Implantable Cardioverter Defibrillators in Primary and Secondary Prevention: A Systematic Review of Randomized, Controlled Trials FREE

–The Editors

Sudden cardiac death accounts for approximately 50% of all deaths from cardiovascular causes (1 - 2). Some patients are at higher risk for sudden cardiac death, particularly those with significant coronary artery disease (CAD) and left ventricular systolic dysfunction. Until recently, prevention of sudden cardiac death has focused on antiarrhythmic drugs, β-blockers, and improved management of the underlying disease processes. Several published trials in the past few years have evaluated implantable cardioverter defibrillators (ICDs) in patients with cardiovascular disease. There seems to be little doubt that this intervention should be routinely considered in some patients, such as those with advanced ischemic cardiomyopathy who are resuscitated after ventricular fibrillation arrest. However, debate continues about the potential benefits of ICDs in other patient groups (3 - 6).

Approximately 85% to 90% of sudden cardiac deaths are due to a first arrhythmic event; the remaining 10% to 15% are due to recurrent events (7). We defined primary prevention as prevention of a first life-threatening arrhythmic event (ventricular fibrillation, sustained ventricular tachycardia, or cardiac arrest) (8). Primary prevention of sudden cardiac death routinely focuses on patients at high risk, including those with CAD and left ventricular systolic dysfunction, up to 60% of whom die of dysrhythmia (9). Secondary prevention refers to the prevention of an additional life-threatening arrhythmic event in survivors of sudden cardiac death or patients with recurrent unstable rhythms.

A recent meta-analysis using individual-patient data from three studies of secondary prevention suggested a survival benefit for ICD therapy compared with amiodarone therapy but did not include any data on primary prevention (10). Another recent systematic review involving 1610 patients included trials of primary and secondary prevention published before January 2000 (11); however, two primary prevention trials involving 1336 patients have been published since then (12 - 13). These trials, in turn, focused exclusively on total mortality and did not calculate summary effect estimates or explore reasons for heterogeneity in the total mortality data. Given the limitations of the existing analyses, the potential impact of ICDs on patient survival, and the major socioeconomic implications of this issue, we performed a systematic review of trials of primary and secondary prevention with ICDs to examine the effects of this therapy on rates of sudden cardiac death and all-cause mortality.

We searched for randomized trials in MEDLINE (1980–24 September 2002), the Cochrane Controlled Clinical Trial Registry (2002, Volume 3), EMBASE (1980–2002), Web of Science, National Library of Medicine Gateway, Cardiosource, the Clinical Trials Registry, Clinicaltrials.gov, the CRISP (Computer Retrieval of Information on Scientific Projects) Database, the National Research Register, the Glaxo–Wellcome Clinical Trials Register, the LILACS (Latin American and Caribbean Health Science Literature) Database, OCLC (Online Computer Library Center) ProceedingsFirst, and the National Health Service Economic Evaluation Database. All databases were last accessed on 24 September 2002. In addition, bibliographies of relevant papers were hand searched and experts, device manufacturers, and primary authors were contacted for information on additional trials. Relevant conference proceedings were also searched. The search was not limited by language. We used the following textwords and Medical Subject Headings: ICD, AICD, implantable defibrillators (exp), heart arrest (exp), sudden cardiac death (exp), sudden death (exp), SCD, cardiac arrest, coronary disease (exp), heart disease (exp), systolic dysfunction, ventricular dysfunction (exp), heart failure (exp), ventric* arrhythmia, ventric* rhythm, ventric* fibrillation, ventric* tachycardia, arrhythmia (exp), anti-arrhythmia agents (exp), anti-arrhythmia drug*, anti-arrhythmia therap*, and antiarrhythmi*.

Two of the study investigators independently reviewed the titles and abstracts of all citations to identify any randomized trials evaluating the efficacy of ICDs versus placebo or ICDs versus antiarrhythmic therapy. Both reviewers used standardized data forms to review the full text of potentially relevant articles. A funnel plot was used to evaluate publication bias.

