Incidence and Predictors of Implantable Cardioverter-defibrillator Therapies After Generator Replacement—A Pooled Analysis of 31,640 Patients’ Data

Among primary prevention implantable cardioverter-defibrillator (ICD) recipients, 75% do not experience any appropriate ICD therapies during their lifetime, and nearly 25% have improvements in their left ventricular ejection fraction (LVEF) during the lifespan of their first generator. The practice guidelines concerning this subgroup’s clinical need for generator replacement (GR) remain unclear. We conducted a proportional meta-analysis to determine the incidence and predictors of ICD therapies after GR and compared this to the immediate and long-term complications. A systematic review of existing literature on ICD GR was performed. Selected studies were critically appraised using the Newcastle–Ottawa scale. Outcomes data were analyzed by random-effects modeling using R (R Foundation for Statistical Computing, Vienna, Austria), and covariate analyses were conducted using the restricted maximum likelihood function. A total of 31,640 patients across 20 studies were included in the meta-analysis with a median (range) follow-up of 2.9 (1.2–8.1) years. The incidences of total therapies, appropriate shocks, and anti-tachycardia pacing post-GR were approximately 8, 4, and 5 per 100 patient-years, respectively, corresponding to 22%, 12%, and 12% of patients of the total cohort, with a high level of heterogeneity across the studies. Greater anti-arrhythmic drug use and previous shocks were associated with ICD therapies post-GR. The all-cause mortality was approximately 6 per 100 patient-years, corresponding to 17% of the cohort. Diabetes mellitus, atrial fibrillation, ischemic cardiomyopathy, and the use of digoxin were predictors of all-cause mortality in the univariate analysis; however, none of these were found to be significant predictors in the multivariate analysis. The incidences of inappropriate shocks and other procedural complications were 2 and 2 per 100 patient-years, respectively, which corresponded to 6% and 4% of the entire cohort. Patients undergoing ICD GR continue to require therapy in a significant proportion of cases without any correlation with an improvement in LVEF. Further prospective studies are necessary to risk-stratify ICD patients undergoing GR.


Introduction
In the United States, about 800,000 people have implantable cardioverter-defibrillators (ICDs), and approximately 150,000 ICDs are implanted annually. 1 Among primary prevention ICD recipients, 75% do not experience any appropriate ICD therapies during the lifetime of their first ICD generator. 2 This trend could be due to goal-directed medical therapy, Multicenter Automatic Defibrillator Implantation Trial-Reduce Inappropriate Therapy (MADIT RIT)-type programming, or because half of all device recipients aged >65 years die within 5 years after device implantation. [3][4][5] Nearly 25% of primary prevention ICD patients experience improvements in their left ventricular ejection fraction (EF) (LVEF) to >35% during the life of the first generator. Clinical guidance on the appropriateness of generator replacement (GR) in this cohort remains ill-defined. 6 It has been found that patients with an LVEF of 35% are more likely to experience an appropriate ICD therapy than those with an LVEF of >35%, as was confirmed in a recent meta-analysis. 2,7 It is undoubtedly true that the decision to replace the ICD generator requires more meticulous consideration than absolute EF in isolation, and frailty is an essential factor to be considered. 8 Complication rates after ICD GR range from 5%-10% and tend to be higher in concert with lead upgrades, multiple replacements, and patient frailty. [9][10][11] Of note, experiencing a major complication during ICD implantation is associated with a significantly higher risk of mortality for up to 6 months after the procedure. 12 Moreover, although ICDs reduce arrhythmic death, 25% of patients die after their first ICD generator replacement without ever experiencing an appropriate ICD therapy, and improvement in EF has not been proven as a predictor of all-cause mortality, thus reiterating that ICDs may not reduce non-arrhythmic death. 2 Therefore, we conducted a proportional meta-analysis to determine the incidence and predictors of ICD therapies after GR and compared the immediate and long-term complications. The need for ICD GR, especially in primary prevention cohorts with no previous therapies and improved LVEFs, has been debated. This study attempts to answer this question, which has important implications as electrophysiologists commonly face this dilemma in their daily practice.

