Applied Clinical Epidemiology

Epidemiologic information that is clinically applicable to the problem of sudden cardiac death extends beyond the collection and analysis of mortality data across various demographic subgroups. The problem is dynamic and so is its epidemiology. Three areas of particular importance include the relation between absolute numbers of events and incidence of sudden cardiac death within defined population pools; the time dependence of risk; and the efficiency of therapeutic interventions.

Risk in Population Subgroups

When sudden cardiac death is measured as the absolute number of events per year in a defined population, it becomes clear that the highest-risk subgroups commonly cited in clinical studies (for instance, patients with low ejection fractions, patients with a history of heart failure, and survivors of out-of-hospital cardiac arrest) do not account for the majority of the events. This statement does not mean that commonly held perceptions about high-risk patients are wrong but rather that the predictive power of these high-risk characteristics applies only to small subgroups. The greater part of the risk for sudden cardiac death is hidden within the larger, more general population pools. These populations have smaller excesses in incidence but generate more events because of the larger size of the population bases from which they emerge. In Figure 1, the magnitude of the risk, expressed as incidence, is compared with the total number of events per year under six different conditions. The estimates are based on published epidemiologic and clinical data [4, 5]. When the more than 300 000 sudden cardiac deaths that occur annually among the unselected adult population in the United States [2, 3] are expressed as a fraction of the total adult population, the overall incidence is 0.1% to 0.2% per year. This calculation includes the 20% to 25% of sudden cardiac death victims whose cardiac arrest is the first clinical manifestation of previously silent or unrecognized heart disease [6-8] plus those with various degrees of increased risk identified by established clinical characteristics or risk factor profiles. When the more easily recognized, very-high-risk subgroups are removed from this population base, the calculated incidence for the remaining patients decreases, further amplifying the problem of identifying specific persons at risk among the general population. On the basis of these estimates, a preventive intervention designed for the general adult population must be applied to the 999 of 1000 people who will not have an event in order to reach and potentially influence the unidentified 1 of 1000 who will. Such limited efficiency impedes application of many active interventions and highlights the need for identifying specific markers of major increases in risk, even among groups with identified risk factors, in order to achieve more effective preventive efforts. Although escalation from a subgroup with several coronary risk factors in the absence of previous clinical events (risk, 1% to 2%/y) to higher-risk subgroups provides the ability to identify high-risk persons with progressively increasing power Figure 1, the absolute number of persons who can be identified decreases with each escalation of risk. Thus, the challenge is not to focus more attention on the highest-risk clinical subgroups that are already clearly profiled but rather to develop new methods that will allow the identification of high-risk clusters within larger subgroups that have lesser degrees of increased risk. The first step toward this goal requires knowledge of the total number of sudden cardiac deaths within a specified population, the fraction of deaths that are sudden within that population, and the total mortality. For example, Kjekshus [9] analyzed a group of studies of heart-failure-related deaths, contrasting the probability of death and the ratio of sudden to total deaths as they relate to functional classification. In studies in which the mean functional class was between class I and class II, the overall death rate was relatively low, but 67% of the deaths were sudden [9]. In contrast, among studies with mean functional classifications close to class IV, the overall death rate was high, but the fraction of sudden deaths was only 29%. In the former circumstance, a therapeutic intervention targeted to all-cause mortality would be less efficient (that is, only a small fraction of the total population exposed to the intervention would have the potential to benefit from it), whereas, for the latter, efficiency would be high. For an intervention specific to the problem of sudden cardiac death, however, efficiency in the former case would be relatively high (67% of all deaths are sudden), whereas for the latter, efficiency for sudden cardiac death would be low because of the dominance of nonsudden deaths.

View larger version (22K): [in this window] [in a new window]   Figure 1. Sudden cardiac deaths among population subgroups. Estimates of incidence (percent per year) and total number of sudden cardiac deaths per year are shown for the overall adult population in the United States and for higher-risk subgroups. The overall estimated incidence is 0.1% to 0.2% per year, totaling more than 300 000 deaths per year. Within subgroups identified by increasingly powerful risk factors, the increasing incidence is accompanied by progressively decreasing total numbers. Practical interventions for the larger subgroups will require identification of higher-risk clusters within the groups. EF = ejection fraction; MI = myocardial infarction; VT/VF = ventricular tachycardia-ventricular fibrillation. The horizontal axis for the incidence figures is nonlinear; see text for details. (From Myerburg, et al [5]; reproduced with permission of the American Heart Association.).

