Author + information
- Received July 8, 2003
- Revision received November 7, 2003
- Accepted November 13, 2003
- Published online April 21, 2004.
- Henry Greenberg, MD*,‡,* (, )
- Robert B. Case, MD*,‡,
- Arthur J. Moss, MD†,
- Mary W. Brown, MS†,
- Elizabeth R. Carroll, MA†,
- Mark L. Andrews, BBS†,
- MADIT-II Investigators
- ↵*Reprint requests and correspondence:
Dr. Henry Greenberg, Division of Cardiology, St. Luke's Roosevelt Hospital Center, 1000 10th Avenue, Room 3B-30, New York, New York 10019, USA.
Objectives The purpose of this study was to determine the efficacy of implantable cardiac defibrillator (ICD) therapy in preventing sudden cardiac death (SCD) in post-infarction patients with advanced left ventricular (LV) dysfunction.
Background The Multicenter Automatic Defibrillator Implantation Trial (MADIT-II) randomized 1,232 post-infarction patients with an ejection fraction of ≤30% to ICD or conventional therapy. In the ICD group, there was a 31% decrease in the risk of total mortality. However, a better understanding of the mode of death is desirable in order to refine therapeutic interventions in high-risk populations.
Methods We evaluated the 202 deaths, using a variation of the Hinkle-Thaler classification system as well as a clinical classification system to determine the incidence of SCD and the incidence of cardiac death due to progressive LV failure.
Results The SCD rates were 10.0% in the conventional group and 3.8% in the ICD group (p < 0.01). The hazard ratio for the risk of SCD in the ICD group compared with the conventional therapy group was 0.33 (95% confidence interval 0.20 to 0.53, p < 0.0001). The ICD had no meaningful effect on non-sudden death (p = 0.32). The effect of defibrillator therapy in reducing SCD was similar in subgroup analyses stratified according to relevant baseline characteristics.
Conclusions The decrease in mortality with ICD therapy in MADIT-II is entirely due to a reduction in SCD, with similar reductions in SCD in a spectrum of subgroups stratified according to relevant baseline characteristics.
Previous randomized studies (1,2)have shown that the survival rate in high-risk patients with coronary artery disease, ventricular dysfunction, and inducible ventricular tachycardia can be increased by an implantable cardiac defibrillator (ICD). The Multicenter Automatic Defibrillator Implantation Trial II (MADIT-II) (3)assessed the value of prophylactic ICD therapy in patients with a history of myocardial infarction (MI) and an ejection fraction (EF) of ≤0.30 without the requirement for spontaneous or inducible ventricular arrhythmias. The primary end point was total mortality. The study showed a 31% reduction in the risk of mortality in the ICD group compared with conventional medical therapy.
In most cases of sudden cardiac death (SCD), the presence of a terminal malignant arrhythmia is not known, because electrocardiographic (ECG) monitoring is rarely available. The MADIT-II End Point Review Committee, established before the start-up of the trial, reviewed all mortality events utilizing a variation of the traditional Hinkle-Thaler clinical classification system (2,4,5)in order to determine the mode of death in the two treatment arms of the trial. This report provides detailed analyses of the terminal events in MADIT-II, with additional insight into the efficacy of ICD therapy in patients with advanced left ventricular (LV) dysfunction.
Details of the methods used in MADIT-II have been presented previously (3). MADIT-II randomized 1,232 patients with previous MI and a left ventricular ejection fraction of ≤0.30 to either ICD or conventional medical therapy in a 3:2 ratio. An End Point Review Committee (Drs. Greenberg and Case), whose members were independent of all aspects of the trial, reviewed the accumulated data on all deaths in MADIT-II, including relevant portions of the ambulance, emergency room, and medical records of patients who died, with a concentrated focus on the clinical details of the terminal event. The narratives supplied by clinical coordinators at the enrolling sites included a description of the circumstances surrounding each patient's death, statements of those present at the terminal event, prodromal symptoms, medications, location of death, and time to death after the onset of symptoms. The coordinators sought specific information on acute coronary symptoms, especially chest pain, in the 24-h interval preceding death. The committee was not blinded as to treatment allocation. The committee did not review any ICD printouts recorded at the time of death, of which there were very few, so as to maintain a uniform clinical assessment of the time to death.
