The placebo
effect in heart failure
Many patients who are enrolled in controlled clinical trials of new drugs for the treatment of heart failure show favorable hemodynamic and clinical responses to placebo therapy. This “placebo effect” results from both the creation of a supportive therapeutic environment and the spontaneous improvement that is commonly seen when measurements of symptoms and cardiac function are repeated frequently over long intervals of time. Three months of treatment with a placebo produces a reduction in symptoms in 25% to 35% of patients, an increase in cardiac output and a decrease in pulmonary wedge pressure, and an increase in exercise tolerance of up to 90 to 120 seconds. Physicians commonly seek to maximize the “placebo effect,” since the goal of treatment in the clinical setting is to improve the quallty of the patient’s life. On the other hand, clinical investigators seek to minimize the “placebo effect,” since the goal of a research study is to test the hypothesis that the new drug is superior to a placebo. (AM HEART J 1990;120: 1579-82.)
Milton
Packer, MD New York, N.Y.
Any attempt to assess the efficacy of new cardiovascular drugs is complicated by the fact that patients may show favorable effects during a study in the absence of a change in active treatment. This phenomenon, known as the “placebo effect,” is well known to occur in various diseases. Patients with coronary artery disease commonly experience a reduction in the frequency of anginal attacks while they receive a placebo.1-3 Patients with systemic hypertension frequently show a sustained decrease in diastolic blood pressure while they receive no active drug.4y5 Patients with ventricular arrhythmias may show spontaneous variation in the frequency and complexity of ectopic activity that may mimic a beneficial drug response even if no new agent has been administered.6-8 Such spontaneous improvement may result from the natural variability of the disease process or from bias introduced by the design of a clinical trial. Because of the common occurrence of the “placebo effect,” placebo-controlled trials have emerged as the principal means of proving the usefulness and worthiness of new drugs in the United States. Interestingly, in contrast to other cardiovascular disorders, recognition of the importance of the “placebo effect” in patients with heart failure is a From School
the Division of Medicine.
of Cardiology,
Department
of Medicine,
Mount
Sinai
Supported by grants ROl-HL-25055 and K04-HL-01229 from the National Institutes of Health. Dr. Packer is the recipient of a Research Career Development Award from the National Heart, Lung and Blood Institute, Bethesda, Md. Reprint Medical 410123942
requests: Center,
Milton Packer, MD, Division One Gustave Levy Pl., New
of Cardiology, Mount York, NY 10029.
Sinai
relatively recent development. When vasodilators were first tested in the 1970s and early 19808, many investigators insisted that placebo-controlled trials were both unnecessary and unethical. The course of heart failure was thought to be so predictable that spontaneous variability was believed to be unlikely. Similarly, the course of heart failure was thought to be so unfavorable that any study that withheld a new promising agent was thought to be unethical. As a result, investigators published numerous uncontrolled observations, which (almost invariably) concluded that the new drug being tested represented a significant therapeutic advance. Such observations assumed that patients with heart failure did not show a “placebo effect.” The purpose of the present review is to establish that the “placebo effect” is a common occurrence in clinical trials of patients with heart failure in a fashion similar to that reported in other cardiovascular diseases. Many of the issues raised in this article have been underscored in similar reviews published during the past 5 years.gv lo THE PLACEBO
EFFECT
ON SYMPTOMS
The most direct approach to the evaluation of patients with heart failure is to inquire about symptoms of dyspnea and fatigue at rest or on exertion. These symptoms are then classified according to a categoric scale, such as that of the New York Heart Association, or a similar “symptom score,” and the efficacy of a new drug is judged using each patient as his or her own historical control. However, many patients seeing a physician or receiving a drug for the first time improve in part because they have entered a new 1579
1580
Packer
therapeutic environment. In this new setting the physician generally reduces any anxiety concerning symptoms, increases the expectation of future benefit, and reinforces compliance with recommendations concerning treatment that may have been previously made but ignored. The creation of such a therapeutic environment occurs commonly in clinical practice, but it is greatly exaggerated in the setting of a formal research trial, in which the personal attention given to the patient reaches extreme levels. All symptoms mentioned by the patient (even those that have received little notice for years) become the focus of scrutiny and are recorded in detail. All frustrations that the patient might experience (which might reduce his enthusiasm for participation) are minimized. Careful instruction about the nature of the disease and the importance of taking prescribed medications may now be provided for the first time. Patients become so pleased by the supportive environment created by the research team that they commonly ask whether they are eligible for a second study after the first trial is completed. Such attention may explain why 25 % to 35% of patients who enter trials of new drugs for the treatment of heart failure show significant clinical benefits from placebo therapy. 11-13The frequency of this occurrence does not appear to depend on the number of patients in the study and has been seen in trials of patients with mild or severe symptoms. THEPLACEBOEFFECTONLEFTVENTRICULAR FUNCTION
Many physicians assess the efficacy of a new drug for heart failure by measuring its effects on left ventricular performance. This assessment can be performed noninvasively (by either echocardiography or radionuclide ventriculography) or invasively (with the use of a pulmonary artery balloon-floatation catheter). Noninvasive measurements. Although mean values for left ventricular ejection fraction (measured either by echocardiography or radionuclide ventriculography) generally show little change during treatment with a placebo, individual patients may show a marked improvement in systolic function in the absence of a change in their medical regimen. Physicians have generally assumed that two consecutive measurements of the left ventricular ejection fraction in the same patient will generally agree within 0.05; this criterion, however, has been validated only in patients with normal left ventricular function, in whom the two tests have generally been performed within several hours or days. Repeated testing is much less reproducible when carried out in patients
