Cardiovascular Drugs and Therapy 4: 657-664, 1990 ~e' Kluwer Academic Publishers, Boston. Printed in U.S.A.
Clinical Aspects of Trial Deson: What Can We Expect from the Cardiac Arrhythmia Suppression Trial? J. Thomas Bigger, Jr. Professor of Medicine, Columbia University, New York, New York
S u m m a r y . The controlled clinical trial, i.e., a prospective study that compares the effect of one or more interventions against a control in human subjects, is the most definitive method to determine the benefit/risk ratio for a treatment. Some of the elements are a clear statement of the primary and secondary hypotheses, carefully defined end points, randomization, double blinding, and concomitant controls for the experimental treatment using either placebo or standard treatment. Randomization tends to produce study groups that are comparable with respect to both known and unknown factors that i n f u e n c e the outcome, and also removes bias in the assignment of subjects to treatments to ensure that statistical hypothesis testing will have valid significance levels. An adequate sample size is critically important for a successful trial. The Cardiac Arrhythmia Suppression Trial (CAST) is a randomized, placebo-controlled, double-blind, international, multicenter clinical trial to determine whether suppression of ventricular arrhythmias after myocardial infarction with long-term antiarrhythmic drug treatment will reduce arrhythmic death. Survivors of myocardial infarction who are < 80 years of age and have -> six ventrieular premature depolarizations (VPD) per hour on a 24-hour continuous ECG recording obtained between 6 days and 2 years after myocardial infarction are eligible for CAST. A sample size of 4400 patients was calculated for the main study using the following assumptions: average follow-up of 3 years, arrhythmic death or cardiac arrest has a 3-year incidence of 11% in the placebo group, suppression of ventricular arrhythmias with antiarrhythmic drugs will reduce sudden death by 30%, type I (cO error = 0.025 (one tail), power (1-13) = 0.85, a drop-in rate of 6%, and a drop-out rate of 30%. The primary end point for CAST is arrhythmic death or cardiac arrest. On April 17, 1989, the CAST Data and Safety Monitoring Board recommended that encainide and flecainide be removed from CAST because of strong evidence that these drugs increased the death rate compared to placebo. The 730 patients randomized to encainide or flecainide and the 725 randomized to a corresponding placebo were followed an average of about 300 days. There were 78 deaths or cardiac arrests. Patients taking encainide or flecainide were 3.6 times as likely to experience arrhythmic death or nonfatal cardiac arrest as those treated with placebo. There were no imbalances between the encainide/flecainide group and the placebo group with respect to baseline risk variables that might confound the apparent adverse treatment effect. The increase in the death rate for patients treated with encainide or flecainide was observed consistently in the various subgroups within the study. The relative risk of dying or experiencing a nonfatal cardiac arrest was almost identical for the two drugs (2.3 for encainide
and 2.7 for flecainide). The CAST finding clearly indicates that marked suppression of ventricular arrhythmias after myocardial infarction by encainide or flecainide does not predict improved survival. CAST continues to enroll patients and to randomize them to moricizine or placebo.
Key Words. clinical trials, encainide, sudden cardiac death, flerainide, cardiac arrest, moricizine
I n this communication, we will discuss the design of clinical trials to evaluate the ability of cardiac antiarrhythmic drugs to reduce mortality and lifethreatening, hemodynamically significant ventricular arrhythmias, e.g., ventricular fibrillation or sustained ventricular tachycardia. We will focus on the design of the Cardiac Arrhythmia Suppression Trial (CAST) in order to discuss what we are likely to learn from this multicenter, international study. Also, we will present a status report from the CAST.
