Author + information
- Received April 29, 2004
- Revision received August 9, 2004
- Accepted August 16, 2004
- Published online December 7, 2004.
- Kumaraswamy Nanthakumar, MD*,* (, )
- Andrew E. Epstein, MD‡,
- G. Neal Kay, MD‡,
- Vance J. Plumb, MD‡ and
- Douglas S. Lee, MD†
- ↵*Reprint requests and correspondence:
Dr. Kumaraswamy Nanthakumar, Division of Cardiology, University Health Network, Toronto General Hospital, 150 Gerrard Street West, PMCC 3-558, Toronto, Ontario, Canada M5G 2C4
Strategies to decrease sudden cardiac death in patients with left ventricular systolic dysfunction are evolving. Recent clinical trials have evaluated the role of prophylactic implantable cardioverter-defibrillators (ICDs) in patients with and without additional risk stratifiers. We pooled studies comparing treatment with and without ICDs from published data and presented abstracts, irrespective of QRS duration and etiology of systolic dysfunction. On the basis of the available clinical trials, implantation of an ICD for primary prevention of death provides a 7.9% absolute mortality reduction (p = 0.003) in patients with left ventricular (LV) systolic dysfunction who were receiving optimized medical therapy. This finding was not sensitive to the exclusion of any individual trial. The ICD is an effective primary preventative measure in patients who are at risk for death; however, the application of this therapy needs to be individualized for the patient, similar to drug therapies in LV systolic dysfunction. In health care settings without unlimited resources, optimal use of this therapy will require better risk stratification methods or lowering of the initial device cost.
Sudden cardiac death accounts for approximately 50% of the mortality for patients with left ventricular (LV) systolic dysfunction (decreased LV ejection fraction) (1). Because the implantable cardioverter-defibrillator (ICD) is effective in terminating ventricular tachycardia and fibrillation, it makes intuitive sense that a strategy of prophylactic ICD implantation in patients with LV systolic dysfunction would decrease arrhythmic mortality (2). Whether this therapy would decrease total mortality is less obvious (3).
Although the number of published prospective clinical trials of ICD implantation has been increasing, policymakers dealing with the financial burden of adopting such a strategy have restricted usage to retrospectively identified subgroups that may have the greatest benefit (4). This decision was mainly driven by a preliminary review of the Multicenter Automatic Defibrillator Implantation Trial II (MADIT II), in which QRS duration >120 ms was thought to be an important risk stratifier. However, subsequent analysis has questioned the validity of excluding patients with QRS width <120 ms (5), as the risk of sudden cardiac death in those patients is not negligible and the mortality benefit of an ICD remains statistically significant. A review of the Multicenter Unsustained Tachycardia Trial (MUSTT) registry revealed that the all-cause mortality of patients who were similar to those in the MADIT II trial was 65% over a period of five years in patients with QRS duration >120 ms and 46% in patients with QRS duration <120 ms. In the MUSTT registry, the all-cause five-year mortality for patients with a QRS duration <120 ms who were treated with an ICD was 17%, compared with 38% for patients with QRS duration >120 ms who were treated with an ICD. These analyses point out the potential weakness of limiting ICD therapy to patients with QRS duration >120 ms. The pitfalls of post-hoc analysis that lead to restriction of therapy to “high-risk” subgroups have been well described (6) and may have particular relevance to ICD therapy. In addition, as a consequence of the current restrictions that limit reimbursement of prophylactic ICDs to patients with QRS width >120 ms, it is possible that some patients with a narrow QRS width may be subjected to unnecessary electrophysiology (EP) studies for the purposes of risk stratification and reimbursement (7).
The ability of individual trials to assess all-cause mortality benefit from prophylactic ICD therapy has been limited by a low baseline risk of death and inadequate power to detect mortality differences between treatment groups (8). A pooled analysis of all-cause mortality in available randomized studies of prophylactic ICD therapy in patients with LV systolic dysfunction would provide a more precise estimate of survival benefit for patients with both ischemic and nonischemic causes of LV systolic dysfunction, without relying on subgroup analysis.
Several trials and meta-analyses have addressed the use of ICDs for preventing death in patients with LV systolic dysfunction who have never had an arrhythmic event. It has been estimated from four randomized trials and one nonrandomized trial that prophylactic ICD implantation reduces the relative risk (RR) for death by approximately 28% to 34% (2,3). When only the four randomized trials were included, the sensitivity analysis indicated that the benefit was not robust; that is, the conclusion was sensitive to exclusion of any one trial from the meta-analysis (3). Since the publication of these analyses, data from five additional trials have become available (9–13).
