Abstract
The best way to appreciate the efficacy of drug and behavioural therapy in the acute and prophylactic treatment of headache is to perform placebo-controlled randomized trials. In order to plan and conduct these studies in the most appropriate way, it is desirable to know which factors influence the placebo response. This paper reviews factors which influence the placebo response in clinical trials, such as expectation, blinding, route of application of drugs and age, gender and geographical distribution. Response rates of placebo in the treatment of acute headache episodes are higher than in headache prophylaxis. Invasive procedures such as injections have a higher placebo response compared with oral drugs. Variables known to influence the placebo response have to be taken into consideration to calculate properly the power of planned randomized trials.
Introduction
The perception of pain is a highly subjective experience that is influenced by cognitive factors such as expectation, attention, anxiety and previous experiences. Placebo analgesia is one of the most striking examples of the cognitive modulation of pain perception (1, 2). It represents a situation where the administration of an ineffective substance produces an analgesic effect when the subject is convinced that the substance is a potent analgesic. Accordingly, any pain therapy inevitably depends upon a significant placebo component, determined by the patient's expectation.
Although placebo effects play a minor role in studies with ‘hard’ end-points (3) (e.g. death, myocardial infarction and stroke), they are of paramount significance when evaluating subjective parameters such as pain or mood. It is astonishing how many therapeutic regimens in pain research and psychiatry prove to be highly effective in open studies. In many cases the efficacy of these successful therapies cannot be replicated in placebo-controlled trials (3). Therefore, regulatory authorities such as the European Medicines Agency and the Food and Drug Administration appropriately require properly conducted placebo-controlled trials for the development and approval of drugs for acute or prophylactic treatment of headache. Unfortunately, these rigorous requirements do not need to be met for approval of medical devices. In order to plan randomized, controlled studies using placebo, variables influencing the placebo response must be known and taken into consideration when calculating the power of a particular trial.
Psychological and physiological mechanisms of placebo analgesia
Placebo analgesia represents a complex phenomenon that can be attributed to different psychological mechanisms, including expectation, Pavlovian conditioning and reduction of anxiety. Placebo analgesia has long been considered to be a purely psychological phenomenon. However, intensive research in previous years has begun to elucidate the neurobiological mechanisms of placebo analgesia (for an excellent review on this topic see (4)). Even though no study to date has specifically assessed the psychobiology of the placebo response for headache pain, several important conclusions can be derived from the available studies of placebo analgesia using experimental pain models.
The psychological mechanisms of placebo analgesia that have been extensively investigated are conditioning and expectancy (5). The expectancy mechanism, which can be manipulated by verbal cues and conditioning protocols, refers to an individual's expectation of an anticipated emotional and/or physiological response to the treatment being administered. The conditioning (learning) mechanism refers to the ability of a previously neutral stimulus/treatment to elicit a response (conditioned response) after having been paired with an active stimulus/treatment (5, 6). The role of conditioning mechanisms is illustrated in a study that demonstrated that patients with experimental ischaemic arm pain, who were pretreated with morphine sulphate, experienced analgesia to treatment with an antibiotic 2 days later, despite knowing that the antibiotic had no analgesic properties (7).
Functional brain imaging studies have contributed substantially to our knowledge of the mechanism of action of placebo. The pioneering neuroimaging study on placebo analgesia was performed by Petrovic and colleagues. This study revealed a shared neural network of the rostral anterior cingulate cortex (rACC) and the brainstem (PAG) underlying both opioid and systemic placebo analgesia (intravenous NaCl) (8). Also, in a model of a somatotopically localized placebo analgesia (induced by applying a placebo cream on one hand), placebo-related activity was found to be located in the rACC. A connectivity analysis identified placebo-dependent contributions of rACC activity, bilateral amygdalae and the PAG (9). A similar network was found to contribute to the effect of an ‘alternative’ placebo intervention, namely sham acupuncture (10), and also to the more general expectation of relief in a recent study (11), testing how expectation attenuates sensory taste transmission of a bitter taste. The converging evidence from these studies supports a common underlying mechanism of expectation-induced placebo analgesia (in models of experimental pain) that depends on the enhanced functional connectivity of the rACC with subcortical brain structures that are crucial for descending inhibition of nociceptive information.
