The Cardiac Arrhythmia Suppression Trial : Casting Suppression in a Different Light

The Cardiac Arrhythmia Suppression Trial (CAST) was instituted in 1986 by the National Heart, Lung, and Blood Institute (NHLBI) after the completion of a 502-patient pilot study (CAPS).^{1 2 3} The initial results of CAST I was published in 1989 and the CAST II results published in 1992.^{2 3} In both trials, antiarrhythmic drugs effectively suppressed asymptomatic ventricular arrhythmias but increased arrhythmic death. Because the suppression hypothesis was refuted, the common practice of using antiarrhythmic drugs to suppress asymptomatic arrhythmias in patients after acute myocardial infarction has been curtailed. In CAST I, encainide and flecainide-treated patients had a 3.6-fold excessive risk of arrhythmic death compared with placebo-treated patients. The CAST results have been extrapolated to include other antiarrhythmic drugs, resulting in a concern of lethal proarrhythmia when using all antiarrhythmic drugs. Consequently, there have been substantial changes in the labeling of antiarrhythmic drugs and significant changes in the regulatory guidelines from the Food and Drug Administration.⁴ There has also been a dramatic restructuring of antiarrhythmic drug development by the pharmaceutical industry.

At least a decade before the initiation of CAST, it was recognized that myocardial infarction patients with frequent and complex ventricular premature depolarizations (VPDs) detected on ambulatory ECG monitoring had an increased risk of subsequent arrhythmic death as compared with patients without these arrhythmias.^{5 6 7 8} Like most noninvasive markers of increased risk, VPDs lack specificity (3% to 6% arrhythmic death rate in 1 year), meaning that most post–myocardial infarction patients with asymptomatic VPDs survive.⁹ Thus, for each 100 patients exposed to empiric antiarrhythmic therapy, only a few (3 to 6 in 100) can benefit (that is, prevention of arrhythmic death), while, unfortunately, all are at risk for potential lethal proarrhythmia from the antiarrhythmic drug. In the design of CAST, only patients whose arrhythmias were suppressed by the antiarrhythmic drug were entered in the trial. Both CAST trials explored and refuted the hypothesis that drug suppression of asymptomatic VPDs in patients surviving myocardial infarction would improve survival.^{2 3}

An extensive literature confirms the independent predictive value of VPDs and nonsustained ventricular tachycardia in identifying post–myocardial infarction patients at increased risk for arrhythmic death. Frequent and complex VPDs occur more commonly in patients with left ventricular dysfunction.^{6 10 11} Some investigators go further, suggesting that VPDs and nonsustained ventricular tachycardia are simply markers of a dying myocardium.¹² The severity and complexity of asymptomatic VPDs after myocardial infarction are related not only to ejection fraction but to the degree of clinical congestive heart failure, the degree of left ventricular dilation, and the presence of left ventricular aneurysm. VPD and ventricular tachycardia suppression are more difficult to achieve in patients with significant left ventricular dysfunction.^{13 14 15 16 17 18} The documentation of VPD suppression in individual patients is also difficult due to the extreme variability of VPDs.^{19 20} Thus, CAST had to establish fairly vigorous criteria for arrhythmia suppression (80% suppression of VPDs) in order to rule out variability in individual patients.⁸ Sudden death was reduced by approximately one third using empiric β-blockers in the Beta-blocker Heart Attack Trial (BHAT) and timolol trial, yet a post hoc analysis of BHAT revealed that propranolol only suppressed approximately one third of the VPDs.^{21 22} The a priori decision on a target of 80% VPD suppression in CAST affected drug selection and dosing. Testing the hypothesis of VPD suppression was at the heart of the study design of CAST. In the present issue of Circulation, Goldstein and colleagues²³ provide a new perspective on the implications of VPD suppression. They report a post hoc analysis from the CAST I and II database observing that CAST-eligible patients whose VPDs were easy to suppress (first dose of antiarrhythmic drug) had a better prognosis relative to arrhythmic death/nonlethal cardiac arrest. To examine the independence of easy suppression as a predictor of arrhythmic death, standard statistical adjustments were made for variables known to be related to ease of VPD suppression and arrhythmic death (including age, previous myocardial infarction, heart failure, and ejection fraction). The authors report that, after adjustment, ease of VPD suppression was an independent predictor of arrhythmic death (relative risk, .66; P=.013). Because active therapy in both CAST trials was associated with an increase in arrhythmic death and easily suppressed patients were more likely to be randomized in the long-term trial, the strength of this association could be underestimated.

