Abstract
Background
Early symptoms prior to a cluster headache bout have been reported to occur days or weeks before the actual beginning of the cluster headache bouts. This study aimed to describe the prevalence of pre-cluster (premonitory) symptoms and examine the predictability of an upcoming cluster headache bout.
Methods
100 patients with episodic cluster headache were included in this retrospective cross-sectional study. All patients underwent a semi-structured interview including 25 questions concerning pre-cluster symptoms.
Results
Pre-cluster symptoms were reported by 86% of patients with a mean of 6.8 days (interquartile range 3–14) preceding the bout. An ability to predict an upcoming bout was reported by 57% with a mean 4.6 days (interquartile range 2–7) before the bout. Occurrence of shadow attacks was associated with increased predictability (odds ratio: 3.06, confidence interval: 1.19–7.88, p-value = 0.020). In remission periods, 58% of patients reported mild cluster headache symptoms and 53% reported occurrence of single shadow attacks.
Conclusions
The majority of episodic cluster headache patients experienced pre-cluster symptoms, and more than half could predict an upcoming bout, suggesting the significant potential of early intervention. Furthermore, the experience of mild cluster headache symptoms and infrequent shadow attacks in remission periods is common and suggest an underlying pathophysiology extending beyond the cluster headache bouts.
Introduction
Cluster headache (CH) is known for its attack with severe unilateral pain in the orbital region accompanied by ipsilateral autonomic symptoms and/or a sense of restlessness (1). It affects one in 1000 in the general population (2), predominantly men, and typically with an onset before the age of 40 (3–5). About 85–90% of patients have episodic CH (eCH) with attacks in bouts lasting weeks to months followed by attack-free remission periods of more than 3 months (1).
A variety of symptoms have been reported to occur before the beginning of the bout (6–8) but they have not been precisely described or defined. These symptoms are in the following referred to as pre-cluster symptoms and arises in the days or weeks before the first attack in an upcoming CH bout, whereas pre-attack or preictal symptoms arise in the minutes or hours before each individual attack (9,10). A group of symptoms, here termed shadow attack, previously referred to as ‘shadows’ (11,12), lacks an official definition. Here, we define shadow attacks as a mild pain that mimics a regular attack in terms of location but is less intense, self-limiting, includes fewer and/or less severe additional symptoms and does not require abortive treatment.
An increased knowledge of pre-cluster symptoms would benefit both clinicians and patients as it could enable use of early intervention and reduce overall disease burden. Furthermore, studies of pre-cluster symptoms could contribute to the understanding of the pathogenic origin of CH, which is still an enigma.
The aim of this study was to investigate and describe the prevalence of pre-cluster symptoms in eCH patients and examine the predictability of an upcoming CH bout. We hypothesised that a large percentage of eCH patients experience pre-cluster symptoms and can predict an upcoming bout.
Methods
Participants
We included 100 eCH patients, recruited from the Danish Headache Center, a specialised tertiary public referral centre. Only patients above 18 years of age with the diagnosis of eCH according to the ICHD-3 criteria (1) were included. All diagnoses were validated by a headache specialist. To minimise recall bias, only patients who had experienced more than five CH bouts or had been in a bout during the last year were included (≤ 12 months). If they suffered from another episodic headache disorder, they had to be able to distinguish pre-cluster symptoms and CH attacks from the other headaches. Patients suffering from chronic headache disorders, significant psychiatric or somatic disease were excluded. All patients were treated by standard of care (13).
The literature on the subject is very sparse; however, we estimated, based on similar studies (6,8,9,14), that 100 participants would be sufficient to reliably detect the prevalence of pre-cluster symptoms.
