Abstract
Probably because of its relative rarity as primary headache, there are few well-controlled clinical trials on cluster headache (CH) patients. Due to the severity of the pain, the placebo response in CH has been considered to be small. During the eighties the first double-blind, placebo-controlled trials were reported, and placebo responses demonstrated. Here we review the placebo response in CH trials in order to assess its magnitude and consider how future studies can be optimized. Six trials were identified with a double-blind, placebo-controlled, cross-over design testing treatments of acute CH. For those with a primary endpoint set to no or mild headache the placebo responses varied from 7 to 42%. In five of seven prophylactic trials, using a double-blind, placebo-controlled, parallel-group design, the placebo was merely used to set a baseline for comparison. The placebo responses were reported in only two trials. Here the response varies from 14 to 43%, the lowest value was reported using the strict endpoint; cessation of headache attacks. We conclude that a placebo response exists in trials of drugs on CH patients. Furthermore, this placebo response is of the same magnitude as that seen in migraine studies. We recommend the use of IHS guidelines when designing new trials. The possibility of a genuine biological mechanism responsible for the placebo response is discussed.
Introduction
The natural course of cluster headache (CH), a disorder that frequently features annual periodicity, periods of different lengths, fluctuating intensity of the pain (1), and spontaneous remissions (2), makes the evaluation of drug efficacy challenging. For many years the placebo response was considered negligible, or at least small when testing treatments on CH patients (3). Most reports of therapeutic trials have therefore been case reports, or have used an open treatment approach (4). It was not until the 1980s that well controlled clinical trials on CH patients were published. Olesen et al. (5) pointed to the need for well-defined clinical criteria in order to get well-defined study populations. The need for well-documented, randomized, double-blind trials was further stressed by Ekbom (3). We have focused on the reported placebo responses in CH trials that have used well-controlled study designs. We consider medicines tested for both acute attack treatment and as prophylactic drugs.
Placebo response in acute attack treatment trials
Six studies were identified that used a double-blind placebo-controlled study design (Table 1). Each of these studies also employed a cross-over methodology.
Acute attack treatment trials of cluster headache patients
CH, cluster headache; sc, subcutaneously.
Fogan (6) published a study on normobaric oxygen as acute treatment of CH attacks. The placebo treatment was inhalation of air. The patients treated up to six attacks with each of the two separate treatments. Nine of 16 study patients (56%) became substantially relieved or headache free after breathing oxygen. One of 14 (7%) study patients had a similar response from breathing air.
Andersson and Jespersen (7) published a study where they tested dihydroergotamine (DHE) nasal spray as an acute treatment of CH attacks. A placebo nasal spray was employed as the control. The patients treated up to eight attacks with each of the two treatments. They reported no difference in frequency or duration of the attacks but there was a significant effect on the intensity of the individual attacks after administration of the active nasal spray. Nineteen out of 133 (14%) placebo treated attacks were also controlled (5%) or strongly reduced. If the number of attacks where the patient experienced a slight reduction in headache intensity were included, the placebo response was 36%.
In a multicentre trial the Sumatriptan Cluster Headache Study Group (8) investigated the effect of subcutaneously injected sumatriptan as acute treatment of CH attacks. Each patient treated one attack with sumatriptan and one other attack with a placebo injection, the order of which was randomized. Both injections were taken within 10 min of onset of the attack. The pain intensity was classified according to a 5-graded scale, where 0 = no pain, 1 = mild, 2 = moderate, 3 = severe and 4 = very severe pain, which from then became the standard scale upon which to assess treatment (headache intensity scale). The primary endpoint was complete, or almost complete, headache relief within 15 min after treatment. The active treatment was significantly better than placebo after five minutes. The placebo response was 8% at five minutes, increasing to 10% and 26% after 10 and 15 min, respectively. The headache attack was completely resolved within 30 min after placebo treatment in 33% of the treated attacks. Considering patients who become pain free the placebo response was 3% and 10% at 10 and 15 min after injection, respectively.
Ekbom et al. (9) performed a multicentre study where CH patients were treated with subcutaneous injections of either sumatriptan (6 mg or 12 mg) or placebo. Each patient received two of the possible three treatments. The patients were instructed to take their medication within 10 min from the onset of the headache attack and when the pain intensity was classified as at least moderate. Primary endpoint was headache relief, or at least a 2-grade decrease on the headache intensity scale, to mild or no headache 10 min after treatment. The reported placebo response was 25% and 35%, 10 and 15 min after injection, respectively. If only headache free patients were considered, the placebo response was 21% 15 min after treatment. If patients were asked to rate their global response to treatment, the placebo response was 13%, 26% and 33% 5, 10 and 15 min after treatment, respectively.
