OUP user menu

Do implantable cardioverter defibrillators improve survival in patients with chronic kidney disease at high risk of sudden cardiac death? A meta-analysis of observational studies

Nader Makki, Paari Dominic Swaminathan, Janel Hanmer, Brian Olshansky
DOI: http://dx.doi.org/10.1093/europace/eut277 55-62 First published online: 20 September 2013

Abstract

Aims Prospective randomized clinical trials show that implantable cardioverter defibrillators (ICDs) can reduce the risk of total mortality in select populations. However, data regarding patients with chronic kidney disease (CKD) are inconclusive. The aim of this study was to evaluate if ICDs affect total mortality in CKD patients at high risk of sudden cardiac death.

Methods and results Two separate meta-analyses were performed to (i) assess the effect of ICD on all-cause mortality in CKD patients at high risk of sudden cardiac death and (ii) assess the effect of CKD on all-cause mortality in patients who already had an ICD for primary or secondary prevention purposes. Medline and EMBASE were searched from 1966 to 2013. A manual search by cross-referencing was performed. Five observational studies with 17 460 CKD patients considered at high risk of sudden cardiac death were included to evaluate the effect of ICDs on patients with severe CKD. Patients with ICD implants had a reduction in all-cause mortality (adjusted hazard ratio (HR) = 0.65, 95% confidence interval (CI) = 0.47–0.91, P < 0.05) compared with a matched control group. Based on 15 observational studies with 5233 patients as part of our second comparison that evaluated the effect of CKD on patients who received an ICD, CKD was associated with higher mortality risk (HR = 2.86, 95% CI = 1.91–4.27, P < 0.05) despite an ICD.

Conclusion The meta-analysis indicates that for patients undergoing ICD implant, CKD is associated with greater risk of dying. However, ICD placement reduces mortality in CKD patients at high risk of sudden cardiac death.

  • Kidney disease
  • Implantable cardioverter defibrillator
  • Sudden death
  • Mortality

What's new?

  • Our meta-analysis represents the most complete meta-analysis of trials focusing on the issue of chronic kidney disease in patients undergoing ICD implantation or already having an ICD

  • Given the lack of randomized trials in CKD patients, this analysis offers a comprehensive overview in the absence of guidelines

  • While patients with chronic kidney disease undergoing ICD implant have a greater risk of dying than patients without chronic kidney disease, ICD implantation in these patients at high risk of sudden cardiac death is associated with lower risk of dying.

Introduction

All patients at risk for sudden cardiac death (SCD) may not necessarily benefit from an implantable cardioverter defibrillator (ICD) if the risk of dying from other causes is too great or the benefit of an ICD is too small. This concept of competing risks highlights the fact that patients may be at high risk of dying from other causes;1 this is particularly pertinent to patients with chronic kidney disease (CKD) as their risk of dying from non-arrhythmic causes exceeds the general ICD population.2

However, SCD, being one of the leading cardiovascular causes of death,3 is particularly high in the CKD population and the risk increases with worsening renal dysfunction.4 In a large study of patients with coronary artery disease enroled in the United States Renal Dialysis System (USRDS), SCD risk increased with worsening glomerular filtration rate (GFR), particularly for those undergoing dialysis.4 Despite this, only 8% of patients undergoing dialysis receive an ICD after a cardiac arrest.5 Reluctance to implant ICDs in patients with CKD may stem from concerns that ICD implantation in CKD patients may carry a higher risk of complications (such as infection and bleeding) than in the general population or anticipation that patients with CKD would have less benefit from an ICD since many of the deaths are non-arrhythmic or non-cardiac, as was shown in the SOLVD trial.69

While ICDs have been shown to improve survival in patients at risk of SCD, major cardiovascular disease trials have excluded patients with CKD or did not provide adequate information on the renal function of enrollees or the effect of interventions on patients with renal disease.810 To understand the value of the ICD in patients with CKD at risk for SCD in a better manner, we performed two meta-analyses of published clinical data to attempt to answer the following questions: (i) What is the effect of ICDs on all-cause mortality in patients with CKD at high risk of SCD? (ii) What is the effect of CKD on all-cause mortality in patients who had an ICD implanted, whether for primary or secondary prevention?