Questionnaire
In this cross-sectional study design, data were collected through a semi-structured interview concerning nine specific local symptoms or signs (e.g. shooting pain/stabbing, neck pain) and 15 general symptoms or signs (e.g. irritability, fatigue) as well as shadow attacks (see complete questionnaire, supplementary data 1). In the following, there will be no distinction between symptoms and signs, and both will be referred to as symptoms. The questions were based on previous studies on migraine and CH and clinical experience in our CH-research group (6,8–10,14,15). All semi-structured interviews were conducted by a medical doctor and/or a specially trained medical student. Pre-cluster symptoms were clearly explained for the patients as CH-related symptoms that arise in the days or weeks preceding an upcoming CH bout. Furthermore, patients were asked about the presence of each individual symptom in the days or weeks prior to the first attack of a CH bout. If a symptom was reported, patients had to estimate how many days before the CH bout they experienced the symptom. Thus, it was explicit that the symptoms mentioned were before an upcoming bout and not in an existing bout. Patients were also asked if they had any additional local or general symptoms not mentioned in the questionnaire. If any pre-cluster symptoms were experienced, the patients were asked whether they felt they could predict their bout, how many days beforehand, and how likely it is that a bout would begin after the appearance of these symptoms. Thus, the prediction was made based on the mentioned pre-cluster symptoms. Patients were not asked which specific symptom they based their prediction on. Instead, the prediction was based on the entire symptomatology. More specific recordings should be done in a prospective study. Furthermore, they were asked if they had any of these specific symptoms in remission periods that were not followed by a bout (out-of-bout symptoms) and if they experienced shadow attacks or single attacks in remission periods. Single attacks were defined for the patients as a rare regular CH attack in remission periods without appearance of other subsequent attacks. If they reported any out-of-bout symptoms, they were asked if they could distinguish which symptoms were followed by a bout and, if so, how they were able to make that distinction.
Statistics
RStudio (version 3.6.0) was used to perform all statistical analyses. Demographic, clinical characteristics and prevalence of pre-cluster symptoms were calculated with absolute number, mean and interquartile range (IQR). Geometric mean was used to calculate the average days before an upcoming bout due to right-skewed data. The dataset was divided into two groups according to ability to predict an upcoming CH bout. Multiple logistic regression was used to generate a model that investigated the odds ratios (OR) of association with 95% confidence interval (CI) between predictability and three pre-defined variables: Gender, number of bouts and shadow attacks. No correction for multiple testing was done on the multiple logistic regression model due to the risk of a type 2 error. Pearson’s Chi-squared test was used to investigate homogeneity between symptom prevalence in the two groups and number of bouts experienced. Two-sample t-test was used to compare the total amount of symptoms experienced in the two groups. These p-values were adjusted using Bonferroni correction for multiple testing. All patients with more than one bout were included in the analysis on predictability and symptom prevalence between groups. A p-value of < 5% was considered statistically significant.
Patient consents and registration
All patients gave their informed written consent according to the Helsinki declaration and the Capital Region of Denmark Ethical Committee approved the study (H-16048941) as a part of a larger biomarker study and it was approved by the Danish data protection agency (RH-2018-71).
Data availability statement
The consent of the participants did not include freely accessible raw data. The dataset contains sensitive information and can consequently not be shared in full form according to Danish data protection law. De-identified data that underlies the results of this article can be shared on researcher request.
Results
A total of 100 patients were included in this study from January 2019 to April 2020 (Figure 1). The male:female ratio was 3.8:1 (Table 1). The mean age was 42.2 years (IQR 34–51) and the mean CH duration was 15 years (IQR 6–20). Furthermore, most patients had experienced more than five bouts (76%), and only seven patients (7%) were in their first CH bout when included in this study. These patients were confirmed episodic on next follow-up visit. The majority (67%) completed the questionnaire while in bout. Only 15 patients (15%) had been out of bout for a year or more when the questions were asked. For these 15 patients, the mean duration since last bout was 22.7 months (IQR 16–26). Additionally, 10 patients (10%) had comorbid migraine.
Flow chart of patient recruitment and inclusion.
Clinical characteristics of included patients, presented with total number (n) and percentage (%), or mean and interquartile range (IQR). The * marks significant difference between the two groups after correction for multiple testing. Difference in number of bouts experienced was the only hypothesis tested in Table 1.
