Abstract
Cluster headache is a trigemino-autonomic cephalgia with a low prevalence. Several population-based studies on its prevalence and incidence have been performed, but with different methodology resulting in different figures. We analysed all available population-based epidemiological studies on cluster headache and compared the data in a meta-analysis. The pooled data showed a lifetime prevalence of 124 per 100 000 [confidence interval (CI) 101, 151] and a 1-year prevalence of 53 per 100 000 (CI 26, 95). The overall sex ratio was 4.3 (male to female), it was higher in chronic cluster headache (15.0) compared with episodic cluster headache (3.8). The ratio of episodic vs. chronic cluster headache was 6.0. Our analysis revealed a relatively stable lifetime prevalence, which suggests that about one in 1000 people suffers from cluster headache, the prevalence being independent of the region of the population study. The sex ratio (male to female) is higher than published in several patient-based epidemiological studies.
Introduction
Cluster headache, a rare and severe primary headache, is characterized by unilateral pain and untreated headache attacks shorter than 3 h, ipsilateral autonomic symptoms and often restlessness. The diagnosis is made according to the criteria of the International Headache Society (IHS), which were published in 1988 (1) and in a revised version in 2004 (2).
Little is known about the population-based epidemiology of cluster headache, most data originate from case series on prevalence surveys. In addition, even fewer incidence studies exist which can be explained by the known very low prevalence of cluster headache. In contrast, several patient sample studies on the characteristics of cluster headache have been published. However, these studies are subject to selection bias such as patients' admittance to healthcare, knowledge about diagnostic criteria and cultural influences.
Although cluster headache is regarded as an idiopathic, purely somatic headache disorder, it cannot be ruled out that cluster headache is also influenced by sociocultural factors. Therefore, it is important to compare epidemiological data from different regions and different cultures to elucidate such relations.
Methods
We searched for all population-based studies on the epidemiology of cluster headache published as a full paper in a scientific journal or book as referenced by MedLine, EMBASE and Currents Contents up to August 2007; in addition, we screened all available monographs on headache and books with congress contributions published as full papers and screened the reference lists of the available publications. As keywords we used ‘cluster headache’, ‘epidemiology’, ‘prevalence’ and ‘incidence’. All studies with population-based data on incidence or prevalence (independent of the respective time frame) and giving at least the absolute number of cases and of the total population were included. Thus, we identified 16 studies. One study from China (3) was excluded because it presented extremely low figures for all headache types, making it likely that different methods were applied (migraine prevalence 0.7%; cluster headache prevalence 0.006%).
All studies used a very similar design. Either by mailed questionnaire or by structured interview, the population was screened for possible cases of cluster headache by asking for the classification criteria of the IHS. This screening was followed by interviews performed by neurologists or trained interviewers and by an evaluation of the possible cases, including neurological examination. Only one study (4) relied exclusively on chart review.
We recorded all available data of the study population including age and sex distribution of the patients, including age, sex and course of cluster headache (episodic vs. chronic). We then recalculated the prevalence (and/or incidence) rate per 100 000 only considering the cases fulfilling the criteria for cluster headache according to the IHS criteria from 1988 or from 2004.
The data were pooled according to the study primary outcome (prevalence in different time frame and incidence) and the rates and the respective 95% confidence intervals (CI) were calculated.
Results
In Table 1, all studies are listed with their respective data. The lifetime prevalence rates of cluster headache ranged from 56 to 381 (one study did not find a cluster headache case). The 1-year prevalence ranged from 3 to 150 per 100 000 (also, one study did not find a cluster headache case). The 1-year incidence was calculated with 2–10 per 100 000 in four studies, three of them performed in the same region.
Epidemiological population-based surveys of cluster headache
Cluster headache according to IHS criteria (excluding probable cluster headache).
c, chronic cluster headache; e, episodic cluster headache; F, female; M, male; NA, not applicable; ND, no data.
We excluded the first studies from San Marino (5) and from Olmsted (6) since later studies explored the same population. After pooling the data of all other studies with respect to their different designs, we calculated a worldwide lifetime prevalence rate of cluster headache of 124 per 100 000 (CI 101, 151) and a 1-year prevalence of 53 per 100 000 (CI 26, 95). Sufficient data for pooling the 1-year incidence was given in only one study (4).
