Abstract
We analysed the adverse events of placebo in acute and preventive randomized, double-blind, placebo-controlled studies for migraine treatment. Fifty-seven trials (oral triptans, non-steroidal anti-inflammatory drugs, nasal ergot alkaloids and preventive agents) were included. From 10 to 30% of subjects reported adverse events after placebo. Most common were features associated with a migraine attack, such as nausea, phono- and photophobia. Other frequent complaints resembled those of the active drug (e.g. chest pressure in triptan trials). A third group of adverse events appeared to be coincidental (e.g. sleep disturbance). Adverse events following placebo are probably related to the drug under study and the symptomatology of migraine; some have no obvious explanation.
Introduction
Placebos have been reported to be effective, thereby reducing outcomes in clinical trials by 30–40% (1). A recently published critical analysis of clinical trials confirmed a placebo effect in pain studies and in trials with continuous subjective outcomes, such as insomnia or anxiety (2). The mechanisms of the placebo response include expectations, beliefs and conditioning (3). An anatomic correlate exists for placebo analgesia. The rostral anterior cingulate cortex and the lower pons, areas rich in opioid receptors, show increased activity following opioid or placebo application when pain is reduced (4). This suggests that the same brain circuits that are involved in response to opioids are involved in the response to placebo. These findings support the hypothesis generated in 1978 that endogenous opioids mediate the pain placebo responses (5).
Unlike the effects of placebo on efficacy, little is known about its adverse events (AEs) in migraine trials (6). In general, AEs are mild and transient. Some AEs, such as nausea, vomiting and headache, are symptoms of the migraine attack itself, whereas others may be independent of the migraine attack and coincidental. Some, such as hypotension, are more common in migraineurs. Placebo AEs may also resemble the AEs of the active drug given for treatment. For example, in triptan trials chest pain and finger tingling, which are AEs of triptans, are also reported following placebo intake (7). Headaches have also been documented as AEs of placebo in non-headache trials. In a clinical study to determine the efficacy of lamotrigine in painful neuropathy, headache occurred as an adverse event in 10% of patients after placebo (8). These headaches are neither classified according to IHS criteria (9) nor reported as the patient's usual condition before entering the study. The prevalence of headache in the population is much higher than 10% (10); therefore it is difficult to come to any conclusion.
To our knowledge, no effort has been undertaken so far to systematically analyse the AEs of placebos in acute and preventive migraine trials. We performed an extensive literature search to ascertain the nature of AEs after placebo in randomized, double-blind, placebo-controlled acute and preventive migraine trials. We discuss factors that may contribute to headache as an AE of placebo.
Methods
We performed a Medline survey and identified randomized, placebo-controlled, double-blind studies for acute and preventive migraine treatment between 1989 and 2002. Trials were included only when migraine was diagnosed according to the 1988 IHS guidelines (9). The clinical trials were classified in two study groups: (I) acute migraine therapy with (a) triptans, (b) non-steroidal anti-inflammatory drugs (NSAIDs) and (c) ergot alkaloids; (II) preventive migraine therapy. Except for nasal formula ergot alkaloids, only studies with an oral drug formulation were included. Patients with active and placebo treatment were identified in each trial, as were the type and number of AEs following placebo treatment. We then averaged the percentage of each AE for those studies in which this particular AE was reported. Results are reported as mean ± standard deviation (SD).
Results
Fifty-seven studies were analysed with a total of 5906 patients in the placebo groups (11–67): triptans (3618 placebo/12496 active drug); NSAIDs (1656 placebo/3966 active drug); ergot alkaloids (258 placebo/677 active drug) and preventive agents (374 placebo/790 active drug). The triptan group had 31 studies: sumatriptan (
In none of the studies was a drug related serious AE reported after placebo. When AEs were considered to be drug related this was assumed so by the investigator of the study. In general AEs were mild after placebo.
Adverse events of placebos in trials with Triptans
In all analysed triptan trials, between 20 and 30% (21 ± 9%) of patients reported at least one AE, 13 ± 4% were believed to be drug related. The most common AEs were phono-/photophobia in 11 ± 6% followed by nausea/vomiting on average in 13 ± 11% of subjects. In one trial,>50% of placebo-treated patients complained about nausea/vomiting as an AE. Other less frequently mentioned AEs were headache/migraine, dry mouth, tingling, neck stiffness, dizziness and somnolence (Figure Fig. 1a, b). Skeletal muscle pain and chest pain were rarely reported (<2%). On average, 5% of patients mentioned other unspecified AEs.
Side-effects of placebo in trials for the treatment of migraine attacks. (a) and (b) show side-effects in studies with oral triptans, (c) illustrates AEs in studies with oral non-steroidal anti-inflammatory drugs and (d) summarizes the frequency of AEs in studies with a nasal formula of ergot alkaloids. The results are expressed as percentage ±SD of subjects experiencing one specific AEs and illustrated as box plots. N indicates the number of studies in which the side-effect was assessed. ∗indicates studies with results outside the 95% confidence interval.