We included any randomized, controlled trials involving patients at risk for sudden cardiac death or ventricular arrhythmia (sustained ventricular tachycardia or ventricular fibrillation) who had evidence of heart failure or CAD (primary prevention), as well as studies in survivors of sudden cardiac death or unstable ventricular rhythm (secondary prevention). The trial outcomes had to include sudden cardiac death or all-cause mortality. Trials in patients with inherited arrhythmic disorders were excluded. We also excluded trials that did not report any of the outcomes of interest or had crossover rates of greater than 50% between study groups.

Intention-to-treat analyses were performed, and the outcome definitions used by the original researchers were accepted. All discrepancies in trial eligibility or data collection were resolved by consensus.

We extracted data on all-cause mortality, sudden cardiac death, total cardiac mortality, and total noncardiac mortality. A priori, we decided to examine the effects of ICD therapy in primary versus secondary prevention. Because we anticipated that the primary prevention trials would encompass a broad spectrum of patients, we subdivided them into those enrolling high-risk patients and those enrolling moderate-risk patients. We defined high-risk patients as those with an expected rate of sudden cardiac death of at least 5% per year (that is, patients with ischemic cardiomyopathy, with or without ventricular arrhythmia) (14).

We used Metaview 4.1 software (Update Software, Oxford, United Kingdom) to calculate summary relative risks (since the outcomes were relatively common) and used the Cochran Q-test to assess for heterogeneity in each outcome of interest. We combined studies using the DerSimonian and Laird random-effects model as well as the Mantel–Haenszel–Peto fixed-effects model; when the results from both models were identical and there was no statistical heterogeneity, we reported only the fixed effects (15). We also conducted sensitivity analyses to examine the effect of year of publication, study quality, and allocation concealment on the results (16 - 17).

The funding sources had no role in the collection, analysis, or interpretation of the data or in the decision to submit the paper for publication.

Of the 385 potentially relevant articles identified in our search, 9 were parallel-group randomized trials. We excluded 1 trial because patients in both study groups received ICD therapy (18). We then analyzed the data from the 8 trials that fulfilled our inclusion criteria (12 - 13,19 - 24)(Figure 1). There was no disagreement about any of the articles selected for final inclusion in the meta-analysis. Funnel-plot analyses did not suggest any marked publication bias.

In the 8 trials we examined, 4909 patients were randomly assigned to study groups. Three of the 8 trials were trials of secondary prevention, and the remaining 5 (3 involving high-risk patients and 2 involving moderate-risk patients) were classified as trials of primary prevention (Table 1). All but 1 study (12) enrolled patients with ischemic heart disease. All trials reported all-cause mortality, and all but 1 (21) used all-cause mortality as the primary outcome. The control groups and crossover rates are described in Table 1, and concomitant medication use is described in Table 2.

The included trials were randomized and controlled. Because of the nature of the intervention, none of the trials were blinded. Randomization and allocation concealment were adequate in all trials. All-cause mortality and sudden cardiac death were reported in all trials, but other end points were not consistently reported. For several trials, we needed to review secondary publications or contact the authors to determine causes of death.

The summary relative risk (RR) for sudden cardiac death was 0.43 (95% CI, 0.35 to 0.53) for all 8 trials. This confirms that ICDs are highly efficacious in preventing sudden cardiac death, both as primary prevention (RR, 0.37 [CI, 0.27 to 0.50]) and secondary prevention (RR, 0.50 [CI, 0.38 to 0.66]) (Figure 2). No appreciable heterogeneity was seen among trials, although no sudden cardiac deaths occurred in either study group in 1 trial (12) because the investigators had recruited lower-risk patients. There was no appreciable difference between Ts of ICD (transthoracic vs. transvenous) in the summary effect estimates for prevention of sudden cardiac death (19 - 20).