Clinical Trials A clinical trial is defined as a prospective study that compares the effect of one o1" more interventions against a control in human subjects. The controlled clinical trial is the most definitive method to determine the benefit/risk ratio for a treatment. The controlled clinical trial is a relatively recent scientific development. The concept of randomization in clinical trials is attributed to R.A. Fisher in 1926 [1], and the first use of the concept in a clinical experiment to Amberson in 1931 [2]. However, the first use of randomization to assign individual subjects to alternate treatments was by A.B. Hill in 1948, and the first randomized, blinded clinical trial was conducted by A.B. Hill in 1950 [3]. A controlled clinical trial to obtain a point estimate of
Address for correspondence and reprint requests: J. Thomas Bigger, Jr., M.D., Arrhythmia Control Unit, Columbia-Presbyterian Medical Center, 630 West 168th Street, New York, NY 10032. 657
658
Bigger
treatment effect on a continuous variable, such as arterial blood pressure, may involve a small number of patients, 20 or less. Controlled clinical trials designed to determine the effect of treatment on mortality in populations suffering from cardiovascular diseases usually include several thousand patients. Prerequisites for a clinical trial The conduct of a clinical trial depends on the determination of feasibility in preliminary studies [4]. There must be some preliminary evidence of efficacy for the treatment that is proposed for the trial. The safety of the treatment must be established, and there must be some stability for the treatment. If the conventional treatment is changing rapidly, e.g., constant evolution of coronary artery bypass graft surgery, then the results of a trial will have doubtful applicability. The appropriate end point for the trial must be defined and methods to evaluate the end points must be available. An excellent example of problems with end points is sudden cardiac death, a term that suggests arrhythmic death. If sudden cardiac death can be equated with arrhythmic death, treatments that suppress arrhythmias should reduce mortality due to sudden death. At present these concepts do not find much scientific evidence to support them. Sudden death is difficult to define in the first place. Unexpected death occun'ing within 1 hour of the onset of new symptoms is a definition that has been used often in recent years. However, there are many problems in applying this definition. It is often difficult to define "unexpected" and to determine the time of onset of new symptoms. Also, most studies include some types of unwitnessed deaths in the definition of sudden death. Attempts to define death as "arrhythmic," such as the HinkleThaler method [5,6] or the definition used in the Cardiac Arrhythmia Pilot Study [6], also have difficulties. Interestingly, definitions of an-hythmic death classify patients quite differently than sudden death, i.e., these terms are far from synonymous [6,7]. Selection of a population The selection of the population for a clinical trial is critically important. Trial experience has taught us that we need to select the sicker patients for the trial. The risk/benefit ratio for participating in a clinical trial is poor for a patient with a very low risk of experiencing an outcome. Also, the patient being enrolled in a trial should be at risk for an outcome that is addressed by the clinical trial. For example, patients who are selected for a trial aimed at the prevention of sudden cardiac death with antiarrhythmic drugs should have a reasonably high risk of arrhythmic death. Usually patients are screened at baseline for a predictor of future
Table I. Eleme~ts q/ clstical trial desig~
Primary and secondary hypotheses stated End points defined Sample size computed 9 Event rate ill control group 9 Magnitude of treatment group 9 Alpha and beta levels 9 Duration of follow-up 9 Drop-in and drop-out rates Randomization procedures selected 9 Stratification 9 Permuted blocks 9 Adaptive methods Double-blind procedures outlined Data collection, quality control procedm'es specified Stopping rules, close-out procedures specified Analysis and interpretation specified
arrhythmic death, e.g., frequent and repetitive ventricular arrhythmias or late potentials in a signal averaged ECG. One of the flaws in the initial antiarrhythmic drug trials in the 1970s was the inclusion of patients with recent myocardial infarction who did not have ventricular arrhythmias. In fact, about 80% of those enrolled did not have significant ventricular arrhythmias. On the other hand, the sickest patients usually are excluded from controlled clinical trials because conventional treatment is used. For example, patients who have ventricular fibrillation or hemodynamically significant sustained ventricular tachycardia are usually managed by programmatic evaluation of antiarrhythmic therapy, rather than being enrolled in controlled clinical trials. In general, patients are excluded from clinical trials if it is judged that they would not benefit from the treatment, would be harmed by the treatment, or would not comply with the treatment. Design of a clinical trial Some of the elements of the design of a clinical trial are listed in Table 1. A clear statement of the primary and secondary hypotheses is the starting point in the design. The end points also need to be defined carefully. Key elements in the design of a clinical trial are randomization, double-blinding, and concomitant controls for the experimental treatment, using either placebo or standard treatment. By definition, a clinical trial must have a control group to contrast with the treatment group. If no treatment is established as having a better risk/benefit ratio than placebo, placebo is the obvious best choice treatment for the control group. If a treatment is established as having a better risk/benefit ratio than placebo, the established
What Can We E.rpect From Cast?
Table 2. Achievement.v qt the Cardiac Arrh!tthmia Pilot Shr
Effective screening and enrollment methods w e r e developed Natural history and variability of VPD arm- infarction determined Determine how to use 24-hour ECGs to determine drug efficacy Tilerapeutie strategies to suppress ventrieular arrhythmias were found 9 Dose ranging 9 Drug changes Adverse effect rates of antiarrhythmie drugs were determined 9 Proarrhythmia 9 Congestive heart failure 9 Other adverse effects 9 Drop-out rates Stress and behavioral effects were evaluated 9 On ventrieular arrhythmia rates 9 On response to antiarrhythmie drugs VPD = ventricuhu"l)remature depolarizations.