A modified Hinkle-Thaler system was used to classify deaths. The original Hinkle-Thaler classification (4)of suspected arrhythmic cardiac death was based almost exclusively on a sudden, unexpected death with or without immediately preceding symptoms, with 95% of those classified as a presumed arrhythmic event dead within 1 h of cardiac symptoms. The ECG evidence of a terminal arrhythmia was not required. In the current analysis, we have further defined this category, which is classified here as SCD, and includes those who: 1) died suddenly and unexpectedly within 1 h of cardiac symptoms in the absence of progressive cardiac deterioration (this category includes several patients whose time course of death was prolonged by unsuccessful resuscitative measures, as well as patients who died within 1 h of the onset of cardiac symptoms in the setting of stable heart failure [HF]); 2) died unexpectedly in bed during sleep; and 3) died unexpectedly within 24 h after last being seen alive. Three patients were found more than 24 h after having been last seen alive and were classified as unknown cause of death.
Patients categorized as SCD were subclassified into those with and without symptoms of severe LV dysfunction (New York Heart Association [NYHA] functional class ≥III HF). Other Hinkle-Thaler categories of death included non-SCD, unclassified cardiac death, non-cardiac death, and unknown/unclassified cause of death when insufficient information was available to make a reasonable decision as to the cause of death. Non-SCD included patients who died of progressive cardiac failure or patients who did not meet the time criteria for sudden death. Progressive cardiac failure was defined as unstable, clinical progression of deteriorating pump function in the setting of active therapy, most often in an intensive care setting. However, patients with advanced HF in whom death was not anticipated as imminent were categorized as sudden death if their terminal event met the time criteria.
In addition, an alternative clinical classification of death developed before the analysis was also utilized. In this clinical classification scheme, SCD (death with 1 h of symptom onset) was subcategorized as primary when there were no preceding symptoms, or as secondary when a complaint of chest pain occurred during the 1-h period of time before death. Marked ECG changes indicative of active myocardial ischemia were not present in any of the reviewed records. A new multiple cause category was created, and it required the presence of several medical problems in which coronary heart disease contributed to, but was not the dominant feature of, the mortality event. A category of death related to cardiac procedures was also included.
Analyses used version 3.0 of the MADIT-II database, released on July 26, 2002. For categorical variables, the chi-square statistic was used to assess group differences. For continuous variables, comparisons were performed using the ttest for independent samples. Cumulative survival curves were determined by the Kaplan-Meier method (6), with a statistical comparison of cumulative mortality by the log-rank method. The Cox proportional hazards regression model (7)was used to calculate the risk for SCD and non-SCD in the total population and in subgroups stratified by relevant baseline characteristics for patients randomized to ICD versus conventional therapy.
Clinical characteristics by outcome in the two treatment groups
The baseline characteristics of those who survived and those who died in the two treatment groups are presented in Table 1A. In both groups, those who died were older and had more advanced HF, a lower EF, and poorer renal function at enrollment than those who survived.
The cardiac morbidity that occurred after enrollment in the two treatment arms among those who survived and those who died are presented in Table 1B. In both treatment arms, those who died had a higher frequency of hospitalization for congestive HF and interim MI after enrollment than those who survived.
Mode of death
The mode of death by treatment group classified by the modified Hinkle-Thaler scheme is presented in Table 2A. Sudden cardiac death made up 51% (49 of 97 patients) of the total deaths in the conventional therapy group, compared with 27% (28 of 105) of the total deaths in the ICD group. Sudden death constituted 61% (49 of 80 patients) of the cardiac deaths in the conventional therapy group, compared with 35% (28 of 79) in the ICD group (chi-square =12.4, p < 0.001). The distribution of sudden, non-sudden, and unclassified deaths as a percentage of the cardiac deaths in the two treatment groups is shown in Figure 1. Similar findings were obtained when using the clinical classification system (Table 2B). Chest pain symptoms suggestive of active myocardial ischemia infrequently preceded sudden death in either treatment group. The two cardiac procedure deaths were both related to coronary artery bypass graft surgery; there were no deaths related to ICD implantation.
The nominal (raw) death rates, calculated from the number of specified deaths per number of randomized patients in each treatment arm, are given in Table 2. The SCD rate was 3.8 % in the ICD group and 10.0 % in the conventional therapy group (p < 0.01). There was no significant difference in non-sudden death rates in the two treatment arms.