The sex ratio (male divided by female) of cluster headache ranged from 1.3 to 14.0. Interestingly, both the highest and the lowest rates were from the same country. When pooling all studies, the overall sex ratio was 4.3. With respect to the course of cluster headache, the pooled data revealed that 96 patients had an episodic type (sex ratio 3.8) and 16 had a chronic type (sex ratio 15.0); insufficient data were obtained for the remaining 71 cluster headache patients. The overall ratio for episodic vs. chronic cluster headache was 6.0.
Discussion
In this paper, we aimed to analyse and pool all available population-based studies on the epidemiology of cluster headache. In contrast to migraine, the low prevalence of cluster headache makes it difficult to obtain exact data on prevalence or incidence because very large population samples are needed. On the other hand, the very specific symptomatology of cluster headache makes it easy to detect a single case unanimously.
The pooled lifetime prevalence is 124 per 100 000 (0.12%) for adults of all ages and both sexes (except for two studies that investigated only male subjects). The 1-year prevalence was, as would be expected, lower with 53 per 100 000. The incidence of cluster headache is very difficult to determine. Only three studies evaluated these data with rates between 2, 2.5 and 10 per 100 000, respectively. The incidence rates of two and 10 per 100 000 were obtained from the same population in a time frame of 10 years. Whether or not the hypothesis of an increasing incidence of cluster headache in the last decade (4) is correct cannot be determined on the basis of the published population-based studies.
Several studies on cluster headache epidemiology in patients have been published with detailed epidemiological data on these selected patient samples. Comparing the data of the population-based studies with those of the patient sample studies, we detected some remarkable differences. The sex ratio in our analysis was 4.3, which is higher than in most of the patient sample studies with 2.5 (7), 2.6 (8), 3.2 (9), 3.5 (10), 3.5 (11), or 3.7 (12). Only one Taiwanese (13) and one British (14) patient sample study showed higher sex ratios with 6.4 and 15.5, respectively; one patient sample study showed a very similar ratio of 4.4 (15). We could also not confirm the finding that the sex ratio has been changing over recent decades with an increasing rate of female cluster headache patients (10, 13, 15). This hypothesis remains controversial in the literature, since our analysis and a patient sample study (7) could not reproduce this finding. The ratio of episodic vs. chronic cluster headache was 6.0 in our analysis (however, with a large proportion of patients with an undetermined course). In patient sample studies, this ratio was 3.8 (7), 3.5 (12), 4.5 (11), 7.5 (10), 8.2 (15) and even 13.7 (14). One study with 104 patients did not identify any patient with chronic cluster headache (13). The sex ratio for episodic cluster headache (3.8 in our analysis) was lower than that for chronic cluster headache (15.0 in our analysis). This confirms previous reports of patient sample studies (10, 15). In summary, the male preponderance was on average higher in population-based studies than in patient sample studies, whereas the ratio of episodic vs. chronic cluster headache was about the same (with a unanimously clearly higher male preponderance in chronic cluster headache).
Regional differences in the prevalence of cluster headache could not really be disclosed by our analysis due to the limited number of studies. However, there was a trend that in the more northern countries the prevalence rates were higher than in those countries nearer to the equator. The studies from Ethiopia, Malaysia and the People's Republic of China suggest that cluster headache might be less frequent in developing countries. Interestingly, no studies from countries south of the equator are available. Studies from the same country, which are available for Scandinavia, Germany and Italy, revealed very similar prevalence rates.
This meta-analysis has several limitations. First, all studies had a different design in the details of identifying and interviewing patients, making it difficult to compare the data. Therefore, the population samples investigated were different with respect to age, sex and how representative the city was of the country. Second, the criteria applied for cluster headache were different. Most of the studies applied the first version of the IHS criteria. However, two studies were performed before and two after that era. Since cluster headache is a very specific clinical entity and since the criteria of cluster headache are very similar between the two classifications, we think that this difference is of minor importance. Third, the sample sizes were very different, leading to different sizes of the confidence intervals. This is a statistical problem, since in smaller studies one single case could have changed the prevalence rates considerably.
In conclusion, cluster headache shows a lifetime prevalence of 124 and a 1-year prevalence of 53 on the basis of the published population-based studies. The male to female ratio is 4.3 (and higher than reported in many patient sample studies). The ratio of episodic vs. chronic cluster headache is 6.0, with a higher male preponderance in episodic compared with chronic cluster headache.