treatment is the best choice for the control group. Randomization has become a standard feature for clinical trials, because it tends to produce study groups that are comparable with respect to both known and unknown factors that influence outcome. Randomization also removes bias in the assignment of subjects to treatments and ensures that statistical hypothesis testing will have valid significance levels. Randomization techniques have become quite sophisticated, ineluding stratification on the strongest proDmstie factors and permuted block randomization to ensure balance in the number of patients assigned to each group. Randomization may also be adaptive on baseline variables, treatment assignment, or response to treatment. Computation of sample size is critically important for a suceessful trial and requires that the following be specified: the event rate in the control gn'oup; the expected magnitude of the treatment effect; the average length of follow-up; one- or two-sided hypothesis; ~ level, the probability of finding a difference between treatment and control gTOUpS when there is none (type I error); [3 level, the probability of failing to find a difference between treatment and control when there is one (type II error); and the drop-in and drop-out rates. The event rate in the control gq'oup usually is known from previous studies, although previous studies often provide information in a slightly different sample than the sample selected fo2" the trial. The previous study may have been done at a time when conventional concomitant treatment was different. The magnitude of the treatment effect is usually unknown and is selected based on a minimum value of clinical interest. By tradition, the u and [3
Inloletant O l ~ Ircre3sed ,3rrhu OF 51ill in tllrat on ,35 o~ ~ 0 89 ,L:7
Case F,rd,ng
a
t
ID
On.g dcso Se 15 consecutive complexes at a rate of -> 120/rain or patients with symptomatic hemodynamically important ventricular tachycardia are excluded from CAST. Also, patients are excluded who, at the time of enrollment, have New York Heart Association class IV congestive heart failure el" Canadian Cardiovascular class IV angina pectoris.
During the open-label drug/dose selection phase, three active drugs are used in a number of sequences that are randomly assigned. For patients with an election fi'action -> 309~, the sequence will contain all three drugs--encainide, flecainide, or moricizine--in one of two sequences:
01"
fleeainide ~ moricizine + encainide. For patients with left ventricular ejection fraction < 30c~, only two drugs, encainide and moricizine, are used. Patients are randomly assigned to start encainide or moricizine first. Each drug has two close levels (Table 3) and dosing always proceeds fl'om the low to the high close. The two closes used are the same as the two lower doses used in CAPS. Suppression is assessed by a 24-hour continuous ECG recording. The CAST definition of suppression is -> 80c~ reduction in VPD fl'equency and -> 90~ suppression of runs of unsustained ventricular tachycardia (;3-14 consecutive VPD at a rate of -> 120/min). Patients whose arrhythmia is suppressed and who tolerate their treatment are randomized to the drug that suppressed their arrhythmia el" to a matching placebo. Randomization is stratified on: a) clinical center; b) lek ventricular ejection fraction, -> 30~ or < 30(2 ; c) time of enrollment, 6 to 90 days or -> 90 clays to 2 years. Randomization to blinded treatment with drug or placebo begins the trial. Enrolhnent will continue l%r ;3 or 4 years (fl'om 1987 to 1990 or 1991) and follow-up for 2 to 5 years (until 1992 or 1993).
Sample size A sample size of 4400 patients was calculated for the main study using the following assumptions (Table 4). The average follow-up will last 3 years. The primary end point, arrhythmic death el" cardiac arrest, has a 3year incidence of llC~ in the placebo gToup. Suppression of ventricular arrhythmias with antiarrhythmic drugs will reduce sudden death by 30~. The type I (a)
What Ca~ We Expect From Cast?
LVEF 30-55% - 24 hours, these symptoms will not be considered part of the death event. Sustained ventricular tachycardia with hemodynamic symptoms and requiring cardioversion is not considered an arrhythmic death. Secondary end points are a) total cardiac mortality, b) combined arrhythmic death and sustained ventrieular tachycardia, and c) total mortality (cardiac and noncardiac).
Substudy for patients whose V P D s a r e not suppressed Assuming that 80% of the patients will achieve the CAST criteria for arrhythmia suppression during open-label titration, less than 1000 patients will be enrolled into the substudy for patients who do not achieve adequate suppression of VPDs during openlabel titration. So far, about 7% of the patients entering open-label titration have failed to have their arrhythmia suppressed and have entered the treatment substudy. It is recognized that 1000 patients is not enough to permit a definitive conclusion about the treatment hypothesis, but it was thought that important trends should be demonstrable. For example, if the group treated with antiarrhythmic drugs without marked suppression of VPDs have a significant reduction in arrhythmic mortality similar to the group that achieved marked suppression of VPDs, it will suggest that treatment not suppression should be the goal of treatment of ventricular arrhythmias after myocardial infarction. Also, the relationship between the magnitude of the reduction in arrhythmic death and the degree of VPD suppression will improve the interpretation of the findings in the main trial. If the degree of VPD suppression is related monotonically and directly to survival rate, it will strongly suggest that suppression of ventricular arrhythmias is causally related to the improvement in survival. Also, a comparison of the placebo group in the suppressed group (main trial) and the nonsuppressed group (substudy) will indicate the degree to which non-suppression is an indicator of disease severity.
Substudy of long-term suppression and ventricular arrhythmia variability It is important to know the degree of continued suppression of ventricular arrhythmias during CAST follow-up, particularly for interpreting a negative result and for evaluating the treatment hypothesis. Approximately 1200 will be randomly selected from the CAST patients to participate in a substudy that will evaluate long-term suppression of ventricular alThythmias. These patients will have two 24-hour Holter recordings, one at the 4-month follow-up visit and another at
662
Bigger
variable times during follow-up to provide information about the entire follow-up period. The purpose of the tapes recorded in the placebo group is to maintain the blind in the study. The variability substudy will characterize the variability in ventricular arrhythmias at two time points, 4 and 12 months after enrollment, to determine the length of Holter monitoring necessary to evaluate initial and continuing suppression in individual patients. For the variability substudy, 300 patients will have two 24-hour Holter ECG recordings 1 week apart at 4 months after enrollment, and another 300 patients will have two 24-hour Holter ECG recordings 1 week apart 12 months after enrollment. The purpose of the tapes recorded in the active treatment group is to maintain the blind in the study.