Time Dependence of Risk

The risk for death after surviving a major change in cardiovascular status is not linear over time for most conditions [5]. Survival curves for both sudden and total cardiac deaths show that the most rapid rate of attrition occurs during the first 6 to 18 months after an index event. By 24 months, the slope of the survival curve begins to approach the configuration of one describing a similar population that has remained free of the interposed major cardiovascular event (Figure 2). Further, the shapes of the curves are likely to be influenced by the magnitude of the increased risk after the index cardiovascular event. For instance, data from the Cardiac Arrhythmia Suppression Trial (CAST) [10, 11] show a linear attrition among the relatively low-risk, randomized placebo group during long-term follow-up of these patients with myocardial infarctions complicated by postinfarction arrhythmias but having relatively good mean ejection fractions. Data from the Multicenter Post-Infarction Program [12, 13], however, showed that subgrouping patients according to interactions between frequency of premature ventricular complexes (PVCs) and ejection fractions after they have survived acute myocardial infarction resulted in progressively increasing risk as the number and cumulative power of risk factors increased. The low-risk subgroups generated linear survival curves, whereas the added mortality in higher-risk subgroups tended to be expressed early (Figure 2). The time dependence of risk within the higher-risk groups limits the opportunity for effective intervention strategies to the early periods after the conditioning cardiovascular event. Curves showing these characteristics have been generated from among survivors of out-of-hospital cardiac arrest [14] Figure 2, from patients with new-onset heart failure, and from patients who have had high-risk markers after myocardial infarction [12, 15]. This property of risk must be integrated into strategies designed to intervene in such patients. Controlled intervention trials that allow enrollment of patients more than 12 to 18 months after the cardiovascular event to which the study is indexed might be confounded by a lower-than-anticipated event rate if such entrants are heavily represented in the study group. When the temporal property of risk is applied to individual patients, moreover, the probability of benefit from a secondary preventive intervention is also a function of time of implementation.

View larger version (22K): [in this window] [in a new window]   Figure 2. Time-dependence of risk after cardiovascular events.Top. Survival curves for hypothetical patients with known cardiovascular disease free of a major index event (curve A) and for patients surviving major cardiovascular events (curve C). Attrition is accelerated during the initial 6 to 24 months after the event. Curve B shows the dynamics of risk over time in low-risk patients with an interposed major event that is normalized to a time point (for example, 18 months). The subsequent attrition is accelerated for 6 to 24 months. (Reproduced from [5] with permission of the American Heart Association.) Middle. Kaplan-Meier estimates representing 3 years of follow-up after acute myocardial infarction. Both the frequency of ventricular premature depolarizations (VPD) and depressed ejection fractions (EF) contribute to increasing mortality rates. As risk increases based on the presence of neither (curve A), one (curves B and C), or both (curve D) of these predictors, increasing risk is expressed preferentially within the first year (curves C and D). (Modified from [13]; reproduced with permission of the American Journal of Cardiology). Bottom. Survival curve showing time-dependence of risk for recurrence among survivors of cardiac arrest. The risk was highest during the first 6 months. (Modified from [14]; reproduced from [5] with permission of the American Heart Association.).