The majority of neuroimaging studies of placebo analgesia indicate that the reduced pain ratings during placebo analgesia are paralleled by decreased activity in the classical pain processing areas, including the thalamus, insula and somatosensory cortices (8, 9, 12, 13). This supports the notion that the altered pain experience during placebo analgesia results from active inhibition of nociceptive input, and does not simply result from the re-appraisal of otherwise unchanged input into the nociceptive brain or report bias.
Another brain area that appears to be important for the generation of placebo analgesia (and other mechanisms of cognitive modulation of pain) is the prefrontal cortex. Wager used a placebo design that allowed for dissociating the expectation phase from the actual placebo perception phase and demonstrated activity in the dorsolateral prefrontal cortex during the anticipation of placebo that correlated with stronger placebo response (12). The prefrontal cortex is thought to be necessary to generate, maintain and integrate internal representations and expectations, which represents a likely explanation for its contribution to pain relief in different behavioural contexts such as placebo analgesia.
The clinical relevance of the role of the prefrontal cortex in placebo analgesia is illustrated in a study that assessed the interaction of verum and placebo mechanisms in patients with Alzheimer's disease (14). Reduced placebo effects were observed with reduced scores in the Frontal Assessment Battery and reduced connectivity of the frontal lobes with other brain areas. The loss of these placebo-related mechanisms reduced treatment efficacy, so that a dose increase was necessary to produce adequate analgesia. This study indicated that loss of the frontal lobe-mediated expectancy mechanism disrupts placebo analgesia and thus makes analgesic therapies (which in the clinical environment always have an additional expectancy component) less effective. These findings highlight the active role of cognition and prefrontal lobes in therapeutic outcome.
To date, there has been no neuroimaging study on the placebo action on experimental head pain or clinical headache conditions. It will be of great interest to determine whether the same and/or additional brain structures specific to migraine or cluster headache are involved in the action of placebo on these headaches. Since the 1970s, pharmacological studies have indicated that placebo analgesia can be antagonized by the opioid-antagonist naloxone, implying that at least some aspects of placebo analgesia depend on the endogenous opioid system (15–17). In support of this view, a recent study using radioligand 11C-carfentanil positron emission tomography has convincingly demonstrated activation of µ-opioid receptor-mediated neurotransmission during placebo analgesia in those cortical and subcortical brain regions previously associated with placebo analgesia (18).
Activation of the endogenous opioid system, however, cannot be the only pathway of pain modulation in humans, as is demonstrated by neuropharmacological studies that have convincingly dissected opioid- and non-opioid-dependent mechanisms of placebo analgesia, depending on the procedure applied to induce the placebo effect and previous drug experience (7). Future studies will be needed to explore the role of other neurotransmitters that are also involved in pain control, such as the dopaminergic, serotonergic and cannabinoid systems, and the selective involvement of different neurotransmitter systems for the placebo response in specific headache syndromes.
In addition to the above-mentioned mechanisms, some special characteristics must be considered when evaluating the potential mechanisms of action of placebo on acute headache episodes. Since migraine and cluster headache are both limited in duration, improvement in clinical trials cannot solely be attributed to the actual placebo response, but may also be due to the spontaneous remission or natural improvement of the headache. This is also important in headache prophylaxis studies, as headache frequency and severity is a dynamic process cycling over time from periods of severe disability to periods of relative remission (19). ‘Regression to the mean’, which refers to a significant reduction in migraine frequency during the baseline period of a clinical trial, must be considered when evaluating the results of both active and placebo-treated patients (20). The clinical correlate of this phenomenon occurs when patients experience a significant reduction in migraine frequency while waiting for their appointment with a physician (21).