Signal-averaged ECG ancillary study The CAST signal-averaged ECG ancillary study will evaluate the predictive value of the sigual-averaged ECG for predicting arrhythmic death, cardiac arrest, and sustained ventricular tachycardia. The sample size calculation of 1400 patients was based on the following parameters: a) 25~ prevalence of abnormal signal-averaged ECG; b) follow-up of 1 year; c) primary event rate of 5% with normal sigual-averaged ECG; d) primary event rate of 10c~ with abnormal sigualaveraged ECG; e) type I (e) error of 0.05, one-tailed test; and f) type II ([3) error of 0.20. The patients for the signal-averaged ECG study will be recruited from the CAST Holter Registry. To be eligible for this study patients must: a) be recruited and have a sig~alaveraged ECG recorded between 6 and 30 clays after myocardial infarction, b) be < 80 years of age at the time of the CAST qualifying Holter recording, c) have a technically adequate 24-hour Holter recording, and d) give informed consent. Patients do not have to meet any VPD frequency or left ventricular ejection fraction criteria to be eligible, i.e., they do not have to be eligible to enroll in CAST. Since the study is studying the entire postinfaretion population, its results should be broadly generalizable. Other baseline information includes historical information, 12-lead ECG, and left ventricular ejection fraction. The primary hypothesis for this study is that the signal-averaged ECG predicts arrhythmic events during follow-up. A secondary hypothesis is that the signal-averaged ECG predicts arrhythmic events during follow-up independently of ventricular mThythmias detected by 24-hour Holter recording or left ventricular ejection fl'action. This large multicenter study of the signal-averaged ECG should clarify the role of late potentials in postinfarction risk stratification. Potentially, this study can improve the selection of patients for clinical studies or conventional treatment with therapies aimed at reduc-
ing arrhythmic death and symptomatic nonfatal arrhythmic events.
Programmed electrical stimulation study One of the objectives of CAST is to detect proarrhythmic episodes and to establish their prevalence, contributing causes, and outcomes. The objective of this study is to perform programmed electrical stimulation in patients who develop sustained ventricular tachycardia or ventricular fibrillation while taking a CAST drug to determine the likelihood that the CAST drug contributed to or caused the sustained arrhythmia. Two studies with programmed electrical stimulation will be done, one on the CAST drug and one off of the CAST drug. If the patient has an inducible sustained ventricular arrhythmia on the CAST drug, but is not inducible off of the CAST drug, the sustained arrhythmia will be attributed to an adverse drug effect.
Population pharmacokinetic study Ancillary studies of the plasma concentration (fl'ee and protein bound), half-time for elimination, and metabolism of the CAST drugs are being conducted to determine whether the plasma concentration of parent compounds or metabolites are related to the degree of suppression of ventricular arrhythmias o1"an effect on arrhythmic end points dm'ing follow-up. A random sample of about 1000 patients will participate in this study. The results should be helpful to clinicians using these drugs to treat patients after myocardial infarction by clarifying the need for assessing plasma drug concentrations, the likelihood that a given dose will produce therapeutic plasma drug concentrations, and the dependence of plasma drug concentration on body weight, left ventricular ejection fi'action and renal or hepatic function.
Encainide and fleeainide increase mortality in C A S T
The CAST protocol permits drugs to be removed from or added to the trial. On April 17, 1989, the Data and Safety Monitoring Board held its regular semiannual meeting, reviewed mortality data complete as of March 30, 1989, and recommended that encainide and flecainide be removed from CAST because it had become virtually impossible to establish any beneficial effect fi'om these drugs if they were continued to the end of study and because of strong evidence that these drugs increased the death rate compared to placebo. The Data and Safety Monitoring Board also recommended that CAST be continued with moricizine, which still has a significant chance of showing benefit,
What Ca~ We Expect From Cast?