While a meta-analysis on the effect of prophylactic ICDs on survival of patients with LV systolic dysfunction would be the ideal way of evaluating the data, all of the trials would first have to be published. Thus, there will be a hiatus during which physicians caring for potential ICD recipients will have to address the issue of sudden cardiac death based on the available data, without the benefit of meta-analyses. The focus of this pooled analysis is to provide an updated review of all-cause mortality from the recently published and presented clinical trials. Our aim is to provide an updated analysis of available data from randomized controlled trials and to present a perspective on the effect of ICD implantation in prevention of death in patients with LV systolic dysfunction, irrespective of etiology of the LV dysfunction and the QRS duration.
The five trials that had been included in previous meta-analyses were the MADIT (14), the Coronary Artery Bypass Graft Patch (CABG-Patch) trial (15), the MUSTT (16), the Cardiomyopathy Trial (CAT) (17), and the MADIT II (18). The trial design, etiology of LV dysfunction, ejection fraction, additional risk marker used for inclusion, and follow-up duration are detailed for each of the trials in Table 1.
The MADIT trial tested the hypothesis that an ICD would improve survival for patients with ischemic heart disease at high risk of arrhythmic death (14). This trial used epicardial leads in one-half of the patients and transvenous endocardial leads in the other half. In this trial the mortality rate in the ICD group was 16% versus 39% in patients randomized to the conventional treatment group after 27 months of follow-up (RR reduction of 59%, absolute risk reduction 23%, p = 0.009) (14).
The CABG-Patch trial tested the role of prophylactic ICD implantation in patients undergoing elective coronary artery bypass surgery (15). In contrast to other trials and similar to one-half the patients in the MADIT trial, the defibrillation leads were epicardial patches placed at the time of coronary artery bypass surgery. The mortality rate in the ICD-treated patients was 23% and 21% in the control patients, with no difference in all-cause mortality at 32 months (p = 0.64) (14).
The MUSTT trial was a randomized trial of two treatment strategies (EP-guided versus empiric) for patients with coronary artery disease and unsustained ventricular tachycardia, and ICD use was not randomized (16). The five-year mortality rate in the EP-guided group was 42%, which compared with 48% in the non–EP-guided group (p = 0.06). However, a nonrandomized comparison of patients in the EP-guided arm demonstrated that the patients who received an ICD had significantly lower all-cause mortality at five years (24%) than those who did not receive a device (55%) (RR reduction of 56%, absolute risk reduction 31%, p < 0.001) (16).
The CAT study included patients with dilated, nonischemic cardiomyopathy. Enrollment in this very small trial was terminated early because the interim analysis showed that the overall one-year mortality rate for all patients was only 5.6% and well below the assumed value of 30%. According to the protocol, the randomization was stopped, and all randomly assigned patients completed the scheduled follow-up period of two years. After two years the mortality rate was 26% in the ICD patients and 50% in the control patients (p = 0.554) (17).
The MADIT II trial tested the hypothesis that prophylactic ICD implantation would reduce mortality in patients with ischemic cardiomyopathy without the requirement of further risk stratification (18). The mortality rate was 19.8% in the conventional therapy group and 14.2% in the ICD group at 20 months (RR reduction of 28%, absolute risk reduction of 5.6%, p = 0.016).
These five trials were included in two previous meta-analyses that showed a relative reduction in all-cause mortality of 34% (p = 0.03) (2,3). Both analyses included the MUSTT trial, which was not a randomized comparison of ICD use (19,20). Sensitivity analysis of the meta-analysis performed by Lee et al. (3) indicated that the primary prevention analysis was sensitive to the MUSTT trial, which was not a randomized comparison of the ICD versus conventional therapy. Excluding the MUSTT trial, the RR of arrhythmic death was 0.45 (p < 0.01), but the risk of all-cause mortality (RR 0.74; 95% confidence interval [CI] 0.51 to 1.08) was no longer significantly reduced (p = 0.12) (3).
Since the publication of the previous meta-analyses, data from five additional primary prevention trials have been reported: 1) the Amiodarone Versus Implantable Cardioverter-Defibrillator Randomized Trial (AMIOVIRT) in patients with nonischemic cardiomyopathy (9); 2) Comparison of Medical Therapy, Pacing, and Defibrillation in Patients With Left Ventricular Systolic Dysfunction Trial (COMPANION) (10); 3) Defibrillators In Non-Ischemic Cardiomyopathy Treatment Evaluation Trial (DEFINITE) (11); 4) Defibrillator in Acute Myocardial Infarction Trial (DINAMIT) (12); and 5) Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT) (13). The key particulars of these trials are summarized in Table 1.