But to what degree should we conclude that ease of suppression by itself is a powerful predictor of mortality? Caution is in order because of interpretation difficulties induced by the unavoidable nonrandom nature of the subgroup’s constitution. This report represents an a priori, nonrandomized subgroup analysis of the CAST I and CAST II populations involving the categories of suppression versus nonsuppression. Entry to these groups was nonrandom. In contrast to the random allocation of group assignment, where imbalances are likely to be small and easily managed by standard statistical adjustments, the nonrandom allocation of risk factors poses a twofold problem. First, the nonrandom selection generally leads to important imbalances in the characteristics of these groups. Second, these imbalances are not conclusively adjusted by statistical procedures.

Multiple studies of a spectrum of antiarrhythmic drugs have noted the association between the degree of left ventricular dysfunction and ventricular arrhythmia suppression;^{15 16 17 18 19 20} there is little debate that the degree of left ventricular dysfunction is the most powerful predictor of mortality after myocardial infarction.^{6 10 11} We are then left with the question of whether the relation between VPD suppression and mortality is independent of left ventricular dysfunction or whether suppression mediates and reflects the more direct influences of left ventricular dysfunction on mortality. Unfortunately, it is these variables that are imbalanced between the comparison groups (left ventricular ejection fraction, prior myocardial infarction, heart failure, and diuretic use), which are each expressions of left ventricular dysfunction and simultaneously are the more powerful predictors of death than ease of suppression itself. In subgroup analysis with substantial confounding, statistical adjustment is not a bromide to establish the validity of a new modest relation in a milieu of well-established, powerful predictors of VPD suppression and mortality.^{6 15 16 17 18 19 20}

Until and unless VPD suppression is shown to improve prognosis in the post–myocardial infarction patient, the results of CAST and other trials emphasize an increased risk of antiarrhythmic therapy not counterbalanced by any demonstrated patient benefit.⁴ In CAST, patients with frequent VPDs were effectively suppressed by moricizine, encainide, and flecainide, but arrhythmic deaths dramatically increased.^{2 3} Yet, in the present report, patients who were easily suppressed had the best outcome.²³ Is this a true paradox? Actually, the present study addresses a new issue, separate from the exploration of the VPD suppression hypothesis refuted by the results of CAST. The previous main CAST results are a reflection of the long-term toxicity of antiarrhythmic drugs despite initial arrhythmia VPD suppression.⁹ The issue addressed by Goldstein et al²³ is a newly proposed marker (ease of VPD suppression) associated with a subsequent low arrhythmic death rate.

If the present relation between ease of VPD suppression and improved prognosis is independently confirmed, one might consider a study based on this marker in distinct contrast to the CAST study design. As illustrated in Fig 1, the original CAST hypothesis was rejected by the use of a study design where only post–myocardial infarction patients with frequent VPDs that were suppressed could be randomized into the trial. In contrast, the study design depicted in Fig 2 could explore the implications of the present report by Goldstein et al.²³ In this proposed design, post–myocardial infarction patients with frequent VPDs would be tested, as in CAST, for arrhythmia suppression. However, those easily suppressed would be ineligible (chronic antiarrhythmic drug therapy in such patients was associated with an increased mortality in CAST). The present report by Goldstein et al²³ concludes that such patients are at low risk for arrhythmic death. Instead, nonsuppressed patients would be the target population, randomized to an empiric antiarrhythmic drug or placebo. Antiarrhythmic drug administration is justified in such a population, based on the observation of an increased arrhythmic death risk compared with patients exhibiting easy suppression. Placebo is justified, based on the facts that these patients have had no symptomatic arrhythmic event and that antiarrhythmic drugs have, in general, proven dangerous to such patients early after myocardial infarction.

Thus, the report by Goldstein et al²³ in the present issue of Circulation is another innovative observation derived from the CAST database: Ease of VPD suppression itself may be a marker of a low arrhythmic death rate in patients surviving myocardial infarction with frequent VPDs. The initial results of CAST and subsequent substudy analyses have enriched our knowledge of arrhythmic death, antiarrhythmic drugs, and ventricular arrhythmias. Our concepts have changed regarding the safety of antiarrhythmic drugs as well as the magnitude of the problem of proarrhythmia as a result of CAST.⁴ CAST continues to have an indispensable influence on antiarrhythmic drug development and the clinical research environment as well as on the practice of arrhythmia management.