Prevalence of pre-cluster symptoms
A total of 86 patients (86%) experienced at least one pre-cluster symptom. The mean debut of the first pre-cluster symptom was 6.8 days (IQR 3–14) before the CH bout. The two most frequent symptoms in the pre-cluster period were a dull/aching sensation and shadow attacks (Figure 2). Appearance of shadow attacks were reported by 60%, with a mean of 6.5 days (IQR 4–14) pre-cluster (Figure 3). A dull/aching sensation were experienced by 62%, with a mean of 5.8 days (IQR 3–13) pre-cluster. Neck pain was the third most common symptom reported by 41% with a mean duration of 4.8 days (IQR 2–7) pre-cluster. One week before an upcoming CH bout, 58% reported the experience of at least one symptom (Figure 4).
Prevalence of pre-cluster symptoms. Total number of patients that reported each specific symptom in the pre-cluster period.
Time debut of pre-cluster symptoms in days before the cluster start, mean and IQR. Geometric mean was used to calculate the average debut of each specific symptom in days before cluster onset. Additionally, the data is presented with IQR.
Progress in prevalence of pre-cluster symptoms. Development of symptoms for the 100 included patients. A total of 86 patients experienced at least one symptom in the pre-cluster period.
One or more cranial autonomic symptoms (CAS) (lacrimation, ptosis, redness of eye and rhinorrhea or nasal congestion) were experienced by 29% in the pre-cluster period with a mean of 4.5 days (IQR 2–7) days before onset of a CH bout. The most frequently reported general symptom was fatigue, present in 37%, with a mean beginning 5.3 days (IQR 2–10) pre-cluster. Irritability was the second most common general symptom, reported by 33%, with a mean start 5.2 days (IQR 3–7) pre-cluster. Sleep-related pre-cluster symptoms were common and 27% felt decreased energy level 4.2 days (IQR 2–7) before, and 19% reported sleep issues (e.g. trouble falling asleep) 5.3 days (IQR 3–7) before. Nocturnal awakenings were reported by 20% with a mean beginning 6.4 days (IQR 3–8) pre-cluster. Furthermore, three patients reported an increase in sleep duration.
Other local symptoms were reported by 13% and included painful sensations (e.g. in the scalp, chin, teeth or shoulder blade) or irritating sensations (e.g. in the eye or nose). Other general symptoms were reported by 20%. Mostly, a tendency to get mild diffuse headache reported by six patients or migraine attacks reported by one patient. The remaining descriptions were very individual (e.g. dizziness, feeling more affected by heat, feeling more stressed, increased craving for nicotine or sweets, dreams of having attacks).
Predictability of upcoming CH bout
Based on the experience of these pre-cluster symptoms, 53 of 93 patients with a CH history of ≥ 2 prior CH bouts (57%) answered that they were able to predict an upcoming bout with a mean of 4.6 days (IQR 2–7). However, the prediction was associated with some uncertainty. Among the 53 patients, 50 patients (94.3%) reported that it was likely (a possibility of at least 50%) that these symptoms were followed by a CH bout. Only 17 patients (32%) reported an almost certainty (a possibility of > 95%) of an upcoming bout when these symptoms appeared (Figure 5).
Self-reported predictability of upcoming cluster headache bout. Number of patients able to predict an upcoming cluster headache bout with estimated probability. The patients that had only experienced one bout were excluded in this figure.
Differences between patients that can and cannot predict an upcoming bout
In the group of patients that can predict an upcoming CH bout, 48 patients (90.6%) had experienced more than five bouts. In comparison, in the group not able to predict an upcoming bout, only 28 patients (59.6%) had experienced more than five bouts and seven patients (13.2%) had experienced only one bout. This difference was statistically significant (p = 0.046).
Symptom prevalence was higher among patients that could predict an upcoming bout, with a mean number of 6.7 symptoms experienced (IQR 4–9) compared to 3.8 symptoms (IQR 2–6) for patients that could not predict an upcoming bout (Figure 6). However, this difference was not statistically significant. When comparing the prevalence of each symptom between the groups, none of the symptoms were significantly more present in any of the groups after correction for multiple testing.
Prevalence of pre-cluster symptoms divided in groups according to ability to predict upcoming CH bout. The patients that had only experienced one bout were excluded in this figure. No significant difference was found in symptom prevalence between the two groups after correction for multiple testing.