The AMIOVIRT trial evaluated patients with nonischemic dilated cardiomyopathy. The trial was terminated after a pre-specified stopping rule for futility was reached. Survival at 3 years was 88% in the ICD group and 87% in the control group, significantly higher than had been predicted, with no difference in mortality between groups (p = 0.8) (9).
The COMPANION trial was a prospective, randomized trial that compared optimized medical therapy, optimized medical therapy with cardiac resynchronization therapy using biventricular pacing (CRT), and optimized medical therapy with CRT and ICD capability (CRT-D) in patients with either ischemic or nonischemic cardiomyopathy. Hence, any difference in all-cause mortality between the CRT and the CRT-D groups provides insight into the utility of prophylactic ICD therapy in this patient population. At 12 months, the mortality rate was 19% in the optimized medical therapy group, 15% in the CRT group, and 12% in the CRT-D group. The relative reduction in all-cause mortality compared with optimized medical therapy was 24% with CRT (p = 0.06) and 36% with CRT-D (p = 0.004) (10).
The DEFINITE trial included patients with nonischemic cardiomyopathy. The mortality rate in patients randomized to optimized medical therapy was 13.8% compared with 8.1% in patients randomized to the ICD, over mean follow-up of 29 months (RR reduction of 41%, absolute risk reduction 5.7%, p = 0.06) (11).
The DINAMIT trial enrolled patients within 40 days of myocardial infarction to a treatment strategy of optimized medical therapy either with or without an ICD. These patients had no difference in all-cause mortality (7.5% in the ICD group vs. 6.9% in the non-ICD group) at a mean follow-up of 2.5 years (p = 0.66) (12).
The SCD-HeFT trial randomized 2,521 patients with ischemic or nonischemic cardiomyopathy to optimized heart failure therapy including beta-blockers, angiotensin-converting enzyme (ACE) inhibitors, and spironolactone; optimized heart failure therapy with the addition of amiodarone; or medical treatment with the addition of an ICD. After a median follow-up of four years, the mortality rate was 22% in the ICD group, 28% in the amiodarone group, and 29% in the control group. This resulted in a 24% RR reduction and a 7% absolute risk reduction in the all-cause mortality in the ICD group as compared with optimized medical therapy alone (p = 0.007) (13).
The number of patients who died in the ICD and non-ICD groups was extracted from published data and presented abstracts. The RR and RR difference with 95% CIs using fixed and random effects models were calculated for the trials separately and for all the trials in the overall analysis (20,21). Statistical heterogeneity between studies was identified using the chi-square statistic. When pooled analysis resulted in significant heterogeneity, the random effects model is presented. Conversely, in the absence of heterogeneity between studies, the fixed effects analysis is reported. Sensitivity analysis was performed to determine whether the overall conclusion of ICD effectiveness was robust (i.e., not overly influenced by an individual randomized trial). We re-analyzed the pooled data after sequentially removing each trial from the pooled analysis.
Figure 1demonstrates the RR reduction from 10 different trials. Among the 3,530 patients randomized to receive an ICD, there were 647 deaths, compared with 976 deaths in the 3,723 patients randomized to a non-ICD group. When the deaths from all the trials were pooled using the random effects model, there was a 25% relative reduction in all-cause mortality with the ICD (p = 0.003) (95% CI 9% to 37%). All-cause mortality in the control group for the 10 studies was 26.4%, compared with 18.5% in the ICD group (absolute mortality reduction of 7.9%).
Table 2demonstrates the results of sensitivity analysis. The conclusions reached are not sensitive to the exclusion of any one trial as evidenced by the 95% CI of the odds ratios and p values. Thus, the conclusions reached by the pooling are robust.
Our analysis indicates that implantation of an ICD for primary prevention of death in patients with LV systolic dysfunction irrespective of etiology and QRS duration significantly reduces overall mortality at intermediate-term follow-up. This finding is not sensitive to the exclusion of any one of the trials from the pooled analysis.