Identification and Control of Transient Risk Factors

The PVC hypothesis assumes that PVCs serve as triggers for the initiation of ventricular tachycardia or fibrillation (VT/VF) and that suppression of PVCs protects against sudden cardiac death by eliminating this triggering function. This concept gained popularity from early observations in coronary care units among patients with acute ischemic events [16, 17], and the view of PVCs as primary triggering events for fatal arrhythmias expanded and became ingrained into clinical dogma. However, the clinical application of the PVC hypothesis remained uncertain in clinical settings of chronic risk. Despite the existence of consistent data supporting chronic PVCs as a risk factor in long-term studies [12, 18-21], special circumstances are required to show the initiation of spontaneous life-threatening clinical arrhythmias by PVCs. Incidental ambulatory recordings have shown that the spontaneous onset of ventricular fibrillation may be preceded by a tendency toward increases in sinus rate and PVC frequency [22-24]. This finding may reflect a change in sympathetic tone or hemodynamic status as the intermediary in the relation between PVCs and VT/VF and supports the concept that transiently active factors, rather than the mere existence of chronic PVCs, may be the key pathophysiologic events. Transient influences destabilize ventricular myocardium at a specific point in time, permitting the generation of potentially fatal arrhythmias [5, 25]. They are distinguished from conventional risk factors such as age, sex, hyperlipidemia, or the hypertension-hypertrophy complex (which predispose to underlying disease) and static risk factors that include the structural substrates of preexisting disease. Conventional and static risk factors are present more or less continuously and are likely to be identified any time that a group or an individual is being evaluated. However, the elusive transient risk factors are more difficult to identify at a specific point in time and yet appear to have a central role in establishing instantaneous risk for a fatal arrhythmia. As a result, the early electrogenic concept of sudden cardiac death has been merged into broader pathophysiologic descriptions, which include the roles of both preexisting structural abnormalities as conditioning factors and transient perturbations as the initiating event [5, 25] (Figure 3).

View larger version (38K): [in this window] [in a new window]   Figure 3. Structure, function, and the pathogenesis of sudden cardiac death. Four categories of the conditioning risk factors (structural abnormalities) predisposing to sudden cardiac death interact with one or more inciting influences (functional perturbations) to cause transient destabilization. This interaction may convert premature ventricular complexes (PVCs), which are pathophysiologically innocuous, into triggering events for ventricular tachycardia or fibrillation (VT/VF). See text for details. (Modified from Myerburg RJ, et al [35]; reproduced with permission of the American Journal of Cardiology).

Interactions between structure and function as a pathophysiologic relation with respect to sudden cardiac death received little attention until recently, despite the recognition for many years that a relation between PVCs and VT/VF had a much higher probability of expression if structural heart disease was present. Healed myocardial infarction received the greatest attention as a structural risk factor, largely because of the predominant role of coronary artery disease among the causes of the sudden cardiac death syndrome in Western societies. More recently, insights into other forms of structural abnormalities, particularly left ventricular hypertrophy and the cardiomyopathies, have emerged. The role of hypertrophy as a risk factor for sudden cardiac death has been recognized in epidemiologic studies [26, 27], and clinical associations have been described [28, 29]. New fundamental information on potentially arrhythmogenic membrane channel and electrophysiologic alterations of hypertrophied myocardium is also emerging [30-32], providing further insight into the role of this specific structural abnormality in the pathogenesis of potentially fatal arrhythmias. That regional hypertrophy is common after healing of myocardial infarction [33, 34] raises more questions about the role of hypertrophy-induced electrophysiologic disturbances. The role of hypertrophy in sudden cardiac death has been discussed in greater specific detail previously [4, 5, 25, 35].

After recognizing that the PVC-VT/VF relation subserves structural conditioning factors, clinical investigators and physiologists began to study those factors directly responsible for initiation of a fatal arrhythmia at a specific point in time. The transient nature of these events makes their clinical elucidation difficult. The development of a base of experimental information on the role of myocardial ischemia in establishing the electrophysiologic risk of VT/VF first led to the concept of a transitional event, in which PVCs initiate VT/VF under a predictable set of circumstances [36]. Subsequently, other functional perturbations began to receive attention. Intense functional changes alone may destabilize the system in the absence of structural abnormalities, but almost all cardiac arrests occur in hearts with structural abnormalities. Clinical techniques for predicting susceptibility to transient risk factors, as well as their application as epidemiologic tools, are in their infancy [37, 38]. Transient risk factors are grouped into four categories Figure 3: ischemia and reperfusion; systemic abnormalities; autonomic fluctuations; and cardiotoxic factors, including the problem of proarrhythmia [25].