Methods

Our analysis is based on the guidelines of the Meta-analysis of Observational Studies in the Epidemiology Group.11

Inclusion criteria

We included prospective or retrospective observational studies with a primary objective to analyse the association between ICD implantation and all-cause mortality in CKD patients at high risk of SCD. Titles and abstracts were evaluated and rejected after initial screening according to the following exclusion criteria: (i) no control group; (ii) no evaluation of renal function; (iii) a primary outcome that is different than all-cause mortality; (iv) publication only in the abstract form; and (v) follow-up duration <1 year.

Search strategies

We searched MEDLINE (1966–2013) and EMBASE (1966–2013) databases to identify relevant studies. We used the following keywords: ‘implantable cardioverter defibrillator’, ‘chronic kidney disease’, ‘severe renal insufficiency’, ‘mortality’, ‘end-stage renal disease’, ‘dialysis’, ‘haemodialysis’, and ‘peritoneal dialysis’. In addition, the ‘Related Articles’ feature on PubMed was used and a manual search was conducted using bibliographies of review articles on this topic. Abstracts of the articles published by the American College of Cardiology, the American Heart Association, the European Society of Cardiology, and the North American Society of Pacing and Electrophysiology were also searched. Titles and abstracts were reviewed independently by two reviewers (N.M. and P.D.S.). Differences were resolved by consensus.

Quality assessment and data extraction

The quality of each study was evaluated according to the guidelines developed by the United States Preventive Task Force and the Evidence-Based Medicine Working Group.12,13 The following characteristics were assessed: (i) clear inclusion and exclusion criteria; (ii) study sample representative of the population; (iii) explanation of sample selection; (iv) full specification of clinical and demographic variables; (v) Follow-up at least 1 year; (vi) reporting loss of follow-up; (vii) definition of CKD; (viii) clear definition of outcomes and outcome assessment; (ix) temporality with renal function measured at baseline and not at outcome assessment; and (x) adjustment of possible confounders in multivariate analysis. Studies were graded as poor if they met <5 criteria, fair if they met 5–7 criteria, and good if they met ≥8 criteria.

Two reviewers (N.M. and P.D.S.) extracted the following data elements: (i) publication details including first author's last name, year; (ii) study design; (iii) characteristics of the study population including gender, race, mean age, comorbidities including hypertension, diabetes; (iv) variables included in the multivariate analyses; and (v) adjusted hazard ratio (HR) or odds ratio (OR) with 95% confidence interval (CI) from the multivariate analyses. Three studies reported outcomes in separate subgroups. Hage et al.14 considered patients who had an ICD implanted for primary prevention and secondary prevention separately. Alla et al.15 considered those with a GFR of 30–60 mL/min/1.73 m2 and those with end-stage renal disease on dialysis. Hager et al.16 analysed patients with CKD stages 3, 4, and 5 as three separate subgroups. In these studies, outcome measures were extracted separately in accordance with guidelines of meta-analysis of observational studies regarding analysing subgroups; this was similarly done in a prior meta-analysis by Korantzopoulos et al.9

Definitions

As we were analysing published studies, we depended on categorization of CKD available in the studies. Chronic kidney disease was defined in various studies using different measures: estimated GFR using the modification of diet in Renal Disease equation17 or the Cockcroft–Gault formula with a cut-off value of 60 mL/min/1.73 m2, serum creatinine with cut-off values that ranged from 1.5 to 2 mg/dL, and dialysis dependence. Severe CKD was defined as having a GFR < 35 mL/min/1.73 m2, while end-stage renal disease was defined as being dialysis dependent. Implantable cardioverter defibrillator use for primary prevention was considered present if a patient had risk of SCD according to the current ACC/AHA guidelines but had not yet had an arrhythmic event. Secondary prevention was considered present if a patient had resuscitated cardiac arrest due to ventricular tachycardia or ventricular fibrillation or had poorly tolerated ventricular tachycardia that was not due to a reversible cause.