It can be argued that some of the trials included in the pooled analysis are confounded by other variables that affect sudden cardiac death. Veenhuyzen et al. (22) quantified the effect of CABG surgery on mortality in the Studies Of Left Ventricular Dysfunction (SOLVD) trials. Prior CABG surgery was associated with a 25% reduction in risk of death and a 46% reduction in risk of sudden death independent of ejection fraction and severity of heart failure symptoms. When they applied these results to a group of patients with LV dysfunction who had not undergone prior surgery (Coronary Artery Surgery Study [CASS] registry), the predicted annual rates of death (8.2%) and sudden death (2.4%) were similar to those observed in the CABG-Patch trial (7.9% and 2.3%, respectively). Those investigators concluded that prior CABG surgery is associated with a significant independent reduction in mortality that could have confounded the effect of ICD therapy in the CABG-Patch trial (22). The DINAMIT trial evaluated patients immediately after myocardial infarction. The confounding effect of ICD shock therapy, ACE inhibition, beta-blocker therapy, and revascularization on the ejection fraction over time is difficult to quantify in this population. However, it must be stated that the fact that the CABG-Patch and DINAMIT trials included treatment effects that may have confounded benefits of an ICD are supposition and merely hypothesis generating. The MUSTT trial was not a randomized trial of ICD therapy, a fact that is known to invalidate pooled analyses when included. Hence, for these reasons we excluded these three trials (CABG-Patch, DINAMIT, and MUSTT) and repeated the pooled analysis. Figure 2demonstrates the seven remaining trials and the relative mortality risk reduction. When the deaths from these seven trials were pooled, there was a 26% reduction in all-cause mortality with the ICD (p < 0.00001) (95% CI 17% to 33%). All-cause mortality in the control group for the seven trials was 24.1%, compared with 17.5% in the ICD group (absolute mortality reduction of 6.6%). Although the test for heterogeneity (the assessment of difference in treatment effect) was significant before exclusion of these trials (p = 0.0005) (Fig. 1), after exclusion of these trials the test for heterogeneity was no longer significant (p = 0.39) (Fig. 2). This supports the removal of the CABG-Patch, DINAMIT, and MUSTT trials from a pooled analysis of prophylactic ICDs in patients with impaired LV systolic function. Thus, the strategy of ICD implantation for primary prevention of death is consistent and significant for the seven randomized trials listed in Figure 2.
ICD therapy and other accepted therapies in LV dysfunction
The absolute mortality reduction of 7.9% (number needed to treat = 13) by prophylactic ICDs should be compared with a 6.1% reduction afforded by ACE inhibitors (23) and a 4.4% by beta-blockers (24) in patients with LV systolic dysfunction at intermediate-term follow-up. Approximately 90% of the patients in this pooled analysis were from later trials (MADIT II, SCD-HeFT, COMPANION, DEFINITE) in which there was protocol-specified optimized medical therapy with beta-blockers and ACE inhibitors for LV dysfunction. Hence, the mortality reduction provided by ICDs was in addition to the reductions afforded by beta-blockers and ACE inhibitors. Like aldosterone antagonists, which reduce mortality by 2.3% when added to beta-blockers and ACE inhibitors (25), ICDs offer benefits that are additive to state-of-the art medical treatment.
The results of our study identify the effect of the ICD as primary preventative therapy on the risk of mortality with greater precision than a previously published meta-analysis (3). Owens et al. (26) described the wide range of incremental cost-effectiveness that could result in sensitivity analyses that vary the estimated RR reduction with the ICD. The narrower range of ICD effect in this study allows for greater precision in estimating the cost-effectiveness of this therapy. The cost of treating patients with systolic dysfunction with ICDs is nonlinear as there is a high initial device cost, hospitalization cost, and lower follow-up cost. Not only is ICD therapy nonlinear with regard to cost, as shown by Salukhe et al. (27), it is nonlinear with regard to life-years saved and number of patients needed to be treated to save one life. The benefit of the strategy of implanting a prophylactic ICD in patients with LV systolic dysfunction is expected to rise dramatically over time. Salukhe et al. (27) have shown that at least for the first three years, this benefit rises as the square of implant duration. This exponential growth of benefit suggests that our analysis represents a conservative conclusion of the benefit of a strategy of implanting ICDs for the prevention of death in patients with LV systolic dysfunction.