Transient Ischemia and Reperfusion

Ischemia occurring during the early phase of acute myocardial infarction (that is, the first 48 hours) has a clearly established association with VT/VF. However, only about 20% of sudden cardiac deaths and survivals after out-of-hospital cardiac arrest are associated with acute transmural myocardial infarction [39, 40], and it is assumed that transient ischemia has a major triggering role in a large number of the remaining events. Its transient nature, however, has precluded conducting systematic clinical and epidemiologic studies. Unstable angina pectoris and silent myocardial ischemia also are capable of initiating potentially fatal arrhythmias [41-44], although there is only limited clinical documentation of such mechanisms [45]. Both are associated with a statistical increase in the risk for sudden cardiac death when they accompany preexisting coronary artery disease.

Clinical and epidemiologic data indicating associations between ischemia and potentially fatal arrhythmia are paralleled by experimental data that show the adverse effects of ischemia, especially in the presence of a previous myocardial infarction. For example, a recent study in dogs with healed myocardial infarction was designed to determine the arrhythmogenic effects of graded reductions in blood flow through a non-infarct-related artery. Lesser decreases in blood flow resulted in inducible ventricular tachycardia or spontaneous ventricular fibrillation in the presence of a previous myocardial infarction compared with control animals without a previous infarction [46]. These and other experimental data [36] are providing insight into mechanisms of ischemia-mediated arrhythmias. The data serve both as explanations for the deranged electrophysiology and as targets for treatment. In addition, the effect of left ventricular hypertrophy on risk, especially in the presence of coronary artery disease and previous infarction, is paralleled by reports of specific ion channel abnormalities in hypertrophied myocytes, some of which become manifest primarily during ischemia. These include differences in ATP-sensitive K⁺ channels during ischemia in hypertrophied myocardium compared with normal myocardium and between endocardium and epicardium in normal [31] and ischemic [47] hearts, as well as changes in Ca⁺⁺_andK⁺ currents under conditions of a metabolic surrogate for ischemia [30]. Although the understanding of interactions between epidemiology and membrane physiology is still new, these relations warrant further exploration.

In addition to the initiation of VT/VF by transient ischemia, the role of reperfusion of ischemic muscle on generation of arrhythmias is beginning to be clarified. Reperfusion appears to induce a different type of electrical instability, characterized by very rapid electrical activity, which may be caused by abrupt changes in refractoriness [48] or by generation of afterdepolarizations and triggered activity [49]. Hypertrophied myocytes appear to be more prone to generate reperfusion-induced early afterdepolarizations and triggered activity than are normal myocytes, apparently due to depressed delayed rectifier current (I_k) during reperfusion in the hypertrophied myocyte [32]. In situ studies of ventricular fibrillation during ischemia and reperfusion in hypertrophied hearts support the potential importance of such observations [50]. In addition, a technique of demodulation of the T wave has identified a pattern of T-wave alternans that may relate to susceptibility to potentially fatal arrhythmias during ischemia-reperfusion sequences [51].

Systemic Factors in Transient Risk

Acute or subacute systemic abnormalities modulate chronic structural cardiac abnormalities, influencing electrophysiologic stability and susceptibility to VT/VF and sudden cardiac death [52]. When transient systemic factors responsible for a cardiac arrest can be identified and controlled, other preventive interventions against recurrences may not be required [40]. Acute hypoxemia, acidosis, and electrolyte imbalances all may contribute to transient destabilization [53-55]; these factors are commonly recognizable clinically and are reversible with appropriate therapy. Chronic electrolyte disturbances, most prominently hypokalemia associated with long-term use of diuretics, are associated with an increased risk for cardiovascular mortality [54]. Because hypokalemia is a recognized cause of polymorphic ventricular tachycardia and torsade de pointes [4, 55, 56], most commonly in patients with abnormal hearts and in the presence of class I antiarrhythmic drugs, the association between diuretics and increased mortality has a plausible pathophysiologic basis.

The most common systemic inciting factor, but one that is difficult to study clinically in a controlled manner, is the role of transient hemodynamic dysfunction in patients with structurally abnormal hearts. Severe acute or subacute hemodynamic deterioration may cause a secondary cardiac arrest, which has long been known to carry a very high short-term mortality risk [53, 57]. However, the relation among chronically impaired left ventricular function, acute hemodynamic modulations, and triggering of VT/VF is less well defined and remains an important future focus. Volume loading of isolated canine left ventricles shortens refractory periods [58, 59] and creates regional disparity in hearts with previous myocardial infarction [59]. Stretch-induced modulation of membrane channels may play a role. Despite clinical recognition of an association between heart failure and VT/VF, studies defining mechanisms in humans are limited.