and that other antiarrhythmic drugs be reviewed as possible additions to CAST. The National Heart, Lung, and Blood Institute and the CAST Executive Committee reviewed the April 17, 1989 recommendation by the Data and Safety Monitoring Board and decided that same (lay to remove encainide and fiecainide from CAST immediately. On April 18, 1989, a message was sent to each CAST clinical center to discontinue treatment in all patients who had been assigned to encainide, fiecainide, or their corresponding placebos. This process was started on April 19 and was completed by April 24. Patients whose encainide or fiecainide was stopped had a 24-hour ECG recording done 7 to 30 clays after discontinuing their study drug. Those that still had sufficient ventricular arrhythmia frequency to qualify after washout of their original CAST medication, had a left ventricular ejection fraction
Clinical Aspects of Trial Deson: What Can We Expect from the Cardiac Arrhythmia Suppression Trial? J. Thomas Bigger, Jr. Professor of Medicine, Columbia University, New York, New York
S u m m a r y . The controlled clinical trial, i.e., a prospective study that compares the effect of one or more interventions against a control in human subjects, is the most definitive method to determine the benefit/risk ratio for a treatment. Some of the elements are a clear statement of the primary and secondary hypotheses, carefully defined end points, randomization, double blinding, and concomitant controls for the experimental treatment using either placebo or standard treatment. Randomization tends to produce study groups that are comparable with respect to both known and unknown factors that i n f u e n c e the outcome, and also removes bias in the assignment of subjects to treatments to ensure that statistical hypothesis testing will have valid significance levels. An adequate sample size is critically important for a successful trial. The Cardiac Arrhythmia Suppression Trial (CAST) is a randomized, placebo-controlled, double-blind, international, multicenter clinical trial to determine whether suppression of ventricular arrhythmias after myocardial infarction with long-term antiarrhythmic drug treatment will reduce arrhythmic death. Survivors of myocardial infarction who are < 80 years of age and have -> six ventrieular premature depolarizations (VPD) per hour on a 24-hour continuous ECG recording obtained between 6 days and 2 years after myocardial infarction are eligible for CAST. A sample size of 4400 patients was calculated for the main study using the following assumptions: average follow-up of 3 years, arrhythmic death or cardiac arrest has a 3-year incidence of 11% in the placebo group, suppression of ventricular arrhythmias with antiarrhythmic drugs will reduce sudden death by 30%, type I (cO error = 0.025 (one tail), power (1-13) = 0.85, a drop-in rate of 6%, and a drop-out rate of 30%. The primary end point for CAST is arrhythmic death or cardiac arrest. On April 17, 1989, the CAST Data and Safety Monitoring Board recommended that encainide and flecainide be removed from CAST because of strong evidence that these drugs increased the death rate compared to placebo. The 730 patients randomized to encainide or flecainide and the 725 randomized to a corresponding placebo were followed an average of about 300 days. There were 78 deaths or cardiac arrests. Patients taking encainide or flecainide were 3.6 times as likely to experience arrhythmic death or nonfatal cardiac arrest as those treated with placebo. There were no imbalances between the encainide/flecainide group and the placebo group with respect to baseline risk variables that might confound the apparent adverse treatment effect. The increase in the death rate for patients treated with encainide or flecainide was observed consistently in the various subgroups within the study. The relative risk of dying or experiencing a nonfatal cardiac arrest was almost identical for the two drugs (2.3 for encainide
and 2.7 for flecainide). The CAST finding clearly indicates that marked suppression of ventricular arrhythmias after myocardial infarction by encainide or flecainide does not predict improved survival. CAST continues to enroll patients and to randomize them to moricizine or placebo.
Key Words. clinical trials, encainide, sudden cardiac death, flerainide, cardiac arrest, moricizine
I n this communication, we will discuss the design of clinical trials to evaluate the ability of cardiac antiarrhythmic drugs to reduce mortality and lifethreatening, hemodynamically significant ventricular arrhythmias, e.g., ventricular fibrillation or sustained ventricular tachycardia. We will focus on the design of the Cardiac Arrhythmia Suppression Trial (CAST) in order to discuss what we are likely to learn from this multicenter, international study. Also, we will present a status report from the CAST.