Placebo effect in treatment of acute headache
Overall, the placebo effect is robust in the acute treatment of migraine attacks with analgesics. Treating migraine attacks with aspirin and metoclopramide resulted in a placebo response of 24–26% for the primary end-point improvement of headache relief (severe or moderate headache diminishes to mild or no headache) (22–24). Studies conducted by Tfelt-Hansen (22) and Henry (23) used oral medication, whereas in the study by Diener, aspirin was delivered by the intravenous route (24).
Bendtsen et al. have published a review that compared 11 placebo-controlled studies of treatment of migraine attacks with analgesics (25). The response to placebo for headache relief was on average 30% [95% confidence interval (CI) 23, 36], whereas the placebo response rate for the parameter pain free (severe or moderate headache declines to no pain) after 2 h was 9% (95% CI 7, 12). Interestingly, there was no difference in the response to placebo for oral compared with parenteral application of the analgesic. This review shows that the placebo rate for the end-point ‘pain free’ is far lower than for headache relief, and CIs were very narrow. The International Headache Society therefore recommends ‘pain free after two hours’ as the primary end-point in studies of acute treatment of migraine attacks (26).
The response rates to placebo in triptan trials were published in a meta-analysis by Ferrari et al. (27) The end-point headache relief had an average response rate to placebo of 30%. The response has been far lower in studies comparing eletriptan and sumatriptan with placebo (24%) compared with other triptans, and higher in studies using almotriptan (35%). Similar to the studies with analgesics, triptan studies have had a significantly lower placebo rate for pain-free end-point (4–9%) in comparison with the headache relief end-point (28). A more recent meta-analysis including 30 981 patients from 98 studies investigating treatment of acute migraine attacks has reported that 28.6% of patients improved after 2 h with placebo, compared with 8.8% who were pain free (29).
For an overview of all studies, see Table 1.
Placebo effect in treatment of acute headache
ASA, acetylsalicylate acid.
Placebo effect in the prophylactic treatment of headache
Holroyd and Penzien have published a meta-analysis of all placebo-controlled studies for the use of propranolol as prophylactic treatment of migraine (30). The study reported data from 787 patients. The response rate (reduction of migraine frequency by > 50%) was 55.1% for propranolol and 14.3% for placebo. This result suggests that the response rate for placebo is lower in migraine prophylaxis than in the acute treatment of migraine attacks. Further studies published subsequently have shown placebo response rates of 14–21% for valproic acid (31, 32), 16% for magnesium (33), 22% for bisoprolol (34) and 31% for propranolol (35). A recent meta-analysis of 32 studies has found a placebo responder rate of 21%. The placebo response rates were significantly higher in parallel group studies compared with crossover trials (36). Altogether, studies of prophylactic treatment of migraine have demonstrated a higher variability in the rate of the placebo response than studies of the treatment of acute migraine attacks. This is probably due in part to the different primary end-points used in studies of migraine prophylaxis, and to the inherent variability in response measured over a period of months compared with one measured over a period of hours.
Meanwhile, results of the placebo response in studies evaluating non-pharmacological prophylactic treatment of migraine are available. In both large-scale studies investigating the efficacy of acupuncture as a prophylactic treatment, traditional Chinese acupuncture was compared with sham acupuncture (37, 38). A third group of patients was put on a waiting list or treated with conventional pharmacological therapy. Both studies revealed a responder rate for real vs. sham acupuncture of 50%. These results indicate that an impressive placebo effect occurs in sham acupuncture for migraine prophylaxis. The response rates to ‘placebo’ acupuncture are much higher compared with pharmacological prophylactic treatment of migraine. A possible explanation is that patients who have a high expectation of benefit from acupuncture are more likely to enter these trials.
For an overview of all studies, see Table 2.
