Karen R. Muñana, DVM, MS, DACVIM (Neurology)
The placebo effect is a widely recognized, but controversial concept. There is considerable disagreement in the literature regarding placebo terminology, and various definitions have been proposed based on differing perspectives on the subject. For the purposes of this presentation, a definition adapted from Shapiro and Shapiro1 will be utilized, in which the term placebo describes any treatment that is used for its ameliorative effect on a symptom or disease but that actually is ineffective or is not specifically effective for the condition being treated. Correspondingly, the placebo effect can be defined as the nonspecific psychological or psychophysiological therapeutic outcome associated with placebo administration. The effect is not due to any pharmacologic properties of the treatment being administered, and may be the effect of spontaneous change over time that is falsely attributed to the placebo.
Extensive research has focused on gaining a better understanding of the mechanism of the placebo effect, fueled by the general public's interest in benign forms of treatment and the scientific community's quest to unravel the complex response that is associated with therapeutic interventions in human medicine. Historically, the placebo effect has played a role in the clinical practice of medicine, wherein placebo administration has been incorporated into specific therapeutic regimens with the goal of maximizing benefits. More recently, the primary significance of the placebo effect has been in the interpretation of clinical trials, where the placebo serves as a control by which the experimental intervention is evaluated. Despite the numerous publications on the subject, a widely accepted explanation for the placebo effect remains elusive. While some researchers believe that the placebo effect is a true, powerful phenomenon to be considered in medical decision making, others regard it as insignificant, arguing that most of its purported effect is due to misinterpretation of statistical analyses.
In contrast to the controversy present in the human medical field, the placebo effect has been largely ignored in veterinary medicine. To the author's knowledge, there are only two veterinary publications, one case report2 and one original research3, that address the issue of the placebo effect in animals. However, with the recent emphasis on evidence based medicine in veterinary practice, it seems appropriate to consider the effect of placebos in our veterinary patients, particularly the extent to which animals may demonstrate an improvement in disease manifestations that may be due to nonspecific effects of a therapeutic intervention. This presentation will provide historical background on the placebo effect, and discuss the placebo effect in human epilepsy trials. Finally, data will be presented to demonstrate a role for the placebo effect in canine epilepsy trials.
The History of Placebo in Medical Practice
The word placebo, Latin for "I shall please", was first included in a medical dictionary in 1785, and defined as a "commonplace method or medication calculated to amuse for a time, rather than for any other purpose".1 This date coincides with the first intentional use of inert substances for clinical purposes. The meaning changed slightly over time, as illustrated in a medical dictionary entry from 1811, which defined placebo as " an epithet given to any medicine adopted to please rather than to benefit the patient".1 It is of interest to note that until the 1950's, a placebo was considered a type of medication.
The use of placebos changed dramatically in the 1950's with the marked increase in randomized controlled clinical studies. Placebos began to be utilized for their inert properties, and provided a control for spontaneous change and psychological variables, thereby permitting differentiation between active and inactive treatments. A pioneering study on the importance of the placebo effect in clinical practice was published in 1955. In this article, entitled "The Powerful Placebo", Beecher analyzed the results from 15 controlled clinical trials on a variety of diseases in order to quantify the placebo effect.4 He documented a therapeutic benefit associated with the administration of placebo, but also reported the magnitude of adverse effects described in patients receiving inert substances as treatment for a medical condition. The publication of these findings led to a widespread acceptance among the medical community of the value of randomized clinical trials in evaluating novel therapies. With this change, the placebo's role evolved from an active therapeutic to its present status as more of a variable to be controlled in experimental studies.