Clinical Trials A clinical trial is defined as a prospective study that compares the effect of one o1" more interventions against a control in human subjects. The controlled clinical trial is the most definitive method to determine the benefit/risk ratio for a treatment. The controlled clinical trial is a relatively recent scientific development. The concept of randomization in clinical trials is attributed to R.A. Fisher in 1926 [1], and the first use of the concept in a clinical experiment to Amberson in 1931 [2]. However, the first use of randomization to assign individual subjects to alternate treatments was by A.B. Hill in 1948, and the first randomized, blinded clinical trial was conducted by A.B. Hill in 1950 [3]. A controlled clinical trial to obtain a point estimate of
Address for correspondence and reprint requests: J. Thomas Bigger, Jr., M.D., Arrhythmia Control Unit, Columbia-Presbyterian Medical Center, 630 West 168th Street, New York, NY 10032. 657
658
Bigger
treatment effect on a continuous variable, such as arterial blood pressure, may involve a small number of patients, 20 or less. Controlled clinical trials designed to determine the effect of treatment on mortality in populations suffering from cardiovascular diseases usually include several thousand patients. Prerequisites for a clinical trial The conduct of a clinical trial depends on the determination of feasibility in preliminary studies [4]. There must be some preliminary evidence of efficacy for the treatment that is proposed for the trial. The safety of the treatment must be established, and there must be some stability for the treatment. If the conventional treatment is changing rapidly, e.g., constant evolution of coronary artery bypass graft surgery, then the results of a trial will have doubtful applicability. The appropriate end point for the trial must be defined and methods to evaluate the end points must be available. An excellent example of problems with end points is sudden cardiac death, a term that suggests arrhythmic death. If sudden cardiac death can be equated with arrhythmic death, treatments that suppress arrhythmias should reduce mortality due to sudden death. At present these concepts do not find much scientific evidence to support them. Sudden death is difficult to define in the first place. Unexpected death occun'ing within 1 hour of the onset of new symptoms is a definition that has been used often in recent years. However, there are many problems in applying this definition. It is often difficult to define "unexpected" and to determine the time of onset of new symptoms. Also, most studies include some types of unwitnessed deaths in the definition of sudden death. Attempts to define death as "arrhythmic," such as the HinkleThaler method [5,6] or the definition used in the Cardiac Arrhythmia Pilot Study [6], also have difficulties. Interestingly, definitions of an-hythmic death classify patients quite differently than sudden death, i.e., these terms are far from synonymous [6,7]. Selection of a population The selection of the population for a clinical trial is critically important. Trial experience has taught us that we need to select the sicker patients for the trial. The risk/benefit ratio for participating in a clinical trial is poor for a patient with a very low risk of experiencing an outcome. Also, the patient being enrolled in a trial should be at risk for an outcome that is addressed by the clinical trial. For example, patients who are selected for a trial aimed at the prevention of sudden cardiac death with antiarrhythmic drugs should have a reasonably high risk of arrhythmic death. Usually patients are screened at baseline for a predictor of future
Table I. Eleme~ts q/ clstical trial desig~
Primary and secondary hypotheses stated End points defined Sample size computed 9 Event rate ill control group 9 Magnitude of treatment group 9 Alpha and beta levels 9 Duration of follow-up 9 Drop-in and drop-out rates Randomization procedures selected 9 Stratification 9 Permuted blocks 9 Adaptive methods Double-blind procedures outlined Data collection, quality control procedm'es specified Stopping rules, close-out procedures specified Analysis and interpretation specified
arrhythmic death, e.g., frequent and repetitive ventricular arrhythmias or late potentials in a signal averaged ECG. One of the flaws in the initial antiarrhythmic drug trials in the 1970s was the inclusion of patients with recent myocardial infarction who did not have ventricular arrhythmias. In fact, about 80% of those enrolled did not have significant ventricular arrhythmias. On the other hand, the sickest patients usually are excluded from controlled clinical trials because conventional treatment is used. For example, patients who have ventricular fibrillation or hemodynamically significant sustained ventricular tachycardia are usually managed by programmatic evaluation of antiarrhythmic therapy, rather than being enrolled in controlled clinical trials. In general, patients are excluded from clinical trials if it is judged that they would not benefit from the treatment, would be harmed by the treatment, or would not comply with the treatment. Design of a clinical trial Some of the elements of the design of a clinical trial are listed in Table 1. A clear statement of the primary and secondary hypotheses is the starting point in the design. The end points also need to be defined carefully. Key elements in the design of a clinical trial are randomization, double-blinding, and concomitant controls for the experimental treatment, using either placebo or standard treatment. By definition, a clinical trial must have a control group to contrast with the treatment group. If no treatment is established as having a better risk/benefit ratio than placebo, placebo is the obvious best choice treatment for the control group. If a treatment is established as having a better risk/benefit ratio than placebo, the established
What Can We E.rpect From Cast?
Table 2. Achievement.v qt the Cardiac Arrh!tthmia Pilot Shr
Effective screening and enrollment methods w e r e developed Natural history and variability of VPD arm- infarction determined Determine how to use 24-hour ECGs to determine drug efficacy Tilerapeutie strategies to suppress ventrieular arrhythmias were found 9 Dose ranging 9 Drug changes Adverse effect rates of antiarrhythmie drugs were determined 9 Proarrhythmia 9 Congestive heart failure 9 Other adverse effects 9 Drop-out rates Stress and behavioral effects were evaluated 9 On ventrieular arrhythmia rates 9 On response to antiarrhythmie drugs VPD = ventricuhu"l)remature depolarizations.