Placebo effect in the prophylactic treatment of headache
Application of treatment and placebo effect
Theoretically, the route of administration (tablet, nasal spray or injection) and the treatment environment (medical office, home, or in-patient) could influence the response to placebo. An analysis of several sumatriptan studies including 1800 patients showed no difference in response to placebo when treated at home or in hospital (39). However, the response rate to placebo was higher when administered subcutaneously in comparison with oral medication. Macedo et al. have summarized six trials with subcutaneous, 63 with oral and four with nasal application of placebo and found pain-free rates for placebo after 2 h of 14.9, 8.8 and 7.5%, respectively (29). Generally, parenteral treatments have a higher placebo rate when used in pain therapy than placebo pills (40). This might explain why the placebo rates were much higher in trials using botulinum toxin than in those using oral medication (41, 42).
Effect of placebo and expectation
Given that placebo analgesia inherently depends on the expectation of pain relief, it is extremely important to control and homogenize the magnitude of expectation in clinical trials and across clinical trials. Several studies have shown the importance of expectation for the size of the placebo response. An important study of Baussel and colleagues (43) had the following design: each subject had a tooth extraction; 100 subjects were treated with Chinese acupuncture and 197 with sham acupuncture; all were asked for their opinion as to which treatment they thought they had received. Concerning elapsed time before being free of pain, there was no significant difference between the two groups. However, those subjects who believed they were being treated with Chinese acupuncture had significantly less pain compared with those who believed they were treated with placebo acupuncture.
Another variable modifying patients’ expectation is the proportion of randomization. In a study comparing zolmitriptan 5 mg, sumatriptan 100 mg and placebo, the proportion of randomization was 8:8:1 (44). All subjects were informed that the probability of active drug:placebo treatment was 16:1. In this study there was no difference for any end-point for zolmitriptan, sumatriptan or placebo. This is an impressive demonstration of how expectation modifies the rate of placebo response. The practical consequence is that randomization should be done in a 1:1 proportion.
However, we are aware how difficult it is to separate a placebo effect from the spontaneous course of a migraine attack. With regard to spontaneous remission of acute attacks, a recent study has evaluated the natural history of untreated attacks of migraine (45). Brief attacks were unusual, most (nearly 75%) lasted 4–72 h, whereas > 20% of subjects in each group reported that their attacks typically lasted > 72 h. Therefore, spontaneous remission within 2 h of treatment, the primary end-point time evaluated in acute migraine trials, especially of attacks that are moderate or severe at the time of treatment, would be expected to be very low—certainly well below the 30–40% response rates to placebo seen in many acute migraine treatment trials.
Duration of placebo effect
Traditionally, it has been believed that the placebo effect lasts for a short time period, and frequency of headache returns afterwards to its prior level, particularly in studies for the prophylactic treatment of migraine. This argument was refuted in the topiramate studies, which had a treatment period of 6 months. The placebo effect was seen after 4 weeks and maintained constant for 6 months (46–49). Similar results have been seen in the treatment of chronic headache with botulinum toxin (41, 42). In the botulinum toxin studies, the placebo effect was not only maintained for a time period of 9 months, but increased after each of three successive injections. These observations have implications for understanding the pathophysiology of migraine and the mechanism(s) of placebo. If a method, a procedure or a drug is able to reduce the frequency of migraine attacks, the therapeutic effect may be sustained over long periods of time. Spontaneous remission, however, is a contributing factor that is difficult to control. As regards spontaneous remission of migraine itself, the botulinum toxin studies are a clear example of a prolonged, sustained and incremental placebo response over 11 months. One would not expect a remission rate of > 40% over an 11-month period of time when these same patients have had active migraine for an average of 15 years’ duration.
Absent placebo effect
All placebo-controlled studies conducted for the treatment of headache show a placebo effect. The only exception to this rule was seen in two recent studies in patients with chronic migraine and medication overuse headache (Diener H, personal communication, 2007). Patients were not withdrawn from the overused medication, but randomized to receive either topiramate or placebo. There was no reduction in migraine attack frequency in either study in those treated with placebo, whereas topiramate significantly reduced the number of migraine days. The absent placebo response under these conditions may reflect a fundamental alteration in the nociceptive and/or pain-modulating system as a result of acute headache medication overuse in such a way that renders them unaffected by a placebo.