Autonomic Fluctuations and Transient Risk

Systemic, central nervous system, and local cardiac neurophysiologic factors are receiving increasing attention as markers of high-risk subgroups and for explaining mechanisms of VT/VF [60]. At the cardiac level, experimental information [60-66] and limited clinical data [67-74] suggest that a previous myocardial infarction and other cardiac abnormalities are accompanied by changes in cardiac autonomic function. Regionally altered ß-adrenoceptor numbers and changes in coupling proteins and adenylate cyclase activity have been noted in hearts with healed myocardial infarction [66]. Experimental and clinical imaging studies have also shown regional disruption of myocardial sympathetic innervation after acute myocardial infarction, with re-innervation after healing [61, 63, 64, 67]. Clinically, isoproterenol-dependent induction of sustained ventricular tachycardia among cardiac arrest survivors and its prevention by ß-adrenoceptor blockers [68] suggest a role for autonomic stimuli in the genesis of potentially fatal arrhythmias.

At a systemic level, neurophysiologic alterations that modulate cardiac activity have been proposed as a means of identifying increased risk for sudden cardiac death. Changes in heart rate variability and baroreceptor sensitivity have been studied [71-75]. Among survivors of myocardial infarction [71, 73] and survivors of out-of-hospital cardiac arrest [72], altered heart rate variability may be a marker for sudden cardiac death risk. Power-spectrum analysis of heart rate variability in the frequency domain suggests specific patterns of blunted heart rate variability that identify high-risk subgroups [73]. In addition, short-term frequency-domain patterns differ before the onset of sustained ventricular tachycardia compared to nonsustained ventricular tachycardia in the same patient [74]. Blunted baroreceptor sensitivity to a phenylephrine infusion has also been suggested as a marker to identify subgroups at risk for ventricular tachycardia and sudden cardiac death after myocardial infarction [75]. An association between the sinus node rate immediately after the onset of sustained ventricular tachycardia and the electrophysiologic and hemodynamic stability of the ventricular tachycardia has been reported [70]. When ventricular tachycardia is electrically or mechanically unstable, the sinus node rate increased more rapidly during the initial 5 seconds and then decreased abruptly. This pattern, possibly indicating autonomic variability, is absent when ventricular tachycardia is stable. These diverse observations support arguments that abnormal patterns of autonomic function contribute to the identification and possible control of risk for VT/VF and sudden cardiac death.

Effects of Toxic Substances on the Heart

The risk for ventricular fibrillation during chloroform anesthesia was the first recognized relation between a clinically used substance and potentially fatal arrhythmias [76, 77]. Subsequently, the relation between quinidine therapy and the risk for torsade de pointes and ventricular fibrillation [78] identified an association with an antiarrhythmic drug. It is now recognized that classical proarrhythmic responses of this type may occur with any of the class I-A antiarrhythmic drugs and many class III drugs and occur in association with prolongation of the Q-T interval on the electrocardiogram. More subtle, but possibly quite important, are the emerging number of nonantiarrhythmic substances that also may prolong the Q-T interval and induce similar proarrhythmic responses. These include diverse categories of medications, such as erythromycin, pentamidine, many of the psychotropic drugs, and terfenadine [79]. In addition, limited clinical data suggest that diverse substances such as organic phosphate insecticides, cocaine [80], and probucol [81] affect the Q-T interval and increase the risk for torsade de pointes in susceptible persons. For some of these substances, prolongation of the Q-T interval has been shown to be related to an effect on repolarizing currents, such as the delayed rectifier current (I_K+) [79, 80]. An inherent ability of a substance to prolong repolarization and individual susceptibility to this effect may explain the sporadic occurrence of these responses. It follows that the ability to identify individual susceptibility to an adverse ion channel effect might provide a method to identify individual risk prospectively [82]. Unfortunately, because such events are more common in patients with underlying heart disease, discerning the distinction between a proarrhythmic response and a confounding clinical arrhythmia may be difficult.