treatment is the best choice for the control group. Randomization has become a standard feature for clinical trials, because it tends to produce study groups that are comparable with respect to both known and unknown factors that influence outcome. Randomization also removes bias in the assignment of subjects to treatments and ensures that statistical hypothesis testing will have valid significance levels. Randomization techniques have become quite sophisticated, ineluding stratification on the strongest proDmstie factors and permuted block randomization to ensure balance in the number of patients assigned to each group. Randomization may also be adaptive on baseline variables, treatment assignment, or response to treatment. Computation of sample size is critically important for a suceessful trial and requires that the following be specified: the event rate in the control gn'oup; the expected magnitude of the treatment effect; the average length of follow-up; one- or two-sided hypothesis; ~ level, the probability of finding a difference between treatment and control gTOUpS when there is none (type I error); [3 level, the probability of failing to find a difference between treatment and control when there is one (type II error); and the drop-in and drop-out rates. The event rate in the control gq'oup usually is known from previous studies, although previous studies often provide information in a slightly different sample than the sample selected fo2" the trial. The previous study may have been done at a time when conventional concomitant treatment was different. The magnitude of the treatment effect is usually unknown and is selected based on a minimum value of clinical interest. By tradition, the u and [3
Inloletant O l ~ Ircre3sed ,3rrhu OF 51ill in tllrat on ,35 o~ ~ 0 89 ,L:7
Case F,rd,ng
a
t
ID
On.g dcso Se 15 consecutive complexes at a rate of -> 120/rain or patients with symptomatic hemodynamically important ventricular tachycardia are excluded from CAST. Also, patients are excluded who, at the time of enrollment, have New York Heart Association class IV congestive heart failure el" Canadian Cardiovascular class IV angina pectoris.
During the open-label drug/dose selection phase, three active drugs are used in a number of sequences that are randomly assigned. For patients with an election fi'action -> 309~, the sequence will contain all three drugs--encainide, flecainide, or moricizine--in one of two sequences:
01"
fleeainide ~ moricizine + encainide. For patients with left ventricular ejection fraction < 30c~, only two drugs, encainide and moricizine, are used. Patients are randomly assigned to start encainide or moricizine first. Each drug has two close levels (Table 3) and dosing always proceeds fl'om the low to the high close. The two closes used are the same as the two lower doses used in CAPS. Suppression is assessed by a 24-hour continuous ECG recording. The CAST definition of suppression is -> 80c~ reduction in VPD fl'equency and -> 90~ suppression of runs of unsustained ventricular tachycardia (;3-14 consecutive VPD at a rate of -> 120/min). Patients whose arrhythmia is suppressed and who tolerate their treatment are randomized to the drug that suppressed their arrhythmia el" to a matching placebo. Randomization is stratified on: a) clinical center; b) lek ventricular ejection fraction, -> 30~ or < 30(2 ; c) time of enrollment, 6 to 90 days or -> 90 clays to 2 years. Randomization to blinded treatment with drug or placebo begins the trial. Enrolhnent will continue l%r ;3 or 4 years (fl'om 1987 to 1990 or 1991) and follow-up for 2 to 5 years (until 1992 or 1993).
Sample size A sample size of 4400 patients was calculated for the main study using the following assumptions (Table 4). The average follow-up will last 3 years. The primary end point, arrhythmic death el" cardiac arrest, has a 3year incidence of llC~ in the placebo gToup. Suppression of ventricular arrhythmias with antiarrhythmic drugs will reduce sudden death by 30~. The type I (a)
What Ca~ We Expect From Cast?
LVEF 30-55% - 24 hours, these symptoms will not be considered part of the death event. Sustained ventricular tachycardia with hemodynamic symptoms and requiring cardioversion is not considered an arrhythmic death. Secondary end points are a) total cardiac mortality, b) combined arrhythmic death and sustained ventrieular tachycardia, and c) total mortality (cardiac and noncardiac).
Substudy for patients whose V P D s a r e not suppressed Assuming that 80% of the patients will achieve the CAST criteria for arrhythmia suppression during open-label titration, less than 1000 patients will be enrolled into the substudy for patients who do not achieve adequate suppression of VPDs during openlabel titration. So far, about 7% of the patients entering open-label titration have failed to have their arrhythmia suppressed and have entered the treatment substudy. It is recognized that 1000 patients is not enough to permit a definitive conclusion about the treatment hypothesis, but it was thought that important trends should be demonstrable. For example, if the group treated with antiarrhythmic drugs without marked suppression of VPDs have a significant reduction in arrhythmic mortality similar to the group that achieved marked suppression of VPDs, it will suggest that treatment not suppression should be the goal of treatment of ventricular arrhythmias after myocardial infarction. Also, the relationship between the magnitude of the reduction in arrhythmic death and the degree of VPD suppression will improve the interpretation of the findings in the main trial. If the degree of VPD suppression is related monotonically and directly to survival rate, it will strongly suggest that suppression of ventricular arrhythmias is causally related to the improvement in survival. Also, a comparison of the placebo group in the suppressed group (main trial) and the nonsuppressed group (substudy) will indicate the degree to which non-suppression is an indicator of disease severity.
Substudy of long-term suppression and ventricular arrhythmia variability It is important to know the degree of continued suppression of ventricular arrhythmias during CAST follow-up, particularly for interpreting a negative result and for evaluating the treatment hypothesis. Approximately 1200 will be randomly selected from the CAST patients to participate in a substudy that will evaluate long-term suppression of ventricular alThythmias. These patients will have two 24-hour Holter recordings, one at the 4-month follow-up visit and another at
662
Bigger
variable times during follow-up to provide information about the entire follow-up period. The purpose of the tapes recorded in the placebo group is to maintain the blind in the study. The variability substudy will characterize the variability in ventricular arrhythmias at two time points, 4 and 12 months after enrollment, to determine the length of Holter monitoring necessary to evaluate initial and continuing suppression in individual patients. For the variability substudy, 300 patients will have two 24-hour Holter ECG recordings 1 week apart at 4 months after enrollment, and another 300 patients will have two 24-hour Holter ECG recordings 1 week apart 12 months after enrollment. The purpose of the tapes recorded in the active treatment group is to maintain the blind in the study.