Search strategy
A systematic review of existing literature on ICD GR in primary prevention patients was performed to search for studies published prior to December 2021. Two physicianreviewers queried PubMed, the Web of Science, and the Cochrane Library of Controlled Trials (CENTRAL) databases for published literature, using keywords such as "implantable cardioverter-defibrillator," "generator replacement," "primary prevention," and "therapies" alone and in combination. Additional literature was sought by searching the references of eligible articles. A third reviewer resolved any discrepancies.

Study selection
For the meta-analysis, we selected observational studies and registries that reported data on ICD GR in primary and secondary prevention patients. Case reports, case series, editorials, and review articles were excluded. Only studies on transvenous ICDs were included. Details are given in the Preferred Reporting Instrument for Systematic Reviews and Meta-analysis flow diagram (Figure 1). 13 Studies were critically appraised using the Newcastle-Ottawa scale (Table S1). 14 This meta-analysis has been registered in PROSPERO (identifier: CRD42021288546).

Data extraction
Baseline characteristics included the type of study, total number of patients, follow-up duration, year of publication, sex ratio, age, diabetes, hypertension, chronic kidney disease (CKD), body mass index, atrial fibrillation (AF), indication (primary or secondary), previous therapies (anti-tachycardia pacing [ATP] therapies or shocks), LVEF, improvement in LVEF, biventricular device, ischemic cardiomyopathy (ICM), New York Heart Association (NYHA) class, QRS duration, heart failure (HF) medications, and anti-arrhythmic drugs (AADs). In addition, outcomes such as procedural complications, therapies (shocks and ATPs), inappropriate shocks, and all-cause mortalities were also extracted from the individual studies.

Outcome
The primary outcomes of the meta-analysis were the incidence rates and determinants of ICD therapies and all-cause mortality, while secondary outcomes were incidence rates of ICD shocks, ATP, inappropriate shocks, and procedure-related complications. Based on the available data, attempts were made to perform covariate analyses of all the primary and secondary outcomes.

Data analysis
To pool the incidence of outcomes across all the selected studies, the R software (RStudio: Integrated Development for R; RStudio, Inc., Boston, MA, USA) using the "metarate" function was used to derive the incidence rate ratio in 100 patient-years. A random-effects model was used to circumvent heterogeneity so that no particular study was given a higher weightage. The outcome data were transformed by the logarithmic method to increase their        statistical value and back-transformed to proportions using the "escal" function in "metafor." 15 Forest plots were drawn with proportional measures and 95% confidence intervals (CIs) using a random-effects model. Study heterogeneity was expressed as I 2 . Influencing and outlier studies were looked for, and sensitivity analyses were conducted. Univariate and multivariate logarithmic covariate analyses of the outcomes were performed using the restricted maximum likelihood (REML) function. R 2 was used to estimate the heterogeneity accounted for, and a bubble plot was drawn to visualize the moderator effects.

Baseline data
Eighty-one percent of patients in the meta-analysis cohort were male. Most of the studies defined an improvement of LVEF as LVEF ≥ 35%; however, House et al. 19 defined it as LVEF ≥ 50%, and Kini et al. 6 and Sebag et al. 17 both defined it as LVEF ≥ 40%. The baseline LVEF was 26.7% ± 3.4% in the entire cohort, with 58.9% of patients overall having ICM, and proportions of patients with NYHA class 3 or 4 symptoms ranged from 14.6%-76.6% across the studies. The common comorbidities included hypertension (59.5%), diabetes mellitus (DM) (31.1%), CKD (21.8%), and AF (36.5%). Overall, 45.6% of patients received cardiac resynchronization therapy with defibrillation (CRT-D), and 78.4%, 84.4%, 39.6%, and 27.9% of patients were on angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers (ACEI/ARBs), β-blockers, mineralocorticoid inhibitors, and digoxin, respectively. AADs were prescribed to 20.7% of patients. These ICDs were programmed in multiple zones with therapy advised for fast and sustained ventricular arrhythmias starting with ATP followed by shock. Only Witt et al. 26 and Narducci et al. 27 reported secondary prevention ICD replacements

Total therapy
ICD therapies were delivered to 22% of the entire cohort ( Figure 3A). The study by Witt et al. reported 51.5% of patients had secondary prevention ICDs; however, excluding this study, the incidence of ICD therapies remained at 21%. Incidence rates of ICD therapies in retrospective and prospective studies were 7.7 and 8.  Table 2). In the multivariate analysis, the use of AAD and previous shocks was found to be associated with higher ICD therapies following GR (Table 2).