Adverse events of placebos in trials with non-steroidal anti-inflammatory drugs
On average 10 ± 5% of patients reported one or more AEs following placebo; they were drug related in 4 ± 2.5%. Nausea and vomiting, dizziness and fatigue were most common, each occurred in about 3% of patients. Headache as a side-effect was mentioned only in one study by 1% of subjects. 3% of patients reported not further characterized AEs after placebo (Fig. 1c).
Adverse events of placebos in trials with ergot alkaloids (dihydroergotamin nasal spray)
Our results for this substance class are hampered by the limited number of placebo-controlled randomized studies after 1988. The most common AEs in all trials were nose/throat symptoms most likely due to the nasal application mode which were reported by 7 ± 5% of subject (Fig. 1d). All other mentioned AEs were only reported in one trial and include dizziness (3%), nausea (7%) and vomiting (1%).
Adverse events of placebos in trials with preventive drugs
In these preventive trials 21 ± 9% of patients reported at least 1 adverse event after placebo. Nausea (11.5 ± 3.5%), paraesthesia (13 ± 8%) and fatigue (8 ± 4%) were the most common complaints. Headache was not mentioned as an adverse event after placebo in any of the trials. In one study with topiramate weight loss was mentioned by 29% of subjects in the placebo group (65). Non-specified AEs were reported by 8% of patients (Fig. 2).
Side-effects of placebo in studies for the prevention of migraine. The results are expressed as percentage ±SD of subjects experiencing one specific adverse event and illustrated as box plots. N indicates the number of studies in which the side-effect was assessed. ∗indicates studies with results outside the 95% confidence interval.
Discussion
Our analysis shows that AEs following placebo in migraine trials are in part dependent on the class of active drug being investigated. The percentage of subjects reporting at least one AE after placebo ranged from 20 to 30% in trials with triptans and preventives to 10% in trials with NSAIDs. Many of these AEs were classified as drug related by the investigators while the study was still blinded. The nature of the placebo encoated sugar or starch tablet is unlikely to account for the AEs reported. Other factors contribute to the placebo response.
In acute migraine trials nausea, vomiting, photophobia and phonophobia were the most frequently reported AEs attributable to placebo. These AEs are part of the migraine attack (68), therefore it is not surprising that they are attributed to a pharmacologically inactive substance. Migraine associated symptoms can occur or recur at any time during the attack even coincidentally after placebo. They are often assumed to be an adverse event. It is surprising that the occurrence of these migraine features is low after placebo treatment. This may be related to a low likelihood of new or worsening symptoms occurring after placebo.
Patient expectation influences the response to a given drug. The adverse event profile of the active substance under study probably influences the adverse event reported after placebo. In one trial with topiramate (65), a drug which causes weight loss, about 29% of patients in the placebo group also report weight loss. Similar results have been observed in non-headache trials (e.g. cardiology) in which tachycardia was a side-effect of the active drug and also a typical complaint after placebo (69). The consent form or the doctor's verbal instructions on the active drug expected AEs and the doctor's expectations affect the type and amount of AEs (70–72). This could account for the occurrence of chest pressure or muscle pain after placebo in triptan trials or asthenia in trials with beta blockers. The literature concerning whether and how patients can be influenced by written or verbalized information about potential side-effects of the active drug is contradictory. In one study verbalized (by the doctor) and written information was given to one study group about the three most common side-effects of the substances investigated (antibiotics, angiotensin converting enzymes or non-steroidal anti-inflammatory drugs) whereas no specific information was provided to the other study group. After a 3-week observation period there was no difference in the incidence of the targeted side-effects indicating no influence due to knowledge of potential AEs (73). In contrast, another investigation indicates that the inclusion of potential gastroinstestinal side-effects of non-steroidal anti-inflammatory drugs in the consent form led to a six fold increase in the number of subjects withdrawing from the study due to subjective minor gastrointestinal symptoms. Major side-effects such as ulcers or bleeding were not different between treatment groups (74). The latter result indicates that disclosure of information on potential AEs can influence their occurrence. A similar mechanism could also contribute to an increased incidence of AEs in the placebo and active treatment group in placebo-controlled, double-blind studies. Indeed, there is evidence in the literature that depending on the character of the patient groups adverse events can even occur more often in the placebo than in the treatment group. For example, in patients with anxiety the incidence of dry mouth as an AE was higher in the placebo group than in the active treatment group (69). This finding is puzzling to us. We speculate that the knowledge of the character of potential AEs of the active drug and the patients expectation may increase the likelihood of the occurrence of AEs. This phenomenon must be taken into account in particular when AEs of placebos are analysed. A remedy for this problem is not easy to find since it is unethical from our point of view to include someone in a study without knowledge of potential AEs.