Statistical analysis

The degree of association between CKD and all-cause mortality in patients with an ICD was measured in terms of HR and OR. All the studies employed Cox proportional hazard models to examine associations of CKD and mortality, thus for the present analyses we assumed ORs to be a valid approximation of HRs, thereby enabling the use of one consistent measure throughout. A subgroup analysis, which excluded ORs, was also performed showing results consistent with the findings below. HRs and ORs were transformed logarithmically since they do not follow a normal distribution. The standard error was calculated from Log HR and Log OR and the corresponding 95% CI. We used the inverse variance method to achieve a weighted estimate of the combined overall effect. Risk estimates (HRs) were extracted as adjusted HRs from the included studies. These studies reported use of multivariate and propensity score models to adjust for potential confounders such as age, race, ejection fraction, gender, coronary disease, hypertension, diabetes, antiarrhythmic drugs, and β-blockers. We assessed the results for heterogeneity in our analysis by examining the forest plots and then calculating a Q statistic, which we compared with the I2 index.18 The Q test indicates the statistical significance of the homogeneity hypothesis and the I2 index measures the extent of the heterogeneity. We considered the presence of significant heterogeneity at the 5% level of significance (for the Q test) and values of I2 exceeding 56% as an indicator of significant heterogeneity.18,19 Both analyses exhibited significant heterogeneity (Q test in both comparisons was associated with P < 0.05; I2 was 60.32% for our primary comparison and 64.36% for our secondary comparison). This prompted us to adopt the random effect model.11,19 This model allows a distribution of the true effect size rather than assuming one true effect size.11 It takes into account within-study and between-study variance. The underlying heterogeneity further prompted us to perform meta-regression analysis to investigate if our study outcome (all-cause mortality) is affected by factors other than our primary treatment (ICD).1,20 We adopted a weighted regression random effect model and estimated between-study variance (Tau square τ2) using empirical Bayes estimate.20 A two-sided P value <0.05 was regarded as significant for all analyses. Data were then analysed using Microsoft Excel version 2010 and represented as forest plots. Potential publication bias was assessed with the Egger test and represented graphically with Begg funnel plots of the natural log of the HR vs. its standard error.21

Results

The literature search yielded 55 potential studies. After screening the titles and abstracts, 25 studies were excluded because they were either review articles or did not satisfy our inclusion criteria. Out of the 30 articles selected for detailed evaluation, 10 were excluded because they either did not report mortality as an outcome or did not have a clear definition of CKD or did not report adjusted risk estimates. For our primary comparison evaluating effect of ICD on patients with severe CKD, we included five studies of 17 460 patients. All the patients had either severe CKD10 or end-stage renal disease22 and all the studies were retrospective. The baseline characteristics of the included studies for our primary comparison are shown in Table 1. The risk estimates of all-cause mortality in patients with CKD at high risk of SCD were extracted as adjusted HRs from these studies, which employed propensity score matching to limit the effect of confounders like age, gender, ejection fraction, and comorbidities. Patients who had severe CKD or were on dialysis and were ICD recipients (total of 2584 patients) were matched to a similar cohort of patients who did not receive an ICD (total of 14 606 patients). There was no significant difference in overall baseline traits between ICD recipients and matched CKD controls in our primary comparison as shown in Table 2. The analysis indicated that CKD patients with ICD implants were at a reduced risk for total mortality vs. a similar cohort not undergoing ICD implant [HR = 0.650, 95% CI (0.466–0.907); P < 0.05] (Figure 1). Advanced CKD is an important risk factor for non-arrhythmic death, thereby potentially reducing the survival benefit of ICD disparately in CKD patients who receive it for primary prevention more than for those who receive it for secondary prevention. Therefore, we analysed the effect of indication for ICD (primary vs. secondary prevention) on the outcome to assess if the benefit of an ICD was mostly derived from studies with secondary preventative ICD patients. Meta-regression analysis showed that the effect size of all-cause mortality in the studies did not significantly interact with the independent variable: percent primary prevention ICD patients (Figure 2).