Implantable cardioverter-defibrillator shocks
Twelve percent of patients of the entire cohort received ICD shocks in our pooled analysis ( Figure S1A). A funnel plot is shown in Figure S1B (Table S2).

Anti-tachycardia pacing
Twelve percent of patients in the entire cohort received ATPs in our pooled analysis ( Figure S2A). There was no evidence of publication bias ( Figure S2B ; R 2 = 52%), but these findings were not found to be significant in the multivariate analysis ( Figure S2D). The incidence of ATP was not found to be affected by LVEF (ES [95% CI], 0.06 [−0.02 to 0.15]; P = .14) ( Figure S2E). The results of bivariate and multivariate analyses are presented in Table S3.

All-cause mortality
The all-cause mortality was found to be 5.52 (95% CI, 3.64-7.77; I 2 = 96%) per 100 patient-years ( Figure 2). The proportion of all-cause mortality was 17% in the total cohort ( Figure 4A). A funnel plot is depicted in Figure 4B Figure 4F) were found to be associated with higher all-cause mortality rates, but they did not independently affect all-cause mortality in the multivariate analysis (

Discussion
In this proportional meta-analysis of 31,640 patients across 20 studies who had received ICDs predominantly for primary prevention, the incidence of total therapies post-GR was 8 per 100 patient-years, whereas the incidence of all-cause mortality was 5.52 per 100 patientyears in the entire cohort. Also, the incidence rates of inappropriate shock and other procedural complications were low at 2.32 and 1.56 per 100 patient-years, respectively. The use of AADs was independently associated with more frequent ICD therapies, and previous ICD shocks but not ATPs were an independent predictor of ICD therapies. Finally, LVEF did not affect total therapies from ICD or all-cause mortality.

Implantable cardioverter-defibrillator therapies
Our meta-analysis showed that using AADs independently increases ICD therapies in primary prevention ICD patients undergoing GR. Patients receiving AADs may experience drug-device interactions, which lead to pro-arrhythmic effects that alter the tachycardia cycle length and/or change the QT interval, which may lead to more frequent ICD therapies. 32,33 Although this may be true due to the risk of pro-arrhythmia with these medications, it is equally plausible that the use of anti-arrhythmic medications is a marker of previous arrhythmias and, therefore, would be associated with an increased risk for future arrhythmias. Our meta-analysis revealed that ICD shocks in the first life of the generator predict overall ICD therapies following GR, which resonated with the findings from Arcinas et al.'s study. 11 In our meta-analysis, patients with appropriate shocks were associated with the use of β-blockers probably because β-blockers are more likely to be used in patients with previous shocks or those at high risk for shock and primary prevention patients with HF with reduced ejection fraction. The Optimal Pharmacological Therapy in Cardioverter-defibrillator Patients (OPTIC) trial evaluated the effect of amiodarone, β-blockers, sotalol, and the combination of amiodarone and β-blockers on ICD shocks and showed that shocks were lower in the combination amiodarone-β-blocker and sotalol arms than in the β-blockers arm. 34 The addition of β-blockers to ICD therapy has been found to reduce HF-related hospitalization and mortality rates. 35 LVEF is currently our best risk-stratification tool when planning for primary prevention ICD implantation. However, our meta-analysis shows that the absolute improvement of LVEF did not affect total therapies, appropriate shocks, or ATP in patients who had undergone GR. Therefore, we propose that LVEF in isolation should not be considered the sole criterion for deciding on GR in these patients. Furthermore, though LVEF might improve the pro-arrhythmic substrate in the form of scars, initiating and sustaining arrhythmogenic foci may linger, thus prompting the need for an ICD to prevent sudden cardiac death.

