Implantable Defibrillators for the Prevention of Mortality in Patients With Nonischemic Cardiomyopathy:  A Meta-analysis of Randomized Controlled Trials FREE

Several randomized clinical trials have demonstrated that implantable cardioverter defibrillators (ICDs) reduce mortality in high-risk subgroups of patients with coronary artery disease.^1- 7 However, nearly 10% to 15% of instances of sudden cardiac death occur in patients with left ventricular dysfunction but no evidence of prior myocardial infarction. While sudden cardiac death in patients with prior myocardial infarction is thought mainly to be the result of reentrant ventricular arrhythmias originating from the subendocardial surface of infarcted myocardium, the mechanism of sudden cardiac death in patients with nonischemic cardiomyopathy (NICM) is less understood.⁸ Identification of patients with NICM at high risk for sudden death is complicated by the inadequate predictive accuracy of both electrophysiologic studies and noninvasive tests in this patient population.^9- 10

No single prospective randomized controlled trial of ICD therapy in NICM has demonstrated convincing evidence of mortality reduction. Observational studies suggest that up to 30% of deaths in patients with NICM are sudden.¹¹ Mortality in medically treated patients with NICM and a prior history of syncope may exceed 30%¹² at 2 years, while those treated with an ICD experience a high frequency of appropriate device therapies.¹³ Two primary prevention trials^14- 15 have demonstrated a trend toward reduced mortality in the subgroup of ICD-treated patients with NICM, but they failed to achieve statistical significance. Two additional small randomized controlled trials of ICD in primary prevention among patients with NICM showed no benefit over therapy with antiarrhythmic drugs.^16- 17

The appropriate role of ICD therapy in the NICM population remains unclear. We performed a systematic review and meta-analysis of randomized controlled trials of ICD therapy for prevention of all-cause mortality in NICM.

All prospective randomized controlled trials of ICD vs medical therapy enrolling patients with NICM were identified using a 2-level search strategy. First, public domain databases including MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials were searched using Web-based search engines (PubMed, OVID). Second, relevant studies were identified through a manual search of secondary sources including references of initially identified articles and a search of reviews, commentaries, and proceedings from national cardiology meetings from 2003-2004. All references were downloaded for consolidation, elimination of duplicates, and further analysis.

The MEDLINE database was searched from January 1966 to April 2004. MeSH keywords included defibrillators, implantable; randomized controlled trials; andclinical trials. A previously developed MEDLINE research methodology filter¹⁸ designed to retrieve scientifically strong studies of treatment was also used with the keyword defibrillator and optimizing for a sensitive rather than specific search. The Cochrane Library and Central Register of Controlled Trials (current through the first quarter of 2004) was searched using OVID exploding keywords including defibrillator, clinical trial, randomized clinical trial, andsudden death. The EMBASE database was searched from January 1991 to the second quarter of 2004 using OVID, exploding keywords including defibrillator, clinical trial, andsudden death.

Studies considered for inclusion met the following criteria: the design was a prospective randomized controlled clinical trial; the study population was patients with resuscitated cardiac arrest, those with documented or symptomatic sustained ventricular tachyarrhythmia, or those with depressed left ventricular function deemed to be at high risk for developing lethal cardiac arrhythmia; patients were randomly assigned to ICD vs medical therapy; and main outcomes included all-cause mortality, cardiac death, or arrhythmic mortality. A QUOROM¹⁹ flow diagram of the study selection process is illustrated in Figure 1.

Because the detailed, final results of SCD-HeFT are not yet published, we relied on the preliminary results as reported by the investigators at a recent national cardiology meeting.¹⁴

All qualifying studies were assessed for adequate blinding of randomization, completeness of follow-up, and objectivity of the outcome assessment. Blinding of the intervention was not used for quality assessment because the intervention (ICD implantation) was surgical. Data regarding detailed inclusion criteria, ICD device type, duration of follow-up, rates of crossover, all-cause mortality, and arrhythmic death were abstracted (as available) from each individual study.