Oral zolmitriptan was tested in a multicentre study (10). One inclusion criteria of note was that patients should routinely have attacks of greater than 45 min duration. Each patient treated a total of three headache attacks all within 10 min of onset and with a pain intensity classified as at least moderate. They tested zolmitriptan (5mg and 10 mg) and placebo, the order of which was randomized. The primary endpoint was headache response: defined as a 2-grade or greater improvement from pretreatment headache intensity 30 min after treatment. The result was reported separately for episodic and chronic cluster headache patients. Considering patients with mild or no pain at 30 min the placebo response was 42% for episodic and 47% for chronic CH patients. Thirty-five percent and 44% of episodic and chronic CH patients, repectively, reported a meaningful headache relief after treatment with placebo.
Intranasal sumatriptan has recently been tested in acute CH attacks in a multicentre study (11). The patients were randomized to treat two attacks of at least moderate pain severity using intranasal sumatriptan or matching placebo. The patients estimated their headache severity at 5, 10, 15, 20 and 30 min after treatment. The primary endpoint was mild or no pain after 30 min. The placebo response was 30% and 33% in attack 1 and attack 2, respectively.
Summarizing remarks
The placebo response in studies of acute treatment of CH is summarized in Fig. 1. Only mild/no pain, or pain free placebo responses are presented. The placebo response varies from 7% to 42% in the different studies. The lowest value was based on only one study patient. The general opinion at that time was that there was no placebo response among CH patients. The last four presented trials have used a standard ordinal catergorial pain scale as documentation of headache intensity prior to treatment and as measure of drug efficacy facilitating comparison of placebo responses amongst studies. Furthermore, these studies all tested triptans as active drugs, but used two time endpoints, 15 and 30 min. For the two studies using injectable placebo responses were 26% and 35% at 15 min, whereas for the intranasal study it was 32% at 30 min and for the oral study 42% at 30 min. Although it is tempting to suggest a relationship between route of administration being tested and placebo response, the outcomes for the injectable and nasal forms were very similar. The placebo response in acute treatment trials may be influenced both by a true placebo effect and by the spontaneous end of headache attacks. With increasing time the last explanation becomes more likely, which might explain the different placebo responses found after 15 and 30 min, respectively.
The number of patients with mild and no headache or no headache after treatment with placebo and active drugs. Studies of oxygen inhalation (Oxygen inhal.) (6), dihydroergotamine nasal spray (DHE intranasal) (7), sumatriptan subcutaneously (Suma 6 mg sc) at 15 min (8, 9), sumatriptan nasal spray (Suma 20 mg intranasal) (11) and zolmitriptan 10 mg (Zolmi 10 mg oral) (10), at 30 min are illustrated. ░ Placebo (mild + no pain), ▪ Placebo (no pain), □ Active.
Placebo response in prophylactic treatment trials
A total of seven trials have used a longitudinal and placebo-controlled study design (Table 2). Six of these have been double-blinded.
Prophylactic treatment trials of cluster headache patients
Ep, episodic; Ch, chronic.
In an early trial by Ekbom (12) the prophylactic effect of pizotifen was investigated by using a single-blind technique. Patients with CH periods of at least one month were included. All study patients received placebo for five days, whereafter they were treated with active drug in three periods each five days long. The dosage of active drug was step-wise increased. The headache attacks were graded in severity using a 3-point scale, where 1 = slight pain, 2 = moderate pain and 3 = severe pain. A headache index (number of attacks times their degree of severity) was calculated for the pretreatment period and for each of the treatment periods. In the placebo group the mean headache index decreased from 12.8 (±2.8) to 11.6 (±3.2), i.e. a reduction of 9%, but this decrease was not statistically significant. One of 26 patients preferred the treatment with placebo, i.e. a placebo response of 3.8%. Two of the patients became entirely free from attacks during the placebo period. This was interpreted as due to the natural course of the disease with a spontaneous remission.