Factors Contributing to the Placebo Effect
The standard of care in human medicine dictates that novel therapeutics undergo safety and efficacy evaluation through randomized clinical trials prior to gaining approval and acceptance of their use. This process has produced a multitude of clinical trials evaluating various forms of therapy in all disciplines of medical practice. Meta-analyses of groups of studies have been performed to evaluate the significance of the placebo effect in different disorders. A beneficial effect of placebo administration has been documented in 60-90% of all human diseases, including hypertension, angina pectoris, congestive heart failure, acne, asthma, gastrointestinal ulcers, arthritis, headache, Parkinson's disease and depression. It is generally accepted that approximately one-third of patients will show improvement in their condition after the administration of a placebo. However, placebo response rates tend to vary considerably between different medical conditions, and even higher rates have been reported with certain diseases. Both subjective and objective improvements have been documented after placebo administration, yet uncertainty exists as to the cause of this effect.
In fact, much of the controversy surrounding placebo administration has centered on the proposed mechanisms for its effect. Efforts have been made to differentiate the "true" placebo effects from the "perceived" placebo effects, and determine the relative contribution of each to the overall response observed. The "true" placebo effect is a change that is caused by placebo administration, or nonspecific effects associated with the therapeutic intervention. The "perceived" placebo effect is a change that is observed after placebo administration, but that is not directly caused by the placebo.5 Some investigators do not believe that the latter should be considered a component of the placebo effect, since these are time and situational effects rather than being a direct result of placebo administration. However, since both of these effects factor into the response seen after a therapeutic agent is administered, they will be considered components of the placebo effect for the purposes of this discussion.
The two main theories cited to explain the "true" placebo effect are classical (unconscious) conditioning and conscious expectation.5 The conditioning theory equates the placebo effect to a Pavlovian response, in which an initially neutral stimulus (the placebo) is presented repeatedly with an unconditioned stimulus that elicits a known specific response. The pairing of the unconditioned stimulus with the placebo eventually leads to the placebo alone eliciting the same response as the unconditioned stimulus. In this way, the placebo effect is a conditioned response that is generated in the therapeutic setting; patients learn to experience improvement after medical interventions because they have benefited from consulting a doctor in the past. The development of this type of conditioned response has been documented in studies on both humans and animals.
The expectancy theory states that the expectations of the patient are raised through the act of administering a medical treatment, which in turn, brings about symptomatic improvements during the treatment period. This model assumes that the effected change was causally associated with the patient's specific expectation of improvement. This effect can be modified through information provided to the patient (such as comments regarding the potency of the medication being administered) as well as the degree of patient motivation. Furthermore, as expectancies are acquired in part through conditioning, extensive overlap exists between the two theories. The neurobiological mechanisms behind the "true" placebo effect appear to be mediated in part by endogenous opioids. In addition, imaging studies have demonstrated that placebos may mimic the effects of functional drug by activating similar areas and circuitry of the brain.6 These mechanisms are plausible explanations for both the conditioning and the expectancy theories.
The "perceived" placebo effects include instances in which the temporal association of placebo administration is associated with some change in the condition, although no causal relationship can be established between the two. Factors that contribute to the "perceived" placebo effect include the natural history of disease, regression to the mean, unidentified parallel interventions and investigator or participant bias.5 The natural history of the disease must be considered when evaluating any therapeutic response. It is well accepted that a disease can either worsen or improve over time, and if a disease naturally improves during placebo administration, this effect may be erroneously attributed to the placebo. Similarly, the possibility of regression to the mean should be considered, which is a statistical term used to describe the fluctuations of biological variables that occur over time and that take the form of a sine wave around a mean. Patients are most likely to seek medical care then their symptoms are at a peak, and this is often when a new intervention is initiated. Over time, regression to the mean will likely result in improvement in the patient's condition, regardless of what treatment is administered. Parallel interventions may involve treatments that are not disclosed to the clinician, or changes in the patient's environment that may influence the outcome being evaluated. Finally, bias on the part of the investigator or the participant can also contribute to the placebo effect. Even in blinded studies, expectations of a response may influence subjective interpretation of experimental results by the investigator that could erroneously attribute a response to placebo administration. The potential for participant bias is well recognized and explained by the Hawthorne Effect, which states that simply by participating in a test, trial or study the participants have a better experience because of the focus of interest toward them which is rewarding for its own stake. Consequently, participants document better results regardless of the change provided or the treatment experienced.