American
December 1990 Marl Journal
with poor ventricular function, especially when the interval between the two tests is as long as 3 to 6 months. In our experience, if an 0.05 increase in ejection fraction is used to indicate a beneficial response to a new drug, 20% to 30% of patients show a significant improvement with placebo. Even if the ejection fraction were resistant to a placebo effect, it is not clear how changes in this variable should be interpreted. Changes in noninvasive measures of left ventricular performance have not been shown to correlate closely with observed changes in clinical status or functional capacity of patients with heart failure.14s l5 Indeed, most vasodilator and inotropic drugs can produce marked hemodynamic and clinical benefits without a notable change in the left ventricular ejection fraction.g Conversely, the left ventricular ejection fraction may increase significantly in patients with heart failure whose clinical state has deteriorated during treatment.16 Invasive hemodynamic measurements. Another commonly used approach to evaluate the efficacy of a new drug for heart failure is to measure its ability to correct the hemodynamic abnormalities seen in patients with the disease. However, spontaneous fluctuations in hemodynamic variables are frequently seen in the absence of drug therapy and can be interpreted as having statistical (and inevitably, biologic) significance, if postdrug effects are not assessed at fixed times and if threshold values (that exceed the level of spontaneous variability) are not used. If the investigator defined a “peak effect” independently for each hemodynamic variable regardless of the time of occurrence, he or she creates a self-fulfilling prophecy that a true drug effect has occurred.17 Even if uniform times and threshold values are used, hemodynamic changes mimicking a beneficial drug response may be seen in the absence of effective therapy, if measurements are performed immediately after catheterization of the right side of the heart or after the ingestion of a normal meallo> l8 In particular, intravascular instrumentation appears to elicit notable systemic vasoconstriction (presumably related to anxiety), which dissipates in 12 to 24 hours. If predrug measurements are performed during this immediate postcatheterization period, any subsequent measurements performed days, weeks, or months later can be interpreted as showing beneficial effects, if these long-term measurements are compared with the initial artifactually vasoconstricted state; similar degrees of systemic vasoconstriction may not be provoked during a second invasive study.lg Unless hemodynamic measurements are delayed by 12 to 24 hours after catheterization, inves-
Volume
120
Number
6, Part
Placebo effect in heart failure
2
tigators may report the occurrence of sustained hemodynamic improvement in patients receiving ineffective treatment. THE PLACEBO
EFFECT
ON EXERCISE
CAPACITY
The most common approach used in multicenter trials to evaluate the efficacy of a new drug for heart failure is to assess its ability to prolong exercise tolerance. However, the duration of exercise in patients with heart failure is highly dependent on the motivation of both the patient and the physician. Repeated testing predictably results in an improvement in exercise performance, in part because of increased familiarity of the patient with the testing procedure and the increased willingness of the physician to encourage the patient to exercise to exhaustion.i21 l3 The magnitude of the placebo effect during repeated exercise testing varies according to the criteria used for baseline reproducibility
before ran-
domization.20 If only one or two baseline exercise tests are performed, exercise duration during placebo therapy can increase by as much as 90 to 120 seconds,
but the magnitude of the placebo effect can be reduced to 10 to 30 seconds if 3 to 10 baseline tests are carried out. The magnitude of the placebo effect is also directly proportional to the number of investigators in the study. Even the use of gas exchange measurements during exercise cannot completely eliminate the placebo response.2L IMPLICATIONS
FOR THE ANALYSIS
OF STUDIES
Because all variables used to assess the efficacy of a new drug for heart failure are subject to a placebo effect, uncontrolled observations carried out to prove the efficacy of a new drug are difficult to interpret. Uncontrolled studies may be useful in elucidating the mechanism of a new drug or its effects on the physiologic abnormalities in heart failure, but they cannot be used to evaluate the potential of a new drug for the treatment of heart failure. Assuming that a placebo-controlled study is carried out, how should it be analyzed? Some investigators inappropriately analyze such studies by focusing only on the significance of the changes seen in the treatment group. If the p value for the treatment group is
effect in heart failure
Many patients who are enrolled in controlled clinical trials of new drugs for the treatment of heart failure show favorable hemodynamic and clinical responses to placebo therapy. This “placebo effect” results from both the creation of a supportive therapeutic environment and the spontaneous improvement that is commonly seen when measurements of symptoms and cardiac function are repeated frequently over long intervals of time. Three months of treatment with a placebo produces a reduction in symptoms in 25% to 35% of patients, an increase in cardiac output and a decrease in pulmonary wedge pressure, and an increase in exercise tolerance of up to 90 to 120 seconds. Physicians commonly seek to maximize the “placebo effect,” since the goal of treatment in the clinical setting is to improve the quallty of the patient’s life. On the other hand, clinical investigators seek to minimize the “placebo effect,” since the goal of a research study is to test the hypothesis that the new drug is superior to a placebo. (AM HEART J 1990;120: 1579-82.)