In the last few years placebo-controlled, double-blind and parallel-group design has been accepted as the standard method for conducting studies in primary headaches (13). However in CH studies, placebo has often been used during a run-in period to get a base-line and the placebo response itself not estimated. Monstad et al. (14) investigated the prophylactic effect of sumatriptan in a multicentre study. All patients treated one headache attack with one subcutaneous injection of 6 mg sumatriptan followed by sumatriptan 100 mg or placebo three times a day for one week. The placebo responses, when looking at the reduction of number of headache attacks and reduction of intensity, were 22 and 42%, respectively. The corresponding values for the active drug were 23 and 43%, respectively. When asking the patients opinion of the oral treatment, 53% of the patients on placebo and 54% of the patients on active drug rated the treatment as ineffective or poor, while 23% in the placebo group rated their treatment response as reasonable as compared with 13% in the sumatriptan group. The remaining patients considered their treatment to be good or excellent, i.e. a placebo response of 24%. The number of patients that experienced at least 50% reduction in the number of severe and very severe headache attacks comparing the last four days of the run-in week with the last four days of the treatment week, was 35% in the placebo group and 22% in the active treatment group. These differences were not statistically significant.
Steiner et al. (15) investigated the prophylactic effect of lithium in episodic CH patients with a remaining expected period of at least three weeks. The patients were randomized to treatment with either 800 mg lithium carbonate or placebo daily for seven days. The primary endpoint was cessation of attacks. The responses were 14% in the placebo and 15% in the active drug group. The difference was not statistically significant. The secondary endpoint was substantially better subjectively within one week of treatment. Here the placebo response was 43% compared with 62% in the lithium group. Because of some assumptions in the study design, superiority over placebo could not be demonstrated, and the trial was stopped.
Leone et al. (16) tested melatonin as prophylactic treatment. Each treatment group consisted of 10 patients and they were randomized to treat with either melatonin or placebo. Only two chronic CH patients were included and both of them were randomized to the placebo group. The mean number of headache attacks during a run-in week was compared with the mean number of headache attacks during the first and second week of treatment. There was no decrease in headache frequency in the placebo group, while in the melatonin group the mean headache frequency was reduced by 43% and 55% during the first and second week, respectively. In a similar study by the same group (17), the prophylactic effect of verapamil was evaluated in patients suffering from episodic CH. Fifteen patients were randomized to each of the two groups. Likewise, no decrease in headache frequency was shown in the placebo group as compared with the group of patients that received active treatment.
Marks et al. (18) compared intranasal capsaicin with intranasal placebo. Headache intensity was recorded on a visual analogue scale and average headache severity ratings pretreatment were compared with ratings during the first and second weeks of treatment. In the placebo group, the average headache severity rating decreased from 6.4 during the run-in week to 4.7 (27%) and 5.15 (20%) during the first and second treatment weeks, respectively. These improvements were not statistically significant. Among the capsaicin-treated patients there was a significant improvement during the second treatment week. However, the study population was small, the groups were not homogenous and blinding intranasal capsaicin has significant challenges.
A cross-over study, investigating the prophylactic effect of misoprostol included only patients with refractory chronic CH (19). Following a run-in period patients were randomized to treat with either misoprostol or placebo. The separate treatments continued for two weeks separated by a two-week wash-out period. No statistically significant differences were reported in headache frequency, duration or intensity.
Summarizing remarks
Five of the seven studies above reported lack of placebo response regarding efficacy. However, in the remaining two trials placebo responses of 14% and 43% were reported. The lowest value was reported when a very strict endpoint was used; cessation of headache. Most studies have used the placebo group as a baseline against which the active drug efficacy was compared. Furthermore, the placebo and the active drug groups were often inhomogenous with regard to inclusion of patients. None of these studies has reported the results separately for episodic and chronic CH patients.
Discussion
There are considerable challenges to be met in designing clinical trials in CH. Firstly, it is a relatively rare disease, with a prevalence of about 0.1% (20, 21), which requires either a multicentre approach, or long recruitment phase. Furthermore, two major subgroups of this disease can be distinguished, episodic and chronic CH (1), both with well defined clinical differences (22–25). In episodic CH different lengths and severity of headache attacks, different lengths of cluster periods with the occurrence of abortive periods (i.e. mild and extremely short periods) and the occurrence of spontaneous remissions can complicate the evaluation of drug efficacy. Among patients with chronic CH irregular frequency of attacks and periods of partial remission add to the list of issues that can lead to false positive drug responses. Further to these specific issues, there are the general difficulties affecting all types of drug trials, i.e. patients may change their life-style during the trial, the positive effect of participating in a study with expectation of receiving an effective medication and statistical issues, such as regression to the mean (26).
Is there a need for placebo in CH trials?
It is important to evaluate which factors can influence the placebo response. In acute attack treatment trials the frequencies of placebo responders were 7%−42%. When testing drugs as acute attack treatment most of the trials reported here have included both episodic and chronic CH patients. Only one of the trials reported the response separately for these two groups (10). This study reported different drug response rate in the two groups, but approximately the same placebo responses. Moreover, the placebo response reported in the different trials remained low if strict efficacy measurements were used as primary endpoints. The evaluation time point is of interest with regard to the placebo response, the longer the period from treatment to evaluation, the higher the placebo response is expected to be due to the natural course of the disease. However, it remains clear from the studies of subcutaneous sumatriptan that a substantial placebo response approaching that seen in acute migraine studies (27, 28) is seen in acute CH.