It has been argued that the inclusion of an untreated group in clinical trials is necessary in order to differentiate the "truc" placebo effects from the "perceived" placebo effects. However, the majority of the medical profession does not consider this a feasible approach due to ethical concerns of withholding treatment from a patient, and therefore separating these two facets of the placebo effect remains difficult.
The Placebo Effect in Human Epilepsy Studies
Nine new antiepileptic drugs have been approved for use in human epilepsy over the last 15 years, and all these drugs have undergone vigorous preclinical and post-approval evaluation through randomized clinical trials. The design of trials evaluating a novel antiepileptic medication as an "add-on" to therapy has been relatively uniform.7 Typically, patients are eligible if they have been diagnosed with medically refractory partial-onset seizures, in that they experience a minimum number of seizures each month while maintaining a constant dose of one to three antiepileptic drugs. Following a prospective baseline phase of approximately 8 weeks, patients are randomized in a double-blind fashion to add-on treatment with the study drug or to add-on placebo in addition to constant doses of their baseline antiepileptic drugs. The double-blind period typically lasts 12-16 weeks. The primary efficacy variable is the change in seizure frequency during the double-blind period compared to the baseline between the study drug and the placebo. From these studies, considerable data is available on the effect of placebo administration in human epileptics.
A recent study aimed to evaluate the magnitude of the placebo effect across human epilepsy trials.8 The analysis documented a response to placebo in individual trials ranging from 0-36%. Data obtained from meta-analyses performed on specific antiepileptic drugs was also evaluated, and demonstrated that the placebo response varied across drug trials, with 9.3%-16.6% of patients exhibiting a 50% or greater reduction in seizure frequency after receiving placebo. Similar results had been obtained in a previous study that compared data obtained from clinical trials on five of the new antiepileptic drugs, in which placebo success rates ranged from 4%-18%, and averaged 10% over all studies.9
Much of the placebo effect seen in human epilepsy trials has been attributed to regression to the mean. Patients are more likely to enroll in a clinical trial at a time when their seizures are under relatively poor control. If during the course of the trial, the seizure frequency in those patients reverts back to its historical average, an improvement in seizure frequency will be noted that is unrelated to a treatment effect. Other possible explanations include improved and more accurate monitoring of clinical events during the clinical trial, improved adherence to treatment with active medication, investigator bias and the Hawthorne effect.
Investigators have attempted to identify baseline characteristics of placebo responders versus nonresponders in studies involving levetiracetam as add-on therapy, in an effort to account for the variability in placebo response rates seen across seizure trials.10 Placebo responders are more likely than nonresponders to be taking only one antiepileptic drug during baseline, as well as have a later age at onset and a shorter duration of epilepsy. Based on these findings, it was concluded that less severely ill patients are more prone to respond to placebo, a finding that has also been demonstrated for other central nervous system disorders.
Although the mechanism behind the response to placebo is still not completely understood, the data on the magnitude of the placebo effect has been used to support the premise that randomized, placebo controlled clinical trials are essential in establishing the efficacy of new antiepileptic medication as adjunctive therapy in patients with medically refractory epilepsy. If a placebo control group is not used, it is impossible to differentiate the true treatment effect from other nonspecific effects that may be occurring concurrently. These nonspecific effects are what are quantified with the inclusion of a placebo group in a study design.
Findings From Canine Epilepsy Studies
There has been considerable interest in the use of the newly approved human antiepileptic drugs for dogs, with the goal of improving seizure control and decreasing side effects in animals that have an unsatisfactory response to conventional drugs. Results of the use of gabapentin,11,12 zonisamide,13,14 and levetiracetam15 for refractory canine epilepsy have been published, and these studies comprise the primary information that is available to clinicians who are attempting to judge the efficacy of such drugs. To date, all of the published studies have been performed in an open label fashion, with baseline data obtained retrospectively and data during the treatment period collected prospectively. No placebo control groups have been utilized. Treatment response rates, defined as a 50% reduction in seizures from baseline, have been reported to range from 41-80%, leading many to conclude that these drugs are efficacious for medically refractory epilepsy in dogs.