View this table:
Table 1

Baseline characteristics of the included studies for our primary comparison evaluating the effect of ICDs in patients with severe CKD

Study or subgroupStudy designTotal patientsHazard ratio (95% confidence interval)ICD indicationMen (%)Mean age (years)EF (%)YearF/U (months)Quality
Charytan et al.23Retrospective11 1600.86 (0.81–0.91)Secondary706527201117Good
Herzog et al.24Retrospective60420.58 (0.50–0.66)Secondary476726200618Good
Khan et al.25Retrospective780.23 (0.06–0.85)Primary736624201031Good
aMADIT-II26Retrospective801.09 (0.49–2.43)Primary716828200620Good
Hiremath et al.22Retrospective1000.40 (0.19–0.82)Both847029201048Good
Overall17 4600.65 (0.47–0.91)
  • There were a total of 17 460 patients with all the studies being retrospective. Mean follow-up was 27 months while median follow-up was 20 months. Mean age was 66 years while mean ejection fraction was 27%.

  • ICD, implantable cardioverter defibrillator; CKD, chronic kidney disease; GFR, glomerular filtration rate; EF, ejection fraction; F/U, follow-up.

  • aMADIT-II: we only included a subgroup that had severe CKD (defined as <35 mL/min/1.73 m2).

View this table:
Table 2

Overall baseline characteristics of ICD recipients with CKD compared with matched CKD controls for our primary comparison evaluating the effect of ICDs in patients with severe CKD

Baseline characteristicsICD recipients with CKDMatched CKD controlsP value
Number of participants285414 606
Age (years)65 ± 1267 ± 11>0.05
Male (%)6661>0.05
Caucasian (%)6562>0.05
Ejection fraction (%)26 ± 1128 ± 7>0.05
Comorbid conditions
Myocardial infarction (%)3531>0.05
Coronary artery disease (%)8581>0.05
Heart failure (%)8177>0.05
Diabetes (%)6662>0.05
  • ICD, implantable cardioverter defibrillator; CKD, chronic kidney disease.

Figure 1

Forest plot of included studies for our primary comparison evaluating the effect of ICD in patients with severe CKD. Implantable cardioverter defibrillator was associated with reduced total mortality in patients at risk of SCD [HR = 0.650, 95% CI (0.466–0.907); P < 0.05]. *Studies were heterogeneous (Q test, P < 0.05) with an I2= 60.32% ± 3.58%. *MADIT-II: we only included a subgroup that had severe CKD (defined as <35 mL/min/1.73 m2). *ICD, implantable cardioverter defibrillator; CKD, chronic kidney disease; CI, confidence interval .

Figure 2

Plot representing results of our meta-regression analysis of all-cause mortality as an effect of the independent variable (percent primary prevention) for studies included in our primary comparison, which evaluates the effect of ICD on all-cause mortality in patients with severe CKD. The independent variable (percent primary prevention) is plotted on the horizontal axis against effect size (log HR). The data points are then plotted according to original data input. However, the size of the circular dot is calculated as follows: (i) as the area of the dot represents the weight, the radius of the dot is calculated as sqrt (wt); (ii) the size of the dot therefore represents the multiple of the smallest effect size in the meta-analysis dataset. It is important to note that there are five circles on the plot with one of the studies represented as a small dot on the horizontal axis at an abscissa equivalent to 100%.

Among the 20 included studies, we identified 15 reports with a total of 5233 patients (1843 patients with CKD vs. 3390 patients without CKD) for our secondary comparison evaluating effect of CKD on patients who had already received an ICD. Two of the studies were prospective; the rest were retrospective. Patient follow-up ranged from 1 to 5 years. The characteristics of the studies are included in Table 3. It was noted that, despite an ICD implant (for primary or secondary prevention), CKD patients had a higher mortality vs. similar patients without CKD [HR = 2.857, 95% CI (1.912–4.269); P < 0.05] (Figure 3). Meta-regression of all-cause mortality by percent primary prevention ICD patients in these studies failed to show any significant interaction (see Supplementary material online, Figure S1)

View this table:
Table 3

Baseline characteristics of the included studies for our second comparison evaluating effect of CKD on all-cause mortality in patients who had already received an ICD