In the prophylactic trials a discrepancy between the frequency of placebo responders is observed. The placebo response in the sumatriptan study (14) was 22–42%. This was the only multicentre study with a large sample. Furthermore, the two treatment groups had an equal distribution of episodic and chronic cases. In the lithium study (15), a placebo response of 14% was reported. In this study a very strict primary endpoint was used: cessation of headache attacks. However, the two groups included only episodic CH patients, but they were included at very different time points from the start of their CH periods. Late inclusion in the study will enrol patients destined to spontaneously remit. In the studies by Leone et al. (16, 17) a lack of placebo response was reported. The study included only episodic CH patients of whom all were included very early in their cluster period. In the beginning of a cluster period, the attack frequency and intensity often shows an increasing trend to a maximum, whereafter a declining trend is observed. This may be the explanation for the low placebo response. In a further two trials (18, 19) no placebo responses were reported. In the former study, the placebo group consisted of five chronic cases, while the group receiving active drug was composed of a mix of five episodic and three chronic ones. In the latter study both groups consisted of refractory chronic cases. Although the numbers are small it is interesting that chronic CH patients seem to have a relatively modest placebo response.
Hyperbaric oxygen (HBO) treatment of CH using a double-blind placebo-controlled cross-over design has been tested in order to see if HBO treatment was able to induce a remission period. Both treatment with hyperbaric oxygen and hyperbaric air was equally effective in initiating such a remission period (29). In contrast 6 of 7 patients treated with HBO for acute CH responded while none of 6 treated with placebo responded (30). The existence of a biological mechanism responsible for the placebo effect has recently become a focus of interest. Recent PET studies performed on patients suffering from Parkinson's disease using 11C-raclopride, a dopamine antagonist, have shown a significantly reduction in 11C-raclopride compared to basal levels in patients that had received placebo (31). This indicates a release of endogenous dopamine as a response to placebo treatment. The levels of such a placebo-induced release was in the same order as therapeutic doses of levodopa or apomorphine. In another PET study investigating opioid and placebo-induced analgesia showed that the rostral anterior cingulate cortex together with brain stem areas are activated in both conditions (32). These areas are not activated during pain-only conditions. Several issues remain unanswered. Is the placebo mechanism always located to the same brain areas independent of drug/treatment under study (a brainstem/hypothalamic area)? Are different brain areas activated for different diseases? Is it the expectation of receiving a potent drug for a specific disease/disorder that activates the brain area where the disease is ‘situated’, and this expectation activates the brain area in same order of magnitude as the expected active drug? To this one must consider the problem how to construct a placebo preparation/treatment that can mimic the active drug to be tested in order to get a genuine double-blind situation, since as many as 80% of the patients can guess correctly to which group they were asigned during a double-blind regime (33).
We conclude that placebo response exists in CH trials and that the magnitude is in the same range as seen in migraine studies. Therefore, placebo must be considered, and for the most part, included in trials testing drug efficacy in CH. In acute treatment trials, as well as prophylactic treatment studies, it is the only unambiguous way to evaluate if positive responses are due to genuine drug effects or to the natural course of the disease. The use of placebo in prophylactic trials must however, be balanced against the ethical point of view. Long observation periods should be avoided in order to protect patients against unneeded discomfort. Furthermore, the patients should be allowed to use but not change their usual acute treatment. However, the amount of used acute treatment must be documented and reported, both in the group of patients receiving active drug and in the placebo group. Furthermore, we conclude that a strict study design needs to be used. When including episodic CH patients they ought to be included at approximately the same time point from the beginning of their CH period. The groups (placebo, active drug) should be homogenous regarding episodic and chronic cases as well as regarding gender, the result being best presented separately for the different subgroups. Patients should also be randomized (33). We recommend the use of IHS guidelines for controlled trials of drugs in CH (34). In addition, a third test arm might be considered, i.e. treatment with a drug that is in clinical use and with a documented clinical effect (32). The only type of study, in our view, where placebo can be excluded is a preference trial.
Footnotes
Acknowledgements
We wish to thank our colleagues in the Headache Masters Program for initiating the present study and for inspiring discussions at our meetings. We also wish to thank Dr Karl Ekbom for helping us with background materials and for always very inspiring discussions.