Signal-averaged ECG ancillary study The CAST signal-averaged ECG ancillary study will evaluate the predictive value of the sigual-averaged ECG for predicting arrhythmic death, cardiac arrest, and sustained ventricular tachycardia. The sample size calculation of 1400 patients was based on the following parameters: a) 25~ prevalence of abnormal signal-averaged ECG; b) follow-up of 1 year; c) primary event rate of 5% with normal sigual-averaged ECG; d) primary event rate of 10c~ with abnormal sigualaveraged ECG; e) type I (e) error of 0.05, one-tailed test; and f) type II ([3) error of 0.20. The patients for the signal-averaged ECG study will be recruited from the CAST Holter Registry. To be eligible for this study patients must: a) be recruited and have a sig~alaveraged ECG recorded between 6 and 30 clays after myocardial infarction, b) be < 80 years of age at the time of the CAST qualifying Holter recording, c) have a technically adequate 24-hour Holter recording, and d) give informed consent. Patients do not have to meet any VPD frequency or left ventricular ejection fraction criteria to be eligible, i.e., they do not have to be eligible to enroll in CAST. Since the study is studying the entire postinfaretion population, its results should be broadly generalizable. Other baseline information includes historical information, 12-lead ECG, and left ventricular ejection fraction. The primary hypothesis for this study is that the signal-averaged ECG predicts arrhythmic events during follow-up. A secondary hypothesis is that the signal-averaged ECG predicts arrhythmic events during follow-up independently of ventricular mThythmias detected by 24-hour Holter recording or left ventricular ejection fl'action. This large multicenter study of the signal-averaged ECG should clarify the role of late potentials in postinfarction risk stratification. Potentially, this study can improve the selection of patients for clinical studies or conventional treatment with therapies aimed at reduc-
ing arrhythmic death and symptomatic nonfatal arrhythmic events.
Programmed electrical stimulation study One of the objectives of CAST is to detect proarrhythmic episodes and to establish their prevalence, contributing causes, and outcomes. The objective of this study is to perform programmed electrical stimulation in patients who develop sustained ventricular tachycardia or ventricular fibrillation while taking a CAST drug to determine the likelihood that the CAST drug contributed to or caused the sustained arrhythmia. Two studies with programmed electrical stimulation will be done, one on the CAST drug and one off of the CAST drug. If the patient has an inducible sustained ventricular arrhythmia on the CAST drug, but is not inducible off of the CAST drug, the sustained arrhythmia will be attributed to an adverse drug effect.
Population pharmacokinetic study Ancillary studies of the plasma concentration (fl'ee and protein bound), half-time for elimination, and metabolism of the CAST drugs are being conducted to determine whether the plasma concentration of parent compounds or metabolites are related to the degree of suppression of ventricular arrhythmias o1"an effect on arrhythmic end points dm'ing follow-up. A random sample of about 1000 patients will participate in this study. The results should be helpful to clinicians using these drugs to treat patients after myocardial infarction by clarifying the need for assessing plasma drug concentrations, the likelihood that a given dose will produce therapeutic plasma drug concentrations, and the dependence of plasma drug concentration on body weight, left ventricular ejection fi'action and renal or hepatic function.
Encainide and fleeainide increase mortality in C A S T
The CAST protocol permits drugs to be removed from or added to the trial. On April 17, 1989, the Data and Safety Monitoring Board held its regular semiannual meeting, reviewed mortality data complete as of March 30, 1989, and recommended that encainide and flecainide be removed from CAST because it had become virtually impossible to establish any beneficial effect fi'om these drugs if they were continued to the end of study and because of strong evidence that these drugs increased the death rate compared to placebo. The Data and Safety Monitoring Board also recommended that CAST be continued with moricizine, which still has a significant chance of showing benefit,
What Ca~ We Expect From Cast?
and that other antiarrhythmic drugs be reviewed as possible additions to CAST. The National Heart, Lung, and Blood Institute and the CAST Executive Committee reviewed the April 17, 1989 recommendation by the Data and Safety Monitoring Board and decided that same (lay to remove encainide and fiecainide from CAST immediately. On April 18, 1989, a message was sent to each CAST clinical center to discontinue treatment in all patients who had been assigned to encainide, fiecainide, or their corresponding placebos. This process was started on April 19 and was completed by April 24. Patients whose encainide or fiecainide was stopped had a 24-hour ECG recording done 7 to 30 clays after discontinuing their study drug. Those that still had sufficient ventricular arrhythmia frequency to qualify after washout of their original CAST medication, had a left ventricular ejection fraction