Milton
Packer, MD New York, N.Y.
Any attempt to assess the efficacy of new cardiovascular drugs is complicated by the fact that patients may show favorable effects during a study in the absence of a change in active treatment. This phenomenon, known as the “placebo effect,” is well known to occur in various diseases. Patients with coronary artery disease commonly experience a reduction in the frequency of anginal attacks while they receive a placebo.1-3 Patients with systemic hypertension frequently show a sustained decrease in diastolic blood pressure while they receive no active drug.4y5 Patients with ventricular arrhythmias may show spontaneous variation in the frequency and complexity of ectopic activity that may mimic a beneficial drug response even if no new agent has been administered.6-8 Such spontaneous improvement may result from the natural variability of the disease process or from bias introduced by the design of a clinical trial. Because of the common occurrence of the “placebo effect,” placebo-controlled trials have emerged as the principal means of proving the usefulness and worthiness of new drugs in the United States. Interestingly, in contrast to other cardiovascular disorders, recognition of the importance of the “placebo effect” in patients with heart failure is a From School
the Division of Medicine.
of Cardiology,
Department
of Medicine,
Mount
Sinai
Supported by grants ROl-HL-25055 and K04-HL-01229 from the National Institutes of Health. Dr. Packer is the recipient of a Research Career Development Award from the National Heart, Lung and Blood Institute, Bethesda, Md. Reprint Medical 410123942
requests: Center,
Milton Packer, MD, Division One Gustave Levy Pl., New
of Cardiology, Mount York, NY 10029.
Sinai
relatively recent development. When vasodilators were first tested in the 1970s and early 19808, many investigators insisted that placebo-controlled trials were both unnecessary and unethical. The course of heart failure was thought to be so predictable that spontaneous variability was believed to be unlikely. Similarly, the course of heart failure was thought to be so unfavorable that any study that withheld a new promising agent was thought to be unethical. As a result, investigators published numerous uncontrolled observations, which (almost invariably) concluded that the new drug being tested represented a significant therapeutic advance. Such observations assumed that patients with heart failure did not show a “placebo effect.” The purpose of the present review is to establish that the “placebo effect” is a common occurrence in clinical trials of patients with heart failure in a fashion similar to that reported in other cardiovascular diseases. Many of the issues raised in this article have been underscored in similar reviews published during the past 5 years.gv lo THE PLACEBO
EFFECT
ON SYMPTOMS
The most direct approach to the evaluation of patients with heart failure is to inquire about symptoms of dyspnea and fatigue at rest or on exertion. These symptoms are then classified according to a categoric scale, such as that of the New York Heart Association, or a similar “symptom score,” and the efficacy of a new drug is judged using each patient as his or her own historical control. However, many patients seeing a physician or receiving a drug for the first time improve in part because they have entered a new 1579
1580
Packer
therapeutic environment. In this new setting the physician generally reduces any anxiety concerning symptoms, increases the expectation of future benefit, and reinforces compliance with recommendations concerning treatment that may have been previously made but ignored. The creation of such a therapeutic environment occurs commonly in clinical practice, but it is greatly exaggerated in the setting of a formal research trial, in which the personal attention given to the patient reaches extreme levels. All symptoms mentioned by the patient (even those that have received little notice for years) become the focus of scrutiny and are recorded in detail. All frustrations that the patient might experience (which might reduce his enthusiasm for participation) are minimized. Careful instruction about the nature of the disease and the importance of taking prescribed medications may now be provided for the first time. Patients become so pleased by the supportive environment created by the research team that they commonly ask whether they are eligible for a second study after the first trial is completed. Such attention may explain why 25 % to 35% of patients who enter trials of new drugs for the treatment of heart failure show significant clinical benefits from placebo therapy. 11-13The frequency of this occurrence does not appear to depend on the number of patients in the study and has been seen in trials of patients with mild or severe symptoms. THEPLACEBOEFFECTONLEFTVENTRICULAR FUNCTION
Many physicians assess the efficacy of a new drug for heart failure by measuring its effects on left ventricular performance. This assessment can be performed noninvasively (by either echocardiography or radionuclide ventriculography) or invasively (with the use of a pulmonary artery balloon-floatation catheter). Noninvasive measurements. Although mean values for left ventricular ejection fraction (measured either by echocardiography or radionuclide ventriculography) generally show little change during treatment with a placebo, individual patients may show a marked improvement in systolic function in the absence of a change in their medical regimen. Physicians have generally assumed that two consecutive measurements of the left ventricular ejection fraction in the same patient will generally agree within 0.05; this criterion, however, has been validated only in patients with normal left ventricular function, in whom the two tests have generally been performed within several hours or days. Repeated testing is much less reproducible when carried out in patients