We conducted a meta-analysis of summary statistics from the individual trials because detailed, patient-level data were not available for all trials. For each study, data regarding all-cause mortality in both the ICD and control groups were used to generate risk ratios (RRs) and 95% confidence intervals (CIs). A priori, the decision was made to analyze the primary and secondary prevention trials separately due to anticipated heterogeneity in the baseline clinical risk of these populations. Study-specific estimates were combined using inverse variance-weighted averages of logarithmic RRs in both fixed- and random-effects models. Between-study heterogeneity was analyzed by means of standard χ² tests, with P≤.10 deemed statistically significant. Where no significant statistical heterogeneity was identified, the fixed-effect estimate was used preferentially as the summary measure. Sensitivity analyses were performed to assess the contribution of each study to the pooled estimate by excluding individual trials one at a time and recalculating the pooled RR estimates for the remaining studies. To assess the impact of differential length of follow-up among the primary prevention studies on the pooled estimate, the effects of ICD therapy on all-cause mortality were explored separately in trials with follow-up of less than 3 years and 3 or more years. Publication bias was assessed graphically using a funnel plot and mathematically using an adjusted rank-correlation test, according to the method of Begg and Mazumdar.²⁰ All analyses were conducted using STATA version 8.2 (Stata Corp, College Station, Tex).

As outlined in Figure 1, our search identified 8 prospective randomized controlled clinical trials of ICD vs medical therapy enrolling at least some patients with NICM. These included 3 trials of ICD therapy for secondary prevention (enrolling patients with previous resuscitated cardiac arrest or symptomatic ventricular tachycardia) and 5 trials of ICD therapy for primary prevention.

Secondary Prevention Trials. The analyzed secondary prevention trials were the Antiarrhythmics vs Implantable Defibrillators (AVID)¹ trial, the Cardiac Arrest Study Hamburg (CASH),³ and the Canadian Implantable Defibrillator Study (CIDS).² All trials included an active antiarrhythmic drug control group, with most non-ICD patients receiving amiodarone (AVID, CIDS). Patients in the CASH study were initially randomized to ICD or 1 of 3 antiarrhythmic drugs: amiodarone, metoprolol, or propafenone. The propafenone arm was terminated early due to evidence of increased mortality, and the final analysis pooled the amiodarone and metoprolol arms for comparison with ICD.³

None of the secondary prevention trials exclusively enrolled patients with NICM. Across all 3 trials enrolling a total of 1963 patients, only 292 patients (14.8%) were classified as having NICM. The AVID and CIDS trials reported a nonsignificant reduction in all-cause mortality with ICD therapy in the subgroup of patients with NICM,^1- 2 but independent outcomes for this subgroup were not reported in CASH.

The baseline patient characteristics for the secondary prevention trials are summarized in the Table. While the AVID and CIDS trials enrolled patients with similar characteristics in a comparable trial design, the CASH trial was different: the mean left ventricular ejection fraction in CASH was higher and, distinct from the other trials, more than 50% of patients in CASH received epicardial ICD systems rather than modern transvenous systems. Ultimately, the CASH trial was excluded from the pooled analysis because of significant qualitative heterogeneity from the other trials, the small number of patients with NICM enrolled, and the lack of reported estimates for ICD efficacy in the subgroup of patients with NICM.

Primary Prevention Trials. The 5 primary prevention trials analyzed were the Cardiomyopathy Trial (CAT),¹⁶ the Amiodarone vs Implantable Cardioverter-Defibrillator Randomized Trial (AMIOVIRT),¹⁷ the Defibrillator in Nonischemic Cardiomyopathy Treatment Evaluation (DEFINITE) Trial,¹⁵ the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT),¹⁴ and the Comparison of Medical Therapy, Pacing, and Defibrillation in Heart Failure (COMPANION) Trial.²¹

There was substantial qualitative heterogeneity in trial design. The CAT, AMIOVIRT, and DEFINITE trials exclusively enrolled patients with NICM, while SCD-HeFT and COMPANION enrolled patients with both ischemic cardiomyopathy and NICM. CAT, AMIOVIRT, and DEFINITE randomized patients to ICD or medical therapy in a 2-arm design, but SCD-HeFT and COMPANION were more complex.

The SCD-HeFT trial randomized patients with congestive heart failure and left ventricular ejection fraction of 35% or less to therapy with ICD, amiodarone, or placebo; a total of 1676 patients were randomized to either ICD or placebo, of whom 792 (47.3%) had NICM. Since data were reported for the efficacy of ICD relative to placebo in the subgroup of patients with NICM (hazard ratio for all-cause death, 0.73; 95% CI, 0.50-1.04),¹⁴ only this subset of patients was included in the analysis. The COMPANION trial randomized patients with heart failure, ICM or NICM, and QRS duration greater than 120 milliseconds in a 1:2:2 ratio to receive optimal pharmacologic therapy alone or in combination with cardiac resynchronization therapy using either a pacemaker or pacemaker-defibrillator. Of the 1520 patients randomized in the trial, 903 were allocated to either the medical therapy or defibrillator arms; of this subset, 397 patients (44%) had NICM. Cardiac resynchronization with a pacemaker-defibrillator reduced all-cause mortality relative to pharmacologic therapy alone in patients with NICM (hazard ratio for all-cause death, 0.50; 95% CI, 0.29-0.88; P = .015),²¹ and this estimate was used in the pooled analysis.