Kaplan-Meier estimates of the effect of ICD on cumulative SCD are presented in Figure 2A. There is an early and progressive separation of the probability of sudden death between the two arms of the trial. The hazard ratio (HR) for SCD was 0.33 (95% confidence interval 0.20 to 0.53; p < 0.0001) indicating a 67% reduction in the risk of SCD at any interval of time among patients in the defibrillator group compared with those in the conventional therapy group.
The Kaplan-Meier estimates for non-sudden death are similarly presented in Figure 2B. There is a slightly higher non-SCD rate in the ICD group than in the conventionally treated group, especially during the first year, but this separation in the two curves is not statistically significant (p = 0.32).
The ICD to conventional therapy HRs for SCD, utilizing selected baseline characteristics (Table 1), were similar for all relevant subgroups, with values in the range of 0.23 to 0.44; there were no statistically significant interactions. That is, all patient subgroups had an equivalent reduction in SCD with ICD therapy. Cox analyses for non-SCD revealed HRs in the range of ∼1.0 for all relevant subgroups, a finding indicating the absence of an ICD effect on non-SCD.
Chronology of terminal event
The times from the onset of the terminal event to cardiac death by treatment group are shown in Table 3. This time course was recorded in 143 patients. A smaller percentage of patients died within 1 h (p < 0.01) and within 24 h of symptom onset in the defibrillator group compared with the conventional therapy group. This finding was also evident in the death rates when combining the 1-h interval with the dead-in-bed category, or when combining the 24-h interval with all unwitnessed deaths (both p < 0.01). Nearly twice as many delayed cardiac deaths (>24 h) occurred in the ICD group.
Location of cardiac death
Patient locations at the time of cardiac death are shown in Table 4. The out-of-hospital/emergency department cardiac death rates were higher in the conventional therapy group than in the ICD group (9.6% vs. 3.8%, p < 0.01). The ICD group had a slightly higher rate of death occurring in an in-hospital/extended care facility than did the conventional group.
Adverse cardiac events in the week before cardiac death
The respective frequencies of the following adverse events in the week before SCD in the conventional (n = 49) versus ICD (n = 28) patients were 4% vs. 4% for syncope, 4% vs. 4% for angina pectoris, 10% vs. 4% for MI, 10% vs. 25% for ventricular arrhythmia, and 16% vs. 43% for congestive HF. The adverse events other than congestive HF were infrequent in both treatment groups, and in no instance were the differences meaningful.
The major finding from this analysis of the mortality events in MADIT-II is that the improvement in survival with ICD therapy is entirely due to a reduction in SCD. The SCD rate was reduced from 10.0% in the conventional therapy group to 3.8% in the ICD group, and this absolute reduction of 6.2% is higher than the total mortality reduction of 5.6% (19.8% to 14.2%) reported in the primary publication (3). The HR is 0.33 for SCD, and this value is meaningfully lower than the HR of 0.69 for total mortality.
The Kaplan-Meier cumulative SCD curves for the two treatment groups show an early and progressive separation (Fig. 2A), in contrast to the delayed divergence when total mortality was used as the end point (3). The early separation in the SCD curves is consistent with effective termination of malignant ventricular arrhythmias beginning early after ICD implantation. The lack of early separation in the total mortality curves (Fig. 2in reference 3) may be explained by the slightly higher non-SCD rate during the first year after enrollment in the ICD group compared with the conventionally treated group (Fig. 2B). We cannot determine whether the ICD contributes to a slight increase in non-SCD during the first year, or whether the conventionally treated group has an unusually low early non-sudden death rate, possibly a reflection of intrinsic variability in naturally occurring non-sudden cardiac events. It should be emphasized that the difference in non-SCD rates between the two treatment arms is not statistically significant (p = 0.32).
There were no significant differences in the reduction of SCD by ICD in subgroup analyses stratified according to age, gender, NYHA functional class, EF, QRS interval, interval since the most recent MI, and beta-blocker use. We were not able to identify any patient subgroup that achieved significantly increased or decreased ICD efficacy for the prevention of SCD.
It is important to consider the meaning of sudden death in this report. Abrupt, unexpected death of an ambulatory patient is usually assumed to be due to a malignant cardiac arrhythmia, except for a small percentage of patients who experience a catastrophic event such as dissecting aortic aneurysm, pulmonary embolus, cerebral hemorrhage, myocardial rupture, or other sudden non-arrhythmic death. However, precise evidence of an arrhythmic cause of sudden death remains elusive, as ECG recordings at the time of a terminal event are infrequent. However, the current findings that the ICD significantly reduces the rate of SCD provide strong support that a ventricular tachycardia/fibrillation mechanism is responsible for the vast majority of SCDs as presently categorized.