Mortality
In this meta-analysis, DM, an ischemic cause of low LVEF, concurrent AF, and greater use of digoxin were associated with higher all-cause mortality, but they did  not independently affect all-cause mortality. LVEF, or an improvement in LVEF, also did not affect mortality. The role of digoxin in HF has been dubious, and studies have shown that digoxin can be related to increased mortality in patients with HF. 36 Merchant et al. recommended that the decision for GR should be made while keeping in mind the relative risks of arrhythmic and nonarrhythmic deaths. It is mandatory to comprehend that, since initial implant patients undergoing GR are older and may have a plethora of existing and or newer comorbidities, they would ostensibly have a shorter lifespan, which may limit the benefit of ICD therapy. Moreover, the probability of additional complications, including infections, cannot be discounted. 37 Thus, quality of life should be considered, and the patient should have a life expectancy of ≥1 year and a reasonable quality of life to justify GR. 38

Complications (inappropriate shocks and procedural issues)
We have shown that the risks of procedure-related complications and inappropriate shocks are very low (6% and 4%, respectively). Rates of procedural complications in ICD GR cases can be notoriously higher than during the first implant owing to patient frailty, and complications tend to be more common also in cases of lead upgrades and multiple GRs. Looi et al. 10 reported the highest complication rate (10%), possibly due to lead fractures in 6% of patients and upgrades to CRT in nearly 35% of the patients, whereas complications were very low (2%) in the large registry by Thomas et al. 29 Lead fracture as a risk for mortality was likely reflective of the Fidelis (Medtronic, Minneapolis, MN, USA) and Riata (Abbott, Chicago, IL, USA) era. Although the rate of ICD failures from lead fractures is much less in the contemporary world, cardiovascular implantable electronic device infections are likely to predict mortality. However, only Arcinas et al. reported 1.4% lead infections in their ICD cohorts. 11 It is also known from the Multicenter Electrophysiologic Device Infection Cohort (MEDIC) trial that the infection risk for ICD replacement increases exponentially with each pocket re-entry. 40 In our meta-analysis, inappropriate shocks were found to be more common in CKD patients, which reflected the finding by Bansal et al., which showed that patients with CKD have a higher probability of having inappropriate shocks, probably owing to a higher incidence of hyperkalemia and AF in this population. 41 The lower incidence of inappropriate shocks with ACE/ ARB therapy might be related to the improving overall HF status.

Limitations
The present work of meta-analysis is based on pooled data across the selected studies. However, we did not have access to individual patient data and thus had to perform the statistical analysis based only on the published data. A huge male bias was observed in the pooled patients' data. The dataset was grossly heterogeneous, and we did not perform a comparative meta-analysis or subgroup analysis (between patients with improved and not improved LVEFs) due to the lack of separate outcome data in many studies. Some studies did not specify the indications for ICDs (primary or secondary). Device programming was not specified in detail and may have variations across the studies. Some studies did not mention the incidences of previous therapies (such as shocks or  ATPs). Most of the studies did not enumerate complications. Insufficient data restricted our ability to perform covariate analysis for complications. Objective assessments of funnel plots depicting incidence rates of appropriate and inappropriate shocks and all-cause mortality were asymmetric, suggesting very heterogeneous data rather than publication bias in the context of proportional meta-analysis. 15 The findings of the covariate analysis of the outcomes in the meta-analysis need further assessment in prospective registries and may not be replicated in clinical practice.

Conclusion
The need for GR of ICDs, especially in primary prevention cohorts, with no previous therapies and improved LVEFs, has been questioned. We contend that ensuring informed consent from and sharing decision-making with patients are essential. Our findings portend a significant residual risk of sudden cardiac death in this cohort irrespective of the improvement in LVEF. In our large cohort, the use of AADs was associated with an increased risk for appropriate ICD therapies. The pathophysiological etiology of this phenomenon remains to be determined. It is also essential to comprehend that, while GR of ICDs prevents arrhythmic death, it has no bearing on all-cause mortality.
Furthermore, the risk of complications associated with GR is minimal in most circumstances. The benefit offered by GR thus squarely outweighs its risk, though not in all cases. Decision-making should thus be individualized, catering to the individual patient-associated mortality and morbidity. A consensus thus arrived at and discussed with the patient will enable an understanding of the risk-benefit ratio for a GR with great precision. A detailed prospective analysis is mandated to substantiate our findings. It is also evident that a better comprehension of the predictors of therapies needs to be defined, which may help us decide on GR.