In total, our meta-analysis included data on 1854 patients with NICM randomized to therapy with ICD or placebo in a primary prevention setting. The baseline characteristics for the patients enrolled in each trial are summarized in the Table. The most notable difference between the trials was the severity of illness of enrolled patients. AMIOVIRT, DEFINITE, and SCD-HeFT enrolled a comparable number of patients with heart failure of at least 3 years’ duration and New York Heart Association (NYHA) class III or IV symptoms. In contrast, the CAT trial enrolled patients with heart failure of short duration (<9 months) and low control group mortality at 1 year (3.7%), suggesting that some of these patients may have had transient ventricular dysfunction. The COMPANION trial enrolled only NYHA class III/IV patients with prolonged QRS duration by design and had a predictably higher control group mortality (19%) at 1 year.

Despite the noted heterogeneity in design between the primary prevention trials, there was sufficient similarity between the populations and the hypotheses to merit inclusion of all 5 trials in the quantitative meta-analysis. The most dissimilar trials (CAT, COMPANION) were sequentially eliminated in sensitivity analyses to assess their impact on the pooled effect estimate.

All 8 of the trials included in the analysis were of comparable quality. Although all trials were randomized, the method of allocation generation was not described in detail for any of the studies. Allocation concealment and blinding were not possible in these trials given that the intervention (ICD implantation) was surgical; however, all trials used either an independent or blinded committee for adjudication of events. Careful accounting for dropouts and crossovers was provided in nearly all cases, and all analyses were conducted by intention-to-treat. While wide variations in crossover rates were noted in the trials, it was anticipated that crossovers would decrease the calculated benefit of ICD therapy relative to medical therapy and bias results to the null in an intention-to-treat analysis. Summary measures of methodological quality for each trial are outlined in the Table. Funding for the majority of the trials was provided, at least in part, by the company whose device was the subject of study.

Secondary Prevention Trials. Of the secondary prevention trials, only AVID and CIDS reported estimates of ICD efficacy in the subgroup of patients with NICM. Individually, while the point estimates in both trials favored ICD therapy over therapy with antiarrhythmic drugs, the results were statistically nonsignificant. When data from these 2 secondary prevention trials were pooled using a fixed-effects model (n = 256 patients with NICM), ICD therapy was associated with a 31% reduction in all-cause mortality relative to medical therapy that remained statistically nonsignificant (RR, 0.69; 95% CI, 0.39-1.24; P = .22 [Figure 2]). There was minimal trial heterogeneity and accordingly little difference in the pooled result from random-effects modeling.

Primary Prevention Trials. Four of the 5 individual primary prevention trials demonstrated a statistically nonsignificant benefit of ICD over medical therapy for all-cause mortality in NICM patients. Only the COMPANION trial, which used a device with both cardiac resynchronization and defibrillation capabilities, demonstrated a statistically significant all-cause mortality reduction with ICD over optimal medical therapy.

Pooled analysis of the 5 primary prevention trials (representing 1854 patients with NICM) demonstrated a statistically significant 31% reduction in all-cause mortality with ICD relative to medical therapy in fixed-effect models (RR, 0.69; 95% CI, 0.55-0.87; P = .002 [Figure 3]). As with the secondary prevention trials, there was minimal trial heterogeneity of results. To assess the impact of qualitative heterogeneity in trial design and patient selection on the pooled effect estimate, we performed several sensitivity analyses. First, we excluded the COMPANION trial, which enrolled a population with more advanced heart failure and used a device capable of both cardiac resynchronization and defibrillation; combining the remaining 4 trials generated an attenuated but still statistically significant result favoring ICD therapy (fixed-effects RR, 0.74; 95% CI, 0.58-0.96; P = .02). Second, we excluded the CAT study, which enrolled patients with heart failure of short duration and low control group mortality. Without CAT or COMPANION, there was still a statistically significant benefit for ICD therapy in pooled analysis of the remaining 3 trials (fixed-effects RR, 0.73; 95% CI, 0.55-0.96; P = .03). In general, exclusion of any single trial from the analysis did not substantively alter the overall result of our analysis. Additionally, eliminating the primary prevention trials with long (≥3 years) follow-up did not substantially change the pooled point estimate (fixed-effects RR, 0.63; 95% CI, 0.46-0.88; P = .006).