Study or SubgroupStudy DesignTotal patientsHazard ratio (95% confidence interval)ICD IndicationMen (%)Mean age (years)EF (%)YearF/U (months)Quality
Alla [1]Retrospective321.81 (1.18–2.78)Both886932201036Good
Alla [2]Retrospective1884.63 (3.02–7.09)Both837328201036Good
Bruch et al.27Prospective1463.55 (1.03–12.2)Both8061 ± 1329 ± 9200722Good
Cheema et al.28Retrospective4410.99 (0.979–0.997)Both776726201040Good
Chen-Scarabelli et al.29Retrospective3362.70 (1.42–5.31)Both1006727 ± 13200730Good
Cuculich et al.30Retrospective22910.50 (4.8–23.1)Primary846725200718Good
Eckart31Retrospective7411.69 (1.32–2.17)Primary5996830200631Fair
Hage et al.14 [1]Retrospective4092.08 (1.34–3.23)Primary7058 ± 1226 ± 12201150Good
Hage et al.14 [2]Retrospective2871.27 (0.81–2.00)Secondary735930 ± 14201150Good
Hager et al.16 [1]Retrospective3451.40 (0.7–2.6)PrimaryNR71 ± 1026.7 ± 7.9201048Good
Hager et al.16 [2]Retrospective543.10 (1.5–6.3)PrimaryNR72 ± 9.427.2 ± 8.1201048Good
Hager et al.16 [3]Retrospective1310.20 (4.2–24.6)PrimaryNR65 ± 7.527.9 ± 6.5201048Good
Koplan et al.32Retrospective1072.00 (1.10–3.50)Both788231 ± 10200640Fair
Levy et al.17Retrospective3468.82 (3.06–25.39)Both806530200842Good
Parkash et al.33Retrospective2283.83 (1.84–7.96)Both796535 ± 16200612Good
Robin et al.34Retrospective5854.23 (1.84–9.74)Both796333 ± 15200626Good
Schefer et al.35Prospective1765.87 (2.48–13.86)Both785340 ± 18200851Good
Turakhia et al.36Retrospective4771.70 (1.08–10.29)Both746131200848Good
Wase et al.37Retrospective932.00 (1.10–3.50)Both716729200442Good
Overall52332.86 (1.91–4.27)
  • There was a total of 5233 patients with only two studies being prospective; the rest of the studies being retrospective. Mean follow-up was 38 months while median follow-up was 40 months. Mean age was 66 years and mean ejection fraction was 30%.

  • Hage included two subgroups: Hage et al. [1] for primary prevention and Hage et al. [2] for secondary prevention; Alla included two subgroups: Alla [1] for CKD with an estimated GFR between 30 and 60 and Alla [2] for end-stage renal disease on dialysis; Hager included three subgroups: Hager et al. [1], Hager et al. [2], and Hager et al. [3] for CKD stages 3, 4, and 5, respectively.

  • ICD, implantable cardioverter defibrillator; CKD, chronic kidney disease; GFR, glomerular filtration rate; EF, ejection fraction; F/U, follow-up.

Figure 3

Forest plot of the included studies for our secondary comparison evaluating all-cause mortality associated with CKD in patients who already had ICD implanted. Chronic kidney disease was associated with increased total mortality despite presence of an ICD [HR = 2.857, 95% CI (1.912–4.269); P < 0.05]. *Studies were noted to be heterogeneous (Q test, P < 0.05) with an I2 = 64.36% ± 5.32%. *ICD, implantable cardioverter defibrillator; CI, confidence interval.

Publication bias

There was no evidence of publication bias for the included studies that assessed all-cause mortality by visual inspection of the funnel plot and by using the Egger test21 (P = 0.08) (Figure 4).

Figure 4

This funnel plot is for studies included in our secondary comparison, evaluating effect of CKD on all-cause mortality in patients who already had ICD implanted. Treatment effect (log HR) is plotted on the horizontal axis against a measure of study size (standard error, SE). Based on visual inspection, the plot is symmetric with largest studies being near the average, and with small studies being spread on both sides of the average. Therefore, there is no evidence of publication bias. This is confirmed by using Egger's asymmetry test (P = 0.08).

Discussion

Our meta-analysis shows that an ICD implant is associated with a survival benefit in CKD patients at high risk of SCD. We also demonstrate that CKD in ICD patients is associated with an increase in all-cause mortality.