Measuring the Effects of Interventions

As new preventive and therapeutic approaches to sudden cardiac death evolve, their effect on mortality must be measured accurately. This task will be difficult because of subtleties in evaluating outcome. The Cardiac Arrhythmia Suppression Trial dramatically highlighted a known principle that was sometimes ignored—namely, that identifying the mortality benefit of an intervention requires a concurrently controlled study, regardless of how well we think we know the natural history of the study group [83, 84]. Because of the design difficulties and cost of implementation of large, concurrently controlled studies, investigators have sometimes had a tendency to use surrogates for controlled studies. With regard to sudden cardiac death, there is the added ethical issue of withholding a therapy that, although unproven, might reasonably be expected to benefit persons at risk [85]. Despite these issues, the results of CAST have forced us back to reality in regard to the requirements for establishing a mortality benefit.

The following four issues must be considered in addressing mortality from sudden cardiac death: therapeutic efficacy versus efficiency; competing risks; extension-of-life end points; and equilibrium between antiarrhythmic effects and proarrhythmia.

Therapeutic Efficiency

This term refers to the proportion of a treated population that has the potential to benefit from the intervention. A low therapeutic efficiency means that a large subgroup among the population will receive no benefit from an intervention because only a small fraction is expected to have an event. Despite this, the entire population is exposed to the side effects and possible hazards of the intervention. Conversely, when risk is sufficiently high that large numbers of patients within the population may benefit from the intervention, efficiency is high. This differs from therapeutic efficacy, which is a measure of the benefit of an intervention expressed only in terms of the subgroup within the population having events. An often-discussed example of the problem of therapeutic efficiency is the analysis of data from lipid-lowering trials. Although statistically significant reductions in cardiovascular event rates were observed in the treated group in both the cholestyramine [86] and gemfibrozil [87] studies, questions about overall effect have been debated [88, 89]. One point raised is that more than 90% of the patients in the cholestyramine study group and more than 95% in the gemfibrozil study group would not have a spontaneous event during follow-up, based on data from their respective placebo groups. Because the potential benefit of the intervention was limited to less than 10% of the population, the therapeutic efficiency was low. Therapeutic efficacy, however, was quite good—a 19% reduction in the event rate in the cholestyramine-treated group and a 34% reduction in the gemfibrozil group. When improvements in outcome are expressed as absolute fractions of the total populations (1.7% reduction in the cholestyramine study and 1.4% reduction in the gemfibrozil study), they remain statistically significant, but the inefficiency of the intervention and absolute magnitude of effect are highlighted. Similar cautions about expressing intervention trial data in terms of percent reductions of events, without parallel emphasis on the absolute risk and reductions, have received comments with respect to other studies. Such analyses lead to questions of how aggressive, how costly, how risky, and how much negative effect on quality-of-life is warranted for a given level of therapeutic efficiency. With regard to sudden cardiac death, these questions must be addressed as each new diagnostic, preventive, and therapeutic method becomes available. In all cases, however, both mortality and morbidity must be analyzed. It is possible that low efficiency with respect to mortality might be accompanied by higher efficiency for nonfatal events.

Competing Risks and Therapeutic Efficacy

Two or more excess mortality risks may be present concurrently in a given individual. When applied to cardiovascular mortality, it can be expressed as the competition between risks for sudden cardiac deaths and nonsudden cardiovascular deaths. The efficacy of an intervention with the ability to prevent sudden cardiac death is limited by the extent to which a reduction in this event rate is replaced by competing nonsudden cardiac death or noncardiac death. Thus, a measured reduction in sudden cardiac deaths must be interpreted in terms of the extent to which the benefit is neutralized by other forms of death in order to determine the true mortality benefit. An intervention that is 75% effective in preventing sudden cardiac death, but generates only a 25% reduction in total mortality because sudden deaths are replaced by other forms of death, has a net efficacy of only 25%. A particularly good model for illustrating this point is the analysis of outcome data from studies of the implantable cardioverter-defibrillator. Abundant data show that the implantable cardioverter-defibrillator can predictably revert VT/VF [90-101], and many believe that this translates to a reduction in total cardiovascular mortality. However, invoking the principle of competing risks, reduction in total mortality is not proportionate to the reduction in sudden cardiac death. Concurrently controlled studies will quantitate these relation better than do the current uncontrolled sources of data [85].