Four trials reported a statistically significant survival benefit in ICD-treated patients (Figure 3). The summary relative risk of 0.74 (CI, 0.67 to 0.82) for all-cause mortality in all 8 trials demonstrates a beneficial effect of ICD therapy. Effect estimates were similar in trials of primary and secondary prevention (RR, 0.72 [CI, 0.63 to 0.84] vs. 0.76 [CI, 0.65 to 0.89]) (Figure 3). Random-effects models yielded similar summary relative risks for overall mortality, all-cause mortality in primary prevention, and all-cause mortality in secondary prevention (0.72 [CI, 0.58 to 0.90], 0.69 [CI, 0.46 to 1.03], and 0.77 [CI, 0.65 to 0.91], respectively).

Substantial heterogeneity in total mortality was observed between primary prevention trials enrolling high-risk patients and those enrolling moderate-risk patients (P< 0.001). Indeed, the latter trials failed to demonstrate any survival benefit with ICD therapy. This is not surprising, however, when the differences between patients enrolled in the 3 high-risk trials and the 2 moderate-risk trials are considered. In the 3 trials demonstrating a substantial survival benefit with ICDs, virtually all of the patients had known CAD and left ventricular dysfunction (Table 2); in 2 of these 3 trials, the patients also had inducible ventricular arrhythmias on electrophysiologic testing (21,24). However, while both of the moderate-risk trials enrolled patients with left ventricular systolic dysfunction, 1 included only patients with nonischemic dilated cardiomyopathy but no inducible ventricular arrhythmia and the other enrolled patients after successful coronary artery surgery; in the latter group, the issue of myocardial ischemia was presumably resolved (20).

Five trials (12 - 13,20,22,24) reported total cardiac mortality, with a summary point estimate of 0.81 (CI, 0.69 to 0.96) for relative risk. Noncardiac mortality did not differ between patients receiving ICDs and those receiving medical therapy alone in the 3 trials that reported this outcome (RR, 0.91 [CI, 0.60 to 1.38]) (13,22,24). Perioperative infection rates after ICD implantation ranged from 0.7% to 12.3% in these 8 trials, and rates of lead fracture or device malfunction ranged from 1.8% to 14%. Rates of serious bleeding ranged from 1% to 6%, and pneumothorax occurred in less than 1% of patients. Not unexpectedly, the complication rates were higher for transthoracic ICD. The more recent trials, which used newer ICD models, reported lower complication rates.

The risk difference for total mortality was 0.08 (CI, 0.02 to 0.13) for all included trials, yielding a number needed to treat for benefit of 13 (CI, 8 to 50). However, this summary estimate is inadequate because its calculation varies depending on baseline risk. For example, for primary prevention, 18 patients with severe ischemic cardiomyopathy (left ventricular ejection fraction ≤ 0.3) would need to receive ICDs to prevent one death over the next 2 years (13). However, if ICDs were implanted only in patients with similar left ventricular dysfunction and inducible ventricular arrhythmias, the number needed to treat for benefit would be 4 (21,24).

Implantable cardioverter defibrillators reduce the relative risk for sudden cardiac death by approximately 50%, regardless of baseline risk. However, the effect of ICD therapy on all-cause mortality varies according to baseline risk. We would expect that the impact of any intervention on a combined end point (such as total mortality) would vary if the intervention affects only one component of that end point (such as arrhythmic death). The effect of an intervention on a combined end point will be greater in high-risk persons (in whom a greater proportion of all deaths will be due to arrhythmia) and less dramatic in moderate-risk persons (in whom other causes of death will predominate) (25). As expected, while ICDs reduced all-cause mortality by approximately one third in survivors of cardiac arrest and in high-risk patients who had not yet had an arrest (for example, patients with CAD and severe left ventricular systolic dysfunction), the trials we examined did not show a significant impact on total mortality rates in patients at lower risk for sudden cardiac death (for example, patients with left ventricular systolic dysfunction but no CAD or inducible ventricular arrhythmias).