There has been considerable controversy regarding utilization of ICDs in patients with heart failure and/or SCD patients who have coexistent CKD because benefit has not been definitively shown in this population. Current guidelines for ICD use in patients with CKD have largely been extrapolated from the ICD clinical trials, which have mostly excluded patients with severe CKD and/or end-stage renal disease.23

Two meta-analyses have evaluated the effect of ICDs in CKD patients.3,9 Sakhuja et al.3 included only seven studies with 2516 patients and reported outcomes in haemodialysis patients with and without ICDs. The authors reported a 2.7-fold increase in mortality in haemodialysis patients after ICD placement. This study did not adjust for confounding factors, such as comorbid health conditions, age, and gender that could potentially affect outcomes. Korantzopoulos et al.9 published a comprehensive meta-analysis (11 studies with 3010 patients) showing that CKD (vs. no CKD) is associated with higher mortality for patients who had an ICD implant for primary or secondary prevention. Our study updates this report by including 4 recently published studies (15 studies total), adding roughly 2000 patients more to the analysis.

Our meta-analysis addresses a question relevant to the current practice guidelines: What is the effect of an ICD on all-cause mortality in patients with CKD at high risk of SCD? Our study shows that ICDs can reduce mortality in patients with CKD at high risk of SCD even though there may be a greater risk of dying from non-arrhythmic causes than ICD patients who do not have CKD.

Chronic kidney disease causes autonomic and electrolyte imbalance, provokes exaggerated inflammatory response to the myocardium, and alters myocardial structure rendering a proarrhythmic state.20 Rates of SCD progressively increase with decreasing GFR.4 The incidence of SCD is highest in patients on haemodialysis and peaks in the 12 h after dialysis and the last 12 h of the longest period without dialysis every week.38 This is partly explained by decrease in myocardial perfusion in the immediate post-dialysis period 39 and an increase in corrected QT dispersion possibly caused by electrolyte shifts after haemodialysis.40 Consequently, this may increase the likelihood of defibrillation-resistant arrhythmias.

No previous report addresses the larger question of primary prevention of SCD in the CKD population. Our meta-analysis indicates that ICD implantation is associated with reduction in all-cause mortality in CKD patients at high risk of SCD.

Limitations

Studies in this meta-analysis included patients who received ICDs for primary and secondary prevention indications. A paucity of studies in the primary comparison limited our ability to analyse the effect of ICD on total mortality in CKD patients with each of these indications separately. While the meta-regression analysis does not show significant effect of ICD indication on the outcome, the mortality benefit of ICD in CKD patients may be derived mostly from studies with patients who received ICDs for secondary prevention and further large randomized studies are needed in CKD patients who qualify for a primary prevention ICD. The definition of CKD was inconsistent among studies. Of the studies analysed, four used elevated serum creatinine29,30,33,36 and two used only dialysis dependence as a marker for CKD,15,34 thus underestimating CKD in the study population. The remaining nine studies used reduced GFR as marker of CKD. The major limitation of our analysis is that it is based on an overwhelming predominance of retrospective studies adding significant heterogeneity to the studies considered. All the studies included in our primary comparison were retrospective, while 13 out of 15 (∼87%) studies for our secondary comparison were retrospective. The major limitations of retrospective review include selection bias, inability to establish a temporal relationship, inability to control exposure, and difficulty in making accurate comparisons between the exposed and the non-exposed groups. In retrospective studies, it may be that the sickest patients were not given ICDs and, therefore, the control groups were inadequate. Another major limitation is difficulty in deriving any definitive conclusion from observational studies as confounding factors remain a major limitation. The potential overlap between the populations may additionally limit the study. Specifically, Charytan et al.23 and Herzog et al.24 both included Medicare patients enroled in the USRDS. Finally, the cause of death in these patients remains unknown.

Conclusion and clinical significance

Implantable cardioverter defibrillator use is associated with reduction in all-cause mortality in patients with CKD at high risk of SCD even though there is risk of dying from non-arrhythmic causes. Despite the paucity of randomized trials in the CKD population, these data support use of ICD for prevention of SCD in patients at risk. This meta-analysis addresses an important clinical issue that remains difficult to assess due to a multitude of issues including financial and ethical concerns and underscores the importance of further review of appropriate use of life-saving devices in a population of patients that is ever growing in every part of the world.

Supplementary material

Supplementary material is available at Europace online.

Conflict of interest: none declared.

References

View Abstract