Extension-of-Life Compared with End-Point Analysis of Death

For sudden cardiac death interventions, data analyzed for outcome at a predetermined follow-up period, such as 5-year survival rates, may be misleading if interval analyses are not included. For example, in a study comparing implantable defibrillators with conventional therapy in matched control participants, no difference in mortality was observed after 36 months of follow-up [102]. However, interval analysis showed a benefit favoring defibrillator therapy for 12 to 18 months after entry, followed by neutralization of the benefit thereafter. This observation and others [98, 103] suggest that the intervention may delay death within a predetermined evaluation interval.

Extension-of-life analysis is a function of time-dependent risk, which has been modified by an intervention. It is a survival function that is appropriate to conditions in which a basic pathophysiologic process continues to progress over time after a major intervening event (such as cardiac arrest in patients with coronary heart disease), in contrast to those in which an intervention may cure the underlying process (such as 5-year survival after successful treatment of a malignancy). In the latter, 5-year survival is often used as a synonym for cure of the underlying disease; in the former, the underlying disease is not interrupted and remains likely to cause an ultimate fatal event. Accordingly, extension of life is a valid measure of efficacy for the problem of sudden cardiac death, although it still must be interpreted in the context of quality of life [85]. Extension-of-life data are readily available in survival curves generated from studies with negative (placebo) or positive (alternative treatment) control groups [98, 102, 103].

Antiarrhythmic-Proarrhythmic Equilibrium

The benefit of a therapeutic intervention and the risk of its adverse effects are independent variables that may be expressed simultaneously and that may confound outcome data. If an adverse effect includes a risk for death by a mechanism that is not easily distinguished from the clinical target of the therapy, the outcome data reflect an equilibrium between the two actions (Figure 4). For example, in the randomized phase of CAST, the risk for arrhythmic death or cardiac arrest in the placebo group was 2.2%, compared with 5.6% for the encainide- or flecainide-treated subgroup, a 2.64-fold excess risk in the treated group [11]. However, if the randomized group had had a higher risk for sudden cardiac death, while the rate of drug-related deaths remained constant or increased to a lesser degree than clinical risk, the observed adverse equilibrium between antiarrhythmic efficacy and proarrhythmic effects may have decreased in magnitude or been neutralized or reversed. In fact, subgroup analysis of the CAST data did show a trend to a lesser magnitude of increased risk in patients with ejection fractions less than 0.30 (relative risk, 1.97) compared with those patients with ejection fractions greater than 0.30 (relative risk, 3.38), suggesting a shift in equilibrium as a function of changing baseline risk. In addition, in CAST II, in which the risk for sudden cardiac death was higher by design, statistically significant excesses in death rate in the treated group disappeared during the long-term phase of the study [104]. This finding again suggests a more favorable equilibrium between risk and benefit among higher risk populations. (Note that this conclusion may be limited by the electropharmacologic differences between moricizine and the class I-C drugs, which may cause different effects on risks for adverse events.)

View larger version (24K): [in this window] [in a new window]   Figure 4. Antiarrhythmic-proarrhythmic equilibrium. The risk for sudden death, as shown in Figure 1, is used to provide examples of antiarrhythmia-proarrhythmia equilibrium. The untreated risks in the same six groups shown in Figure 1 are indicated by the open bars. Patients with "previous coronary events" and "cardiac arrest survivors" are estimated to have the antiarrhythmic benefit (residual risk) indicated by the superimposed black bars. When the risk for proarrhythmia is added (right panel) in patients with a previous coronary event, based on data from CAST, the residual risk plus the added proarrhythmic risk exceeds untreated risk and results in a net adverse outcome. Among survivors of cardiac arrest, a 10% residual risk with antiarrhythmic therapy based on reported data, plus a 5% added risk for fatal proarrhythmic events, yields an equilibrium that is lower than the estimated untreated risk, resulting in a net benefit. See text for further details.