The drug formulation and mode of administration may contribute to the placebo's AEs. In trials with nasal ergot alkaloids nose/throat symptoms or application side AEs were common. These AEs are probably due to the application mode of the nasal spray. Some patients may simply not like nasal sprays or the application of drugs to the nasal and oral mucosa. The route of administration also affects the placebo response; subcutaneous placebos have stronger effects than oral placebos potentially due to enhanced release of endogenous opioids by a ‘stronger’ treatment (75). The nasal application of a drug could create the same response resulting in a stronger treatment effect and potentially also worsen the adverse event profile.
Some AEs after placebo may be nothing more than the coexistence of common disorders in the general population. The prevalence of sleep disturbances in otherwise healthy subjects in western societies is about 35% (76). Thus sleep disturbances often occur but are independent of drug or placebo. A similar phenomenon may account for the occurrence of headache as an AE in non-headache trials. Lifetime prevalence of unspecified headaches is between 93% in males and 99% in females in western society and the prevalence of tension type headache in the same study for the previous year was 63% in males and 86% in females (77). In contrast, severe unspecified headaches for the previous year were reported by 21% of subjects in a Swedish study (78). Data for the one month prevalence of headaches are rarely described, but, e.g. the one month migraine prevalence is 3.0% in young men and 7.4% in young females in the US (79). Thus headache is an extremely frequent complaint that may be recorded by chance during a clinical study as an AE of placebo.
Headache could also be a response to emotional, physical and other stress factors (80). Participation in a clinical trial and the intake of a unknown drug might be a stress factor which can result in headache. Could the development of headache after placebo be a conditioned response due to repeated association between a conditioned response and an unconditioned stimulus (81)? Although this conditioning may rather account for the efficacy of placebos it is also possible that this phenomenon contributes to the development of AEs after placebo.
In rare cases allergic symptoms (e.g. rash or urticaria) to colouring agents in drugs have been reported, mostly in subjects with a history of allergies and tartrazine is the most frequently accused agent (82, 83). Case reports describing allergies due to ingredients such as sunset yellow, indigo carmine or amarath have also been published but systematic studies do not exist (84). We cannot exclude that in some individuals allergic events occur after placebo in headache trials due to colouring. However, these events tend to be exceptions and should occur similarly after placebo and the active drug in clinical trials and may not significantly contribute to the data we have summarized.
Results are hampered by a different reporting style of AEs among the studies we analysed. Some authors list AEs that were reported by more than 2% of patients in any group; other authors AEs that occurred in>5% of patients. There is huge difference in the listing or classification of AEs in the published reports. Some contain detailed data of each AEs, others announce the AEs overall incidence. Our analysis is also affected by the assessment of AEs. Some studies use a structured reporting sheet, others use free reporting such that the patient has to write the adverse event down. Many studies do not contain any information how AEs are assessed. Because of these inconsistencies it is impossible to analyse how different assessment techniques influence the incidence of AEs in this report. These differences make it also difficult to compare AEs of placebos as well as those of the active drug investigated among the different acute and preventive migraine drug trials. The methods of pain assessment influence the outcome in pain trials and it could be possible that these differences also affect outcome of AEs (3). Therefore, we propose to establish standardized evaluation forms (questionnaire) for the assessment of AEs in clinical headache trials. These standardized forms should give the patient the opportunity, e.g. by using a ranking scale to indicate whether a symptom was perceived as threatening for the subject (important) or rather trivial. The patient should also be asked whether the AE was due to psychological reasons, e.g. a particular reaction that always occurs in this person with stress or excitement like in a clinical trial. This could possibly allow differentiation between a pharmacological and a non-pharmacological source of side-effects in clinical headache trials but it will not allow differentiation between drug related and non-drug related AEs.
It is difficult to objectively determine whether an adverse event after placebo is drug related considering the biological effects of the placebo in the brain (4, 5);. Indeed, activation of the endogenous opioid system could cause a variety of symptoms, which may not necessarily be adverse events but rather a physiological response of the body to a certain stimulus.
The investigator's classification of AEs as drug related could be biased by personal experience and the preexisting knowledge of the AEs of the drug studied. Some symptoms may fit in the investigators scheme of the drug and may falsely be recorded as drug related. Therefore, we feel no benefit for the quality of the study by a rating of the blinded investigator whether an AE is drug related or not. Instead, it is helpful to determine whether the investigator (physician) and the patient consider an adverse event important or not.
Based on the nature of the placebo, the endogenous effects of the placebo and other factors which all contribute to the effects and AEs after placebo intake we hypothesize that ‘real’ drug related AEs due to placebo do not exist. Therefore, this category should be omitted in headache trials (and other trials with subjective outcome measures) in future. We feel that a rating serious vs. harmless AE by the patient and the physician is more valuable for the study. Patients may be scared by AEs and this could prevent a second use of the drug in future. Whether an AE (non-serious AE) is drug related or not may be of lesser importance for the patient.
AEs after placebo in migraine trials are usually mild. Often they resemble the migraine attack or the adverse event profile of the active drug and in some instances their occurrence appears to be related to a coexistent disorder. It is probably not due to the effects of the ingredients of the placebo tablet. Our analysis should help us to define more precise criteria as to what type of AEs should really be reported in clinical trials in order to be more meaningful.