The percentage of patients categorized as dying of SCD in the conventionally treated group was 61%, a value very similar to the 63% found in the recent national report from the Center for Disease Control (8). Although SCD was significantly reduced in MADIT-II by ICD, 35% of the cardiac deaths that occurred in the ICD group were classified as SCD (Table 3). This important residual SCD rate may reflect the absence of an ICD-reversible ventricular arrhythmia, a depressed or ischemic myocardial state that cannot withstand the physiologic challenge of a malignant arrhythmia, a non-arrhythmic catastrophic event, ICD malfunction, or a combination of more than one of these factors. The recent report by Mitchell et al. (9)provides useful information in this regard. The mode of death was investigated in over 300 ICD-treated patients by post-mortem ICD interrogation. The mean EF for the group was 0.27. Of 68 cardiac deaths categorized as sudden, 29% had post-shock electro-mechanical dissociation, 25% had ventricular tachycardia/fibrillation uncorrected by shock, 16% had primary electro-mechanical dissociation, 13% had incessant ventricular tachyarrhythmias, and 16% had other causes of their terminal event.
We and others have used the Hinkle-Thaler approach for classifying SCDs subcategorized by the degree of LV dysfunction (1,2,5). In our study, a reduction in SCD provided important insight into the mechanism of ICD efficacy. However, there are problems with the application of the Hinkle-Thaler schema to current ICD trials. Foremost among these is the inability of the Hinkle-Thaler classification to categorize potential mechanisms of death that are amenable to reversion by an ICD. The Hinkle-Thaler schema directs attention exclusively to the terminal event. Such a focus does not provide information on those who may and may not have benefited from ICD therapy. For example, signs and symptoms of HF that often precede death by weeks to months may provide important information on the factors contributing to the terminal event and whether the patient is likely to benefit from appropriate ICD therapy, but this information is not utilized in the Hinkle-Thaler approach. We attempted to improve on the Hinkle-Thaler approach with our clinical classification schema, and there was some improvement (Table 2), but minimally so.
This analysis reinforces the assumption that the ICD reduces arrhythmic death in high-risk coronary patients, despite a lack of certainty in individual cases. It is also clear that a clinical algorithm, such as the Hinkle-Thaler schema, which looks at the terminal event of such patients, permits a better insight into the mechanisms of death than does total mortality. It is also clear that neither the Hinkle-Thaler nor the clinical algorithms used in this study attain a precision that reflects the potential benefit of ICD, in that neither schema is able to identify subpopulations of patients who will not benefit from ICD therapy. In the future, the clinical classification of cardiac death in ICD trials should be integrated with information obtained about the terminal rhythm from post-mortem ICD interrogation in order to better identify the mechanism of the terminal cardiac event and the true potential of ICD to prevent SCD (9), again helping to identify patients who may not benefit from ICD therapy. The burden of positive expectation, the monetary expenditure associated with ICD therapy, and the recent functional improvement seen with biventricular pacing (10)require greater clarity in our understanding of the mechanisms of cardiac death, both arrhythmic death and progressive HF death in at-risk populations.
☆ This study was supported by a research grant from Guidant Corporation, St. Paul, Minnesota, to the University of Rochester School of Medicine and Dentistry.
- ejection fraction
- heart failure
- hazard ratio
- implantable cardiac defibrillator
- left ventricular
- Multicenter Automatic Defibrillator Implantation Trial
- myocardial infarction
- New York Heart Association
- sudden cardiac death
- Received July 8, 2003.
- Revision received November 7, 2003.
- Accepted November 13, 2003.
- American College of Cardiology Foundation
- Hinkle L.E.,
- Thaler H.T.
- Bigger J.T.,
- Whang W.,
- Rottman J.N.,
- et al.
- Cox D.R.
- ↵State-specific mortality from sudden cardiac death—United States, 1999. Center for Disease Control, MMWR, February 15, 2002. Available at: http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5106a.htm,2139. Updated June 15, 2002, accessed October 2002.
- Mitchell L.B.,
- Pineda E.A.,
- Titus J.L.,
- Bartosch P.M.,
- Benditt D.G.