Effect of placebo and cultural differences
Acute migraine studies involving eletriptan and sumatriptan have demonstrated significantly higher rates of response to placebo in the USA compared with Europe. The response rate for placebo for eletriptan was 24.1% in the USA and 20.1% in Europe (51, 52). The respective numbers for rizatriptan were 38.1 and 29.2% (53). The sumatriptan-naratriptan meta-analysis revealed the highest rates of response to placebo for Canada and Sweden with 36%, and the lowest rates for Belgium and Denmark (Diener H, personal communication, 2007). Overall the placebo rate for European countries was 27%, for the USA 31% and for Canada 36%. The 2-h pain-free end-point had a placebo rate of 9% in Europe and 14% in Canada (Diener H, personal communication, 2007). On the other hand, the meta-analysis by Macedo et al. (29) reported pain relief after placebo in 15 studies in Europe of 31.1% and in 22 studies in North America of 28.8%. Cultural differences were also seen in the occurrence of side-effects with placebo. These were significantly more frequent in North America (29%) compared with Europe (17%) (54).
Effect of placebo in childhood and adolescence
It is well known that children and adolescents exhibit far higher placebo response rates than adults. This may be due to the fact that expectation is more pronounced than in adults. The high placebo rate is responsible for the fact that nearly all trials of the use of oral triptans for children and adolescents have failed to demonstrate superiority over placebo. Although the response rates in children and adolescents mirror those seen in adults, the response to placebo is significantly higher in children and almost matches the response to active drug (55–61). In the study by Winner et al. (61), adolescents who had no prior experience with triptans had a higher placebo rate than those who had. These findings make it imperative that the design of acute migraine trials in children and adolescents is changed (62). Crossover studies in which up to five attacks are treated alternatively with active drug or placebo would provide more valid information.
In the sumatriptan-naratriptan database, 1153 migraine attacks after intake of placebo and 2881 attacks after intake of sumatriptan were analysed (Diener H, personal communication, 2007). This analysis showed a linear correlation between age and efficacy—triptan efficacy improved with advancing age. On the other hand, the response to placebo increased dramatically in adolescents < 16 years old. These data point to the fact that different study designs will be necessary in children and adolescents compared with adults.
Placebo effect in the treatment of acute attacks in cluster headache
Interestingly, a placebo effect has also been seen in the treatment of cluster attacks. A review in 2003 of six crossover studies showed response rates to placebo of 7–42% in the treatment of acute attacks of cluster headache (63). The fact that cluster attacks are of such short duration has to be taken in consideration for designing randomized, placebo-controlled trials. The time to collection of end-point data such as ‘pain free’ should be 15 or 30 min and not after 2 h as in migraine trials. Even taking this into consideration, early sumatriptan trials in cluster headache have revealed pain-free response rates to subcutaneous sumatriptan of 25% at 10 min and 35% at 15 min (64).
Side-effects after intake of placebo
Placebo can surprisingly cause side-effects. In a trial published in 2001, we compared 5 mg flunarizine, 10 mg flunarizine and 160 mg propranolol in the prophylactic treatment of migraine (65). All patients were informed that flunarizine could result in weight gain, but the patients did not know about a 4-week placebo run-in phase in which all subjects were treated with placebo before being randomized to one of three treatment groups. In the placebo run-in phase we observed significant weight gain in 10% of patients, whereas in the following treatment phase significant weight gain was seen only in 6.5% of all patients. Reuter and co-workers analysed 57 trials for the acute treatment of headache concerning adverse events (AEs) following administration of placebo (66). The incidence of AEs ranged from 10 to 30%. The most frequently reported side-effects were nausea, angina pectoris-like episodes and sleep disturbance. The meta-analysis of Macedo et al. (36) observed a higher rate of AEs with placebo in studies performed in North America compared with Europe.