American
December 1990 Marl Journal
with poor ventricular function, especially when the interval between the two tests is as long as 3 to 6 months. In our experience, if an 0.05 increase in ejection fraction is used to indicate a beneficial response to a new drug, 20% to 30% of patients show a significant improvement with placebo. Even if the ejection fraction were resistant to a placebo effect, it is not clear how changes in this variable should be interpreted. Changes in noninvasive measures of left ventricular performance have not been shown to correlate closely with observed changes in clinical status or functional capacity of patients with heart failure.14s l5 Indeed, most vasodilator and inotropic drugs can produce marked hemodynamic and clinical benefits without a notable change in the left ventricular ejection fraction.g Conversely, the left ventricular ejection fraction may increase significantly in patients with heart failure whose clinical state has deteriorated during treatment.16 Invasive hemodynamic measurements. Another commonly used approach to evaluate the efficacy of a new drug for heart failure is to measure its ability to correct the hemodynamic abnormalities seen in patients with the disease. However, spontaneous fluctuations in hemodynamic variables are frequently seen in the absence of drug therapy and can be interpreted as having statistical (and inevitably, biologic) significance, if postdrug effects are not assessed at fixed times and if threshold values (that exceed the level of spontaneous variability) are not used. If the investigator defined a “peak effect” independently for each hemodynamic variable regardless of the time of occurrence, he or she creates a self-fulfilling prophecy that a true drug effect has occurred.17 Even if uniform times and threshold values are used, hemodynamic changes mimicking a beneficial drug response may be seen in the absence of effective therapy, if measurements are performed immediately after catheterization of the right side of the heart or after the ingestion of a normal meallo> l8 In particular, intravascular instrumentation appears to elicit notable systemic vasoconstriction (presumably related to anxiety), which dissipates in 12 to 24 hours. If predrug measurements are performed during this immediate postcatheterization period, any subsequent measurements performed days, weeks, or months later can be interpreted as showing beneficial effects, if these long-term measurements are compared with the initial artifactually vasoconstricted state; similar degrees of systemic vasoconstriction may not be provoked during a second invasive study.lg Unless hemodynamic measurements are delayed by 12 to 24 hours after catheterization, inves-
Volume
120
Number
6, Part
Placebo effect in heart failure
2
tigators may report the occurrence of sustained hemodynamic improvement in patients receiving ineffective treatment. THE PLACEBO
EFFECT
ON EXERCISE
CAPACITY
The most common approach used in multicenter trials to evaluate the efficacy of a new drug for heart failure is to assess its ability to prolong exercise tolerance. However, the duration of exercise in patients with heart failure is highly dependent on the motivation of both the patient and the physician. Repeated testing predictably results in an improvement in exercise performance, in part because of increased familiarity of the patient with the testing procedure and the increased willingness of the physician to encourage the patient to exercise to exhaustion.i21 l3 The magnitude of the placebo effect during repeated exercise testing varies according to the criteria used for baseline reproducibility
before ran-
domization.20 If only one or two baseline exercise tests are performed, exercise duration during placebo therapy can increase by as much as 90 to 120 seconds,
but the magnitude of the placebo effect can be reduced to 10 to 30 seconds if 3 to 10 baseline tests are carried out. The magnitude of the placebo effect is also directly proportional to the number of investigators in the study. Even the use of gas exchange measurements during exercise cannot completely eliminate the placebo response.2L IMPLICATIONS
FOR THE ANALYSIS
OF STUDIES
Because all variables used to assess the efficacy of a new drug for heart failure are subject to a placebo effect, uncontrolled observations carried out to prove the efficacy of a new drug are difficult to interpret. Uncontrolled studies may be useful in elucidating the mechanism of a new drug or its effects on the physiologic abnormalities in heart failure, but they cannot be used to evaluate the potential of a new drug for the treatment of heart failure. Assuming that a placebo-controlled study is carried out, how should it be analyzed? Some investigators inappropriately analyze such studies by focusing only on the significance of the changes seen in the treatment group. If the p value for the treatment group is