Analysis of the 7 primary and secondary prevention trials in combination demonstrated a statistically significant 31% survival benefit for ICD over medical therapy (RR, 0.69; 95% CI, 0.56-0.86; P  = .002). Again, there was minimal heterogeneity of results and the finding of reduced death risk was insensitive to exclusion of any single trial.

To assess publication bias we generated a funnel plot of the logarithm of effect size vs the standard error for each trial. There was no evidence of significant publication bias (P = .81 by Begg adjusted rank-correlation test).

The results of our analysis suggest that ICD therapy may reduce all-cause mortality by 31% over medical therapy in patients with NICM. The results from primary prevention are the most compelling, with data from 1854 patients with NICM randomized in 5 different trials demonstrating a 31% reduction in all-cause mortality with ICD therapy relative to medical therapy. This result was robust in sensitivity analyses, even accounting for possible bias introduced by the inclusion of COMPANION, which randomized patients to a device capable of both cardiac resynchronization and defibrillation rather than ICD. No individual trial was able to demonstrate a statistically significant benefit to ICD therapy, likely due to lower than anticipated control group mortality and systematic underpowering of these trials in the design phase.

Low mortality among medically treated patients in the primary prevention setting also implies a relatively small absolute mortality benefit to ICD therapy in NICM. Assuming a mortality of approximately 7% per year (the averaged control group mortality from CAT, AMIOVIRT, DEFINITE, and SCD-HeFT, affirmed in the recently published Marburg Cardiomyopathy Study¹⁰) and a 31% relative risk reduction, ICD therapy could result in an absolute reduction of approximately 2% per year in all-cause mortality. At this rate, 25 patients with NICM would need to be treated to prevent one death at 2 years. By comparison, in patients with ischemic cardiomyopathy, the number needed to treat to achieve the same primary prevention benefit is approximately 18.⁵

The data from randomized controlled secondary prevention trials are less robust, primarily due to the small number of enrolled patients with NICM. Though underpowered (as reflected in the wide CIs and lack of statistical significance), the observed 31% reduction in mortality in the patients receiving ICD is consistent with a previous meta-analysis⁷ of secondary prevention trials demonstrating a 28% reduction in mortality with ICD therapy in unselected patients (both ischemic and NICM). Furthermore, our conclusions are consistent with the results of prior uncontrolled observational studies in secondary prevention suggesting that patients with NICM and prior unexplained syncope or resuscitated cardiac arrest are at substantial risk for sudden cardiac death.^{12- 13,22- 26} While few patients with NICM have been randomized in the secondary prevention setting, the unacceptable mortality associated with medical therapy in these patients would likely make design of an additional randomized controlled trial ethically problematic.

Our analysis must be viewed in the context of its limitations. First, we used only data from randomized controlled trials. Patients enrolled in randomized trials may not be representative of patients typically seen in clinical practice. However, because randomized trials balance both known and unknown confounders across treatment groups, this is the study design least vulnerable to bias.

Second, our results may be influenced by a publication bias favoring ICD therapy. This risk was minimized through an exhaustive search of the available literature and through the inclusion of an unpublished study (SCD-HeFT) in our analysis. Though the statistical tests did not indicate publication bias, there is clearly limited power to detect such bias, given the small number of studies examined.

Third, our pooled analysis combines the results from clinical trials that enrolled patients during a period of rapid evolution in the efficacy of medical therapy in improving survival among patients with NICM. It is this evolution that is in large part likely responsible for the observed reductions in control group mortality seen in the more recent ICD primary prevention trials. Of the trials analyzed, only COMPANION reported widespread use of aldosterone-receptor antagonists (≈ 50%) in the medical therapy group. In addition, rates of β-blocker use were low in both CAT and AMIOVIRT, and this may have led both trials to overestimate the benefit of ICD therapy.

Despite these acknowledged limitations, we found that, based on a systematic review and meta-analysis, ICD therapy is likely effective in both primary and secondary prevention of all-cause mortality in the subgroup of patients with NICM, and ICD should be considered for patients who meet the criteria for enrollment in the randomized trials discussed above. It is important to emphasize, however, that mortality reduction may not imply greater clinical benefit among patients with advanced heart failure. The decision regarding ICD implantation must be influenced by careful consideration of patient preferences and quality of life. Progressively worsening functional status may lead some patients, in concert with their physicians, to engage advance directives that elect against resuscitation, necessitating ICD deactivation. ICD implantation, while effective in reducing sudden cardiac death, does not forestall the progression of heart failure.