However, based on the magnitude of the placebo effect reported in human epilepsy trials, it is likely that the efficacy results reported in these canine studies are exaggerated. Many of the mechanisms that contribute to the placebo effect can also explain the beneficial treatment results with an open label study design. For example, it is probable that regression to the mean accounts for some of the observed treatment effect. Owners tend to seek veterinary care for their epileptic dog when seizure frequency increases, and this is likely the time that a new treatment will be initiated. Over time, seizure frequency should tend to revert back to its historical average, and this fluctuation may be erroneously attributed to the novel treatment administered. Other factors that may play a role in the interpretation of study results are those previously discussed for human epilepsy trials, including investigator and participant bias and improved compliance with medication administration.
The author has undertaken three randomized, placebo-controlled studies evaluating add-on therapy for dogs with medically refractory therapy. Each study involved a prospective period of baseline data collection ranging from 2-3 months, followed by randomization and administration of the study treatment and placebo. Two studies were cross over in design, such that each dog received both placebo and the study treatment, while the third study was parallel in design, with enrolled dogs received either the study treatment or placebo. In each instance the change in seizure frequency between the treatment period and baseline was compared for the study treatment and placebo. Data from each of these studies was statistically evaluated to determine the magnitude of the placebo effect. Results will be presented. The results suggest that the placebo effect is of significance in canine epilepsy trials, and demonstrate the importance of performing randomized clinical trials to establish an accurate assessment of efficacy of new treatments.
The placebo effect is considered a noteworthy phenomenon in human medicine, and indeed is a source of controversy regarding mechanisms, differentiation of true versus perceived effects of placebo administration, and relative significance in clinical practice. However, there is consensus within the human medical profession that placebo controlled studies are necessary to determine the efficacy of novel therapies. Randomized clinical studies have been the standard in human medicine for the past 50 years, yet are only now beginning to gain momentum in veterinary medicine.
The placebo effect has been shown to be of considerable magnitude in human epilepsy trials, and data reported here suggests that the same is true for canine epilepsy trials. This effect makes it difficult to make determinations of treatment efficacy from studies that are not controlled. Although such studies do provide useful information, particularly with regard to the safety of the treatment and the epidemiology of the disorder, the practitioner should be aware of their shortcomings. More controlled studies are needed in veterinary medicine in order to provide the scientific basis for treatment recommendations in the management of canine epilepsy.
1. Shapiro AK, et al. The Powerful Placebo. The Johns Hopkins University Press, Baltimore, 1997.
2. McMillan FD. J Am Vet Med Assoc 1999; 215(7):992.
3. Jaeger GT, et al. Acta Vet Scand 2005; 46(1-2):57.
4. Beecher H. J Am Med Assoc 1955; 159(17): 1602.
5. Ernst E. Drug Discov Today 2007;12(9-10):413.
6. Klosterhalfen S, et al. Autonom Neurosci 2006; 125(1-2):94.
7. Beydoun A, et al. Epilepsy Behav 2003; 4(1 ):4.
8. Burneo JG, et al. Epilepsy Behav 2002: 3(6):532.
9. Cramer JA, et al. Epilepsia 1999: 40(5):590.
10. Niklson I, et al. Epileptic Disord 2006; 8(1):37.
11. Platt SR, et al. Vet Rec 2006; 159(26):881.
12. Govendir M, et al. Aus Vet J 2005; 83(10):602.
13. Dewey CW, et al. J Am Anim Hosp Assoc 2004; 40(4):285.
14. von Klopmann T, et al. J Sm Anim Pract 2007; 48(3):134.
15. Volk HA, et al. Vet J 2007 (Epub).