure 6B).

Assessment for publication bias and small study effects

The funnel plots appear asymmetric with the fitted lines from standard Egger’s regression (Supplementary material online, Figure S2A and B). However, the contour-enhanced funnel plots suggest that publication bias is less likely (Supplementary material online, Figure S2C D).

Table 2 summarizes of the main findings of this meta-analysis. Sensitivity analyses were performed through exclusion of each individual study at a time; these failed to identify one specific study driving the results in one direction or the other across the various tested outcomes (Supplementary material online, Figures S3 and S4).

Discussion

Our meta-analysis of 34 studies, involving nearly 8000 participants for assessment of VA risk and nearly 6000 participants for assessment of all-cause mortality, revealed that the pooled annual rate and RR of appropriate ICD therapy, a surrogate of clinically relevant VA, and all-cause mortality are significantly lower in CRT responders compared with CRT non-responders. However, we did observe a significant residual risk of VA in CRT responders. We observed that CRT-SR with full normalization of systolic function (LVEF ≥ 50%) carry very low pooled absolute risk of VA, compared with CRT patients with LVEF < 50%. There was no significant difference in the risk of inappropriate ICD therapies between responders and non-responders. The time from CRT device implantation to evaluation for CRT-response varied across studies with a median of 7.9 months, which is slightly longer than the accepted traditional least time interval of 6 months post implantation.⁸ However, these observational studies are probably reflective of real-world clinical practice.

In the current study, we synthesized a larger number of studies compared with previous meta-analyses, and confirmed and extended the finding that there is a lower risk of VA and appropriate ICD shocks in CRT responders vs. non-responders.^15–17,45 Prior meta-analyses have shown that improvement in LVEF to > 35% vs. ≤ 35% is associated with a significantly reduced risk of VA and mortality, in combined ICD and CRTD groups as well as in ICD-only and CRTD-only subgroups during follow-up,⁴⁵ and at generator change.⁴⁶ Individual studies with combined outcomes not allowing for inclusion in the present meta-analysis, have shown that CRT response is associated with significant reduction in composite outcomes of mortality, heart failure hospitalization, left ventricular assist device implantation, and heart transplant.^47–50

We found a substantial residual absolute risk of VA among CRT responders, despite having a lower RR when compared with non-responders. Prior studies and systematic reviews also demonstrated the non-trivial residual risk of ventricular arrhythmias in patients with improvements in left ventricular systolic function >35% after ICD or CRTD implantation at follow-up or at generator change, suggesting continuous clinical benefit from defibrillator therapy despite these improvements.^{2,45,46,51–55} Our meta-analysis is the first to synthesize data and highlight this residual risk quantitatively. Randomized controlled trials comparing risk of VA or sudden cardiac death or all-cause mortality in CRTD vs. CRTP or ICD vs. no ICD for primary prevention in patients with LVEF improvement with the 35–50% range are lacking. In the absence of clinical trials, the observational evidence that is presented in this meta-analysis and other studies might justify the continuation of defibrillator therapy in CRT responders undergoing pulse generator replacement.

Prior studies showed that CRTD is associated with significant reduction in VA compared with ICD only,^16,17 while CRT non-response is associated with increased risk of VA when compared with ICD only.¹⁷ This suggests that trans-coronary sinus or epicardial left ventricular pacing might be inherently pro-arrhythmic in the absence of reverse remodelling.¹⁷ In the present meta-analysis data were not sufficient to compare the risk of VA between CRT-D non-responders and patients with ICD without improvement in LVEF.