Since it has been proven that ICDs are efficacious in reducing sudden cardiac deaths, the challenge is to accurately stratify patients by baseline risk to identify those most likely to benefit from ICD therapy. Our data support the policy of considering ICD therapy for secondary prevention (for example, in survivors of cardiac arrest) or for primary prevention in high-risk patients (26). However, additional research is required before we can more accurately estimate the risk for sudden cardiac death in particular patients.

A collaborative individual-patient meta-analysis of the three available trials of secondary prevention established that ICDs were most beneficial in patients who were resuscitated after cardiac arrest and had ejection fractions less than or equal to 0.35 (10). A preliminary subanalysis of the data from the Multicenter Automatic Defibrillator Implantation Trial (MADIT) II found that ICDs had the greatest benefit in patients with a QRS duration greater than 0.12 seconds (27). The Amiodarone versus Implantable Defibrillator in Patients with Nonischemic Cardiomyopathy and Asymptomatic Nonsustained Ventricular Tachycardia (AMIOVIRT) Trial, a trial of primary prevention, involved 103 patients with nonischemic cardiomyopathy, ejection fraction less than 0.35, and nonsustained ventricular tachycardia. Two sudden cardiac deaths occurred in the amiodarone group, and one occurred in the ICD group. However, preliminary results from this trial did not reveal any significant difference between therapy with ICD or with amiodarone (28). These results are similar to the negative results of the Cardiomyopathy Trial (12) and confirm our findings that the benefits of ICDs on total mortality are sensitive to baseline risk for sudden cardiac death. The Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT), in which 2500 patients with ischemic or nonischemic cardiomyopathy and an ejection fraction less than 0.35 have been randomly assigned to amiodarone or ICD and are being followed for 2.5 years, will further refine the evidence base for primary prevention. Results are expected in 2004.

As a first step, a meta-analysis using individual-patient data and incorporating the results from the 8 trials examined in our study and 6 ongoing trials with mortality end points (Appendix Table) would be invaluable in refining the definition of risk and the potential impact of ICDs in particular patient subgroups. Such a meta-analysis would also permit exploration of potential interactions between ICDs and cardiac medications, such as β-blockers.

Other issues that remain to be addressed by ongoing studies include the cost-effectiveness of ICD therapy and its impact on quality of life. The latter may be important in determining whether an ICD should be used because some patients may have poorer quality of life after defibrillator implantation, particularly if they receive frequent shocks. Much work is needed to clearly define which patients are at risk for worsened quality of life (29 - 31). A recent analysis suggested that the incremental cost-effectiveness ratio for ICD placement compared with drug therapy was $66 677 per year of life saved in patients resuscitated after cardiac arrest (32 - 33). As expected, given the variability in baseline risks among patient subgroups, incremental cost-effectiveness ratios are estimated to range from $17 700 to $138 800 per year of life saved (34 - 39). The wide variability in cost-effectiveness ratios again emphasizes the importance of accurate risk stratification tools to assist in determining who should receive ICD therapy.

Our conclusions are limited by the paucity of trials in this area, particularly for some groups of patients (such as those with nonischemic cardiomyopathy). However, we used the rigorous methods of the Cochrane Collaboration to maximize the validity of our results. Given the nature of the trial reports we examined, we were unable to examine the relationship between specific baseline characteristics and the efficacy of ICDs; such analysis can be performed only in a meta-analysis using individual-patient data or a very large trial. Finally, in any systematic review, there is always concern about the potential for publication bias. However, we conducted an exhaustive search, including contact with primary study authors, device manufacturers, and experts in the field, to ensure that we had identified all relevant randomized trials, including trials with negative results.

Implantable cardioverter defibrillators are clearly more beneficial than drug therapy for secondary prevention of sudden cardiac death and for primary prevention in certain high-risk groups. However, further research is needed to develop accurate risk stratification tools, to determine the economic impact of ICD therapy in different subgroups of patients, and to evaluate quality-of-life issues.