The incidence of side-effects depends on the disease treated with placebo. An analysis conducted by Gauler and Weihrauch (67) revealed a rate of headache as a side-effect during placebo treatment of 16.7% for anxiety disorders, whereas the rates for angina pectoris and stroke were, respectively, 3.4 and 0.3%. In conclusion, this analysis showed that the incidence of headache as a side-effect depends of the disease being treated and has the highest prevalence in anxiety disorder trials. Anxiety disorders and primary headache disorders are comorbid and, because they are usually not excluded from headache studies, these patients may influence the placebo response rate of both treatment arms.
Ethical aspects of the use of placebo
The availability of effective treatments for a particular condition is frequently used to support the argument that the use of placebo in testing drugs for particular diseases is unethical. As regards migraine trials, the counter-argument is that patients participate voluntarily in clinical trials after signing informed consent in which available treatment options are detailed. Second, patients are permitted to use rescue medication after only 2 h. Finally, the placebo response rate in acute pain trials is high, and determining the true biological effect of a new drug is therefore of paramount importance before it is introduced for general consumption. In addition, only a placebo group can reveal whether the examined population is comparable to patients in other studies.
Even more important is the use of placebo in trials for the preventive therapy of primary headaches. For a limited time, German investigators were not permitted to include a placebo arm in prophylactic migraine trials according to ethics committee rulings (65, 68). In these trials, flunarizine and propranolol and metoprolol and aspirin were compared. Due to the lack of a placebo group, we could not assess whether a responder rate of 31% for acetylsalicylate acid was a real effect or if the placebo rate would be approximately the same.
Final considerations
This review has summarized present knowledge about effects and side-effects of placebo. Due to the requirements of the recent German/European drug laws it has become almost impossible to conduct experimental trials to assess the efficacy of different instructions and different routes of administration of placebo. The drug law states that the patient information sheet has to contain information about the use and route of administration of placebo, even if this variable is the one under study. Therefore, we must encourage colleagues in countries outside the EU to perform trials of this kind in order to learn more about the mechanism of action of placebo.
Competing interests
H-C.D. has received honoraria for participation in clinical trials, contribution to advisory boards or oral presentations from: Addex Pharma, Allergan, Almirall, AstraZeneca, Bayer Vital, Berlin Chemie, CoLucid, Böhringer Ingelheim, Bristol-Myers Squibb, GlaxoSmithKline, Grünenthal, Janssen-Cilag, Lilly, La Roche, 3M Medica, MSD, Novartis, Johnson & Johnson, Pierre Fabre, Pfizer, Schaper and Brümmer, SanofiAventis, Weber & Weber. Financial support for research projects was provided by Allergan, Almirall, AstraZeneca, Bayer, GSK, Janssen-Cilag, Pfizer. Headache research at the Department of Neurology in Essen is supported by the German Research Council (DFG), the German Ministry of Education and Research (BMBF) and the European Union. H.C.D. has no ownership interest and does not own stocks of any pharmaceutical company. D.D. has received honoraria for consulting activities with the following companies: Allergan, Pfizer, OrthoMcNeil, Endo, Solvay, Addex, St Jude, GlaxoSmithKline, HS Lundberg, Merck, Neuralieve, Coherex. He is the site principle investigator of clinical trials sponsored by the following companies: Advanced Neurostimulation Systems, Medtronic. He has no ownership interest and does not own stocks of any pharmaceutical company or device manufacturer.
Acknowledgements
This review is based on a lecture given by the first author at the IHC Congress in 2005 in Kyoto and, in addition, on material known to the authors and a Medline search up to September 2007 using the search terms: ‘placebo’, ‘migraine’, ‘tension-type headache’, ‘cluster headache’, ‘treatment’, ‘therapy’ and ‘prophylaxis’. In addition, the journals Headache and Cephalalgias were hand searched by the first author. Supported by the German Ministry of Education and Research (BMBF O1EM0513).