These comments highlight the difficulty of interpreting therapeutic equilibrium in a circumstance in which therapeutic failure (recurrent clinical arrhythmia) and the major life-threatening adverse event (proarrhythmia) are defined by the same generic event (potentially fatal arrhythmia) (Figure 4). This contrasts sharply with a circumstance in which a life-threatening complication of therapy is easily distinguished from its clinical target. For instance, the failure of penicillin in the treatment of a bacterial pneumonia is recognized by the continuation or progression of a pneumonic febrile illness and a septic state, which is clinically distinct from the life-threatening complication of an acute anaphylactic reaction. Separation of therapeutic failure from adverse events is easy when the differences are so stark. For arrhythmia-proarrhythmia, however, the similarity is important for understanding how to use outcome data and for the design and interpretation of future intervention trials. In Figure 4, examples of antiarrhythmic therapy used in a patient who has had a myocardial infarction (similar to the CAST population) and for a survivor of out-of-hospital cardiac arrest illustrate the concept. In the former, the untreated risk for sudden cardiac death is small, as is the potential benefit of antiarrhythmic therapy. However, the fatal proarrhythmic risk, although not terribly large in absolute terms, does exceed the benefit of therapy, resulting in a net adverse outcome [11]. In the cardiac arrest survivor, the natural history risk is much higher, with greater potential for benefit from therapy. In the example shown, proarrhythmic risk increases to a lesser degree than therapeutic benefit, and a net benefit occurs. The assumptions are based on published data on risk for recurrence, measures of efficacy of therapy, and proarrhythmic risk in such patients. Finally, if individual patients susceptible to proarrhythmic risk can be identified prospectively, a promise derived from new insights into cell membrane physiology [105-107], the equilibrium concept will achieve direct clinical usefulness.

The evolution of definitions of proarrhythmia is another area of interest [25]. Classic proarrhythmia refers to a specific relation between a drug and a host, likely related to specific membrane channel responses to a substance in susceptible persons. It is a new arrhythmia occurring by a mechanism different than the target arrhythmia, with expression shortly after initiation of therapy. It characteristically appears as torsade de pointes and may be caused by any of the class I-A antiarrhythmic drugs and class III drugs. During the evolution of new monitoring and electrophysiologic testing techniques for identifying and quantitating arrhythmias in the 1970s and 1980s, definitions of proarrhythmia expanded to include an unrelated group of changes in rhythm status with undefined long-term risk. Clinical definitions of the expanded forms of proarrhythmia include increases in PVC frequency, the onset of new benign arrhythmias, changes in inducibility status, and evolution of spontaneous nonsustained repetitive forms [108-113]. They occur with most categories of antiarrhythmic agents. The long-term risk implied by these changes is unknown because almost all of the data were generated from short-term studies. A third category, however, highlighted by the CAST data, introduces yet another concept of proarrhythmia-transient proarrhythmic risk that has important clinical implications. In this circumstance, no inherent anomalous relation exists between drug and host in terms of drug effects on specific membrane channels, as in classical proarrhythmia. Rather, the drug and host remain in harmony until a third factor intervenes, which transiently changes the effects of the drug on cardiac electrophysiology. For example, a drug that has powerful Na⁺ -channel blocking action (such as a class I-C agent) may exert a stable and uniform effect on cardiac electrophysiology under steady-state conditions. When transient ischemia occurs at an unpredictable point in time, a disproportionately powerful drug effect on the ischemic area may create greater regional heterogeneity than occurs during ischemia in the absence of the drug, establishing a condition for arrhythmogenesis [114]. Such a phenomenon can be inferred from experimental data reported previously [115] and may also occur as a consequence of regional changes in autonomic status or an interaction with deteriorating hemodynamics. Transient pathophysiologic changes may alter steady-state relation between the drug and the host, creating a transient risk. The continuing class I-C proarrhythmic risk over time in the CAST data [10, 11], in contrast to the early expression expected in classic proarrhythmia [116], can be partially explained by the temporal distribution of ischemic events after a myocardial infarction, with consequent drug-ischemia interaction.

Conclusion

Top Conclusion Author & Article Info References

Both the risk for sudden cardiac death and the potential benefit of interventions should be evaluated in the context of changing patterns of interaction. The static models of sudden cardiac death are not adequate to address the problem: Whether considering population characteristics, temporal distribution of events, risk factors directly responsible for cardiac arrest, or the benefits and adverse effects of preventive therapy, the issues in sudden cardiac death are complex, and progress will come only from addressing the problem in a way that will adjust to these dynamic targets.