A key finding of this meta-analysis is that CRT-SR with LVEF ≥ 50%, have a very small absolute risk of VA and low risk of all-cause mortality when compared with patients with LVEF <50%. To the best of our knowledge, this meta-analysis is the first to evaluate the pooled effect of CRT-SR. It has been shown that cessation of CRT (biventricular pacing) leads to recurrence of negative remodelling and significant left ventricular systolic dysfunction in super-responders of CRT, suggesting the need for long-term pacing to preserve recovered ventricular function.⁵⁶ Cardiac resynchronization therapy super-responders (LVEF ≥50%) are known to have comparable survival with the age–sex-matched general population, and super-responders with CRT pacemakers have similar survival to those with CRT defibrillators.⁵ There are many definitions of CRT-SR without an agreed consensus;⁹ in this meta-analysis, we used the definition that almost all of our included studies used. The optimal management of CRTD patients who develop full recovery of LVEF with no prior appropriate ICD therapies at time generator change remains unclear. Based on the observed very low absolute and RR of VA and lower risk of mortality in patients who achieved LVEF ≥50% and given risk of inappropriate shocks, some investigators have suggested that these patients could be considered for downgrade from CRTD to CRTP at the time of battery depletion if no VA have occurred.^2,3 However, a joint Task Force report (ACCF/HRS/AHA/ASE/HFSA/SCAI/SCCT/SCMR) on the appropriate use criteria for ICD and CRT, suggested that it may be appropriate for patients with primary prevention ICD and no clinically relevant VA and normalized LVEF of ≥50% at time of ERI to proceed with the generator replacements.⁵⁷ Therefore, amidst this uncertainty, risk re-stratification using LVEF quantification, history of prior appropriate ICD therapies, comorbidities, life expectancy, ischaemic vs. non-ischaemic cardiomyopathy, and goals of care in a shared decision-making manner between the implanting clinician and patient may be reasonable. It is not known whether conducting an EP study or performing cardiac imaging to look for myocardial scaring in these patients, might also be helpful in risk re-stratification. Nonetheless, specific outcome-related criteria for risk re-stratification at generator change of CRT devices remain elusive and should be subject of future studies. Notwithstanding these considerations, the annual incidence of ventricular arrhythmia in CRT responders seen in this meta-analysis was 4.5% and 0.9% in CRT-SR (LVEF ≥50%), with the later still >15-fold higher than the incidence rate seen in the general population of 0.57 per 1000 person-years.⁵⁸

Limitations

This meta-analysis has some important limitations. First, the included studies were mainly non-randomized observational cohort studies or registries, many of which were retrospective with all the inherent biases and confounding associated with such a methodology. Secondly, ICD programming varied across studies and the arrhythmia detection and therapies zones were not reported in a consistent manner, with some studies lacking device programming details. However, the review of each individual study suggested that programmed detection and tachycardia therapy parameters did capture and treat the clinically relevant VA. Due to incomplete reporting and design of studies, subgroup analysis of types of anti-tachycardia therapies (shocks vs. anti-tachycardia pacing) and ischaemic vs. non-ischaemic substrates could not be performed. Thirdly, the use of antiarrhythmic medications may have impacted arrhythmia incidence and ICD therapies, but there was a non-uniform and incomplete reporting of their use, with the possibility of residual confounding. Fourthly, the CRT response rate based only on echocardiographic criteria in this meta-analysis of majority observational studies was about 46%, as we excluded other criteria of CRT response like NYHA class, exercise duration (6-min walk test or VO_2max), quality of life measures, etc. This is lower than the 65–70% CRT response rate seen in clinical trials, raising the possibility of misclassification of some CRT responders to non-responders which could dilute the associations of CRT response with biasing of outcomes towards the null. Finally, we used LVEF ≥ 50% to define CRT-SR as this is what the identified primary studies used and which is clinically more relevant to implanting physicians, instead of decrease in LVESV ≥30% that is sometimes used or other criteria. We do not have data on how much responder/super-responder status is dependent on an ideal LV stimulation.

Conclusions

Despite a significant reduction in all-cause mortality and the risk of VA in CRT responders, there remains a substantial non-trivial residual risk of these outcomes. However, the absolute risk of VA is very low in CRT-SR (LVEF ≥ 50%). These findings suggest that there might be continued clinical benefit of defibrillator therapy in CRT responders, whereas consideration of risk re-stratification might seem appropriate for CRT-SR with fully normalized left ventricular systolic function at the time of generator change.

Supplementary material

Supplementary material is available at Europace online.

Funding

S.A.E. was supported by funding from the Department of Veterans Affairs, Veterans Health Administration, VISN 1 Career Development Award. S.A.E. also received funding from Center for Aids Research, The Rhode Island Foundation, and Lifespan Cardiovascular Institute.

Conflict of interest: none declared.

References