The paper by Huikuri et al. (1) in this issue of the Journal takes on special relevance in view of the recent demonstration that the implanted cardioverter defibrillator (ICD) improves survival in high-risk patients with chronic coronary heart disease (2- 3). The question, simply put, is whether data from clinical and noninvasive cardiac tests collected during an acute myocardial infarction (AMI) can identify a subset of patients who are specifically at high risk for sudden cardiac death (SCD) in the post-hospital period during long-term follow-up. It is already well-known that the mortality risk in post-infarction patients is inversely related to the left ventricular ejection fraction (4) and that SCD makes up roughly 50% of cardiac mortality (5), although the percentage varies widely in reported studies (6). Thus, are there risk stratifiers that can accurately identify patients at high risk for SCD?

Huikuri et al. (1) have attempted to answer this question by studying a consecutive series of 675 post-infarction patients who were enrolled during a 2.5-year period beginning in 1996 and followed after hospital discharge for an extended period of time. Several clinical and state-of-the-art noninvasive cardiac tests were obtained on these patients before hospital discharge, and the findings from these tests were related to sudden and non-SCD during follow-up that averaged almost four years. The patients were optimally treated with beta-blockers, with 95% of the patients receiving beta-blockers two years after hospital discharge. Classification of sudden and non-SCD utilized standard clinical criteria, with two end point committees categorizing the mode of death. The total mortality during a mean follow-up of 43 months was 15%, with 8.7% (54 patients) classified as cardiac death—37 categorized as non-SCD and 17 as SCD. The relatively small number of cardiac deaths during the extended follow-up may explain, in part, the limited power of the study to identify patients at risk for SCD in this study. Only ejection fraction <0.40, nonsustained ventricular tachycardia on Holter recording, and a positive signal-averaged electrocardiogram entered a multivariate Cox model with hazard ratios significantly >1.0 for differentiating SCD from survival. However, the positive predictive accuracy of each of these three risk factors was <15%, and these findings indicate the clinical uselessness of these parameters for predicting patients who are likely to die of SCD.

Several studies, including the one by Huikuri et al. (1), have evaluated various clinical and non-invasive electrophysiologic markers for cardiac death, but none of these markers discriminated between sudden and non-SCD. Heart rate variability, baroreceptor sensitivity, QT-dispersion, QRS-interval duration, and the signal-averaged electrocardiogram all fall into this category. Huikuri et al. (1) suggest that their inability to predict SCD is due to the excellent adherence of their study patients to beta-blocker therapy, because beta-blockers have been shown to prevent SCD. They also note an unexpectedly low frequency of SCD in the first year after hospital discharge following the index myocardial infarction. But other reasons for limited findings may be operative. Currently, patients are being hospitalized with smaller-sized myocardial infarctions as a result of more sensitive enzyme markers for myocardial necrosis. The cardiac damage caused by AMIs is being attenuated by early thrombolysis and aggressive acute coronary revascularization procedures. The net result is that patients are being discharged from the hospital with better left ventricular function in the “beta-blocking era” than in the past, with many factors other than beta-blockers that could account for the improved outcome. In brief, the landscape of acute coronary disease has changed dramatically in the past decade with resultant improvement in the clinical course of patients after discharge from the hospital. The post-infarction mortality rate has declined during the past decade, but has this decline resulted in a shift from sudden death to non-sudden death? That is a question that cannot be answered by any of the recent studies.

Mortality is an end point that is unequivocal, for dead is dead (7). Categorization of the mechanism of cardiac death, especially when death occurs out-of-hospital, is imprecise for it is based on retrospective clinical data involving the circumstances and the location of death, pre-morbid clinical symptomatology that is often difficult to ascertain after the fatal event, and the time course of the terminal event. The error rate associated with the subclassification of cardiac death into sudden and non-SCD may be quite large, whether autopsy findings are available or not. Analyses by the Multicenter Automatic Defibrillator Implantation Trial (MADIT-II) End-point Review Committee using the Hinkle-Thaler mortality classification system (8) are shedding some light on this topic (H. Greenberg, personal communication, 2003). Sudden cardiac death accounted for 62% of the cardiac deaths in the conventionally treated MADIT-II patients, and 35% in the ICD group. The reduction in mortality with the ICD was almost exclusively the result of a decrease in cardiac death categorized as SCD. Using various assumptions about the efficacy of the ICD in preventing sudden death in MADIT-II, it appears that the imprecision, that is, the misclassification rate, in categorizing death as sudden due to suspected ventricular tachycardia/fibrillation may be an overclassification in the range of 20% to 30%.

Is our failure to identify patients at risk for SCD due to the lower post-infarction cardiac mortality in the current therapeutic era, the non-specificity of the presently available non-invasive electrophysiologic tests, and the potential error rate in accurate classification of the end point SCD? In part, yes, but I don't think these issues are the complete answer. It was previously thought that induction of ventricular tachycardia or fibrillation during invasive electrophysiologic (EP) testing would identify patients at subsequent risk for SCD. A recent prospective study showed that coronary patients who were non-inducible at EP testing had essentially the same risk for sudden death as those who were inducible (9). Preliminary findings from MADIT-II indicate that ICD-treated patients who were non-inducible at EP testing at the time of ICD implantation had more appropriate ICD shocks for ventricular fibrillation during their subsequent clinical course than patients who were inducible at EP testing. Thus, our current array of non-invasive and invasive tests performed during the hospital phase of AMI are inadequate to specifically identify who is at increased risk of dying from sudden rather than non-sudden death during follow-up time after an index coronary event.

What determines the development of malignant ventricular arrhythmias in patients with coronary heart disease, and can we identify those at increased risk for fatal arrhythmias? The evidence is accumulating that the occurrence of an abrupt ventricular arrhythmia is a multifactorial, time-dependent process involving a changing complex interplay of myocardial scar, ischemia, adrenergic factors, electrical heterogeneity, time, and possibly genetic factors, all superimposed on a vulnerable myocardial substrate that is “acquired” as a result of occlusive and progressive coronary artery disease. This multifactorial process has considerable overlap with factors that contribute to progressive heart failure. We are most likely dealing with a probabilistic event in which each of the currently measured risk factors identifies only a small fraction of the multifactorial risk process. Clearly, more specific physiologically based tests, better mathematical modeling involving the time-dependent process, and greater precision in determining when sudden death is really sudden are needed to identify patients at increased risk for sudden death. At present, it is probably best to use an ejection fraction <0.30 or 0.35 to predict coronary patients at risk for both sudden and non-SCD and simply assume that SCD accounts for approximately 50% of all cardiac deaths.