Individualization of ICD therapy
Although prophylactic ICD implantation improved overall mortality in the studies reviewed, these devices are unlikely to be appropriate for all patients who meet the entry criteria. Just as some patients with LV systolic dysfunction do not tolerate beta-blockers because of bradycardia or hypotension, or ACE inhibitors because of cough or hyperkalemia, there are patients who may not be ideal candidates for a prophylactic ICD. Indeed, the care of patients with LV systolic dysfunction must be individualized, after discussion with them. Some patients may elect not to undergo an operative procedure even when fully informed of the expected survival benefit. For example, patients with intractable congestive heart failure may decline ICD implantation because of concerns regarding prolongation of a very poor quality of life. Others may have such anxiety regarding defibrillation shocks (that the expectation of an increased probability of survival is outweighed by an adverse impact of the device on their quality of life). In addition, there are often comorbid conditions that have competing effects on mortality that must be considered by both the patient and the physician. Finally, patients entering clinical trials may be selected for the likeliness they will comply with follow-up and concomitant medical therapies. Whether unselected populations will have the same degree of benefit from prophylactic ICD implantation remains to be determined.
On the basis of currently available clinical data, implantation of an ICD for primary prevention of death in patients with LV systolic dysfunction provides a significant reduction in overall mortality at intermediate-term follow-up. The absolute mortality reduction is additive to accepted drug therapies for LV systolic dysfunction. However, just like other therapies for LV systolic dysfunction, the application of this treatment needs to be individualized for each patient. The ICD therapy differs substantially from drug therapy, particularly from the standpoint of initial cost. The optimal use of this therapy will require better risk stratification methods or lowering of initial device cost.
Dr. Nanthakumar is supported by the clinician scientist program of the Canadian Institute of Health Research. We would like to disclose that there is potential conflict of interest in that all authors of this manuscript except Dr. Lee’s care for patients with these devices. Dr. Epstein chairs events committees for St. Jude Medical and Guidant Corp., and has been a speaker for Biotronik, Guidant Corp., Medtronic Inc., and St Jude Medical. Drs. Kay, Epstein, and Plumb are investigators for Guidant Corp., St. Jude Medical, Medtronic Inc., and Biotronik. Dr. Nanthakumar is an investigator for Medtronic Inc. In addition, Dr. Kay is a consultant for Guidant Corp.
- Abbreviations and acronyms
- angiotensin-converting enzyme
- Amiodarone Versus Implantable Cardioverter-Defibrillator Randomized Trial
- Coronary Artery Bypass Graft Patch trial
- Cardiomyopathy Trial
- confidence interval
- Comparison of Medical Therapy, Pacing, and Defibrillation in Patients With Left Ventricular Systolic Dysfunction trial
- cardiac resynchronization therapy
- Defibrillators In Non-Ischemic Cardiomyopathy Treatment Evaluation trial
- Defibrillator In Acute Myocardial Infarction Trial
- implantable cardioverter-defibrillator
- left ventricular
- Multicenter Automatic Defibrillator Implantation Trial
- Multicenter Unsustained Tachycardia Trial
- relative risk
- Sudden Cardiac Death in Heart Failure Trial
- Received April 29, 2004.
- Revision received August 9, 2004.
- Accepted August 16, 2004.
- American College of Cardiology Foundation
- Uretsky B.F.,
- Sheahan R.G.
- Lee D.S.,
- Green L.D.,
- Liu P.P.,
- et al.
- Centers for Medicare and Medicaid Services
- Packer D.L.,
- Peterson M.A.,
- Haffley G.,
- Lee K.,
- Buxton A.E.
- ↵SCD-HeFT: Experts look for closure on ICD reimbursement debate after MADIT. http://www.theheart.org/viewEntityDispatcherAction.do?primaryKey=52611. Accessed April 22, 2004.
- Strickberger S.A.,
- Hummel J.D.,
- Bartlett T.G.,
- et al.
- ↵DINAMIT study: ICD immediately after MI of no benefit. http://www.theheart.org/viewEntityDispatcherAction.do?primaryKey=530153. Accessed April 22, 2004.
- ↵NHLBI study finds implanted defibrillator reduces heart failure deaths. http://www.nhlbi.nih.gov/new/press/04-03-08.htm. Accessed April 22, 2004.
- Bansch D.,
- Antz M.,
- Boczor S.,
- et al.
- Mulrow C.,
- Cook D.
- Veenhuyzen G.D.,
- Singh S.N.,
- McAreavey D.,
- Shelton B.J.,
- Exner D.V.
- Owens D.K.,
- Sanders G.D.,
- Harris R.A.,
- et al.
- Salukhe T.V.,
- Dimopoulos K.,
- Sutton R.,
- Coats A.J.,
- Piepoli M.,
- Francis D.P.