Adjusted for age, gender and number of previous bouts, the odds ratio for predicting an upcoming bout was three times higher if patients experienced shadow attacks (OR: 3.06, CI: 1.19–7.88, p = 0.020). All patients with more than one CH bout were included in this analysis (93 patients).
Out-of-bout symptoms
In remission periods, 58 patients (58%) reported one or more symptoms that were not followed by a CH bout. Furthermore, 53% of the total patient population had experienced single shadow attacks and 33% had experienced single attacks in out-of-bout periods. However, 19% of the patients were able to clearly distinguish which symptoms are followed by a CH bout and thus able to predict an upcoming bout. In the group of patients that could not predict a bout, 23 patients (48.9%) reported out-of-bout symptoms and none were able to tell if the symptoms were followed by a bout.
The specific explanation behind this ability varied among the patients. However, the most frequent explanation reported by seven patients was that they knew the symptoms would not develop into a bout because they were too mild, too short, or too few. One patient reported that it had to be at the specific time of the year for the symptoms to develop to a bout. Another patient reported a specific nasal congestive feeling as a specific predictive symptom and two patients could not explain further.
Discussion
This study investigated the prevalence of pre-cluster symptoms in 100 well-characterised eCH patients and their ability to predict an upcoming CH bout. Our main findings were that 86% had experienced at least one pre-cluster symptom with a mean debut 6.8 days pre-cluster and 57% of the patients reported the ability to predict an upcoming bout. The experience of shadow attacks was associated with a three times higher predictability than in those without shadow attacks. Furthermore, we found that 58% also reported various milder symptoms and 53% also had single shadow attacks in remission periods.
The description of pre-cluster symptoms in CH literature is scarce, and to our knowledge this is the first study to investigate the predictability of CH bouts in a well-defined sample. In 1999, pre-cluster symptoms were reported by Blau and Engel (termed “premonitory symptoms”); however, only 12 out of 117 patients (10%) experienced symptoms 1 day to 8 weeks anteceding eCH bouts (6). Later, a prevalence of 33% was reported by Raimondi (7) without specifying the number of patients included in the analysis.
Compared to these publications, our study found a much larger percentage of patients that experience pre-cluster symptoms (6,7). Additionally, the current study identified a high prevalence of multiple symptoms that have never been reported before. There may be several explanations for this high occurrence compared to the previous studies. First, it is unspecified how the patients were asked in the previous studies (6,7). It may be that patients were not asked specifically but reported their experience in open questions. Our study used closed-ended questions for multiple specific symptoms, which may increase the patient’s ability to remember the pre-cluster period. Furthermore, the clinical characteristics of the included patients were not specified in the previous studies, which makes it difficult to compare the patients included in the different studies. It may also be that these studies included patients at the time of diagnosis, and consequently the patients might have had a lower disease awareness. Our study population had a mean disease duration of 15 years, which likely means that these were more experienced CH patients.
A third study by Torelli and Manzoni reported a prevalence of pre-cluster symptoms (in that study termed premonitory symptoms) in 17/42 eCH patients (40.5%) with a mean beginning 23.5 days pre-cluster (SD 18.8, range 3–75) (8). They used questionnaires with a list of 65 symptoms that the patients filled in themselves while in bout. Furthermore, the patients included had a mean disease duration of approximately 15 years. The main difference between Torelli’s study and the current one is that the former was based on self-reported questionnaires and the present study on semi-structured interviews.
The large percentage of patients that reported pre-cluster symptoms indicates that CH is not only restricted to the actual CH bout. The debut of symptoms is reported by 58% at least 1 week before an upcoming CH bout and increases dramatically in the days before the first CH attack. Thus, there is a potential for early intervention; that is, start-up preventive medication and arrangements for oxygen and triptans. The first-line drug of prophylactic treatment for CH is verapamil (16), which is effective in 50–94% of eCH patients (17) after approximately 2 weeks (18–20). Therefore, it is possible that verapamil treatment initiated early may improve effectiveness at the beginning of the bout, reducing the suffering for those patients that can predict attack onset. However, such an approach may increase the risk of over-treatment as not all symptoms develop into a manifest CH bout. In this regard, a controlled prospective study is needed, and a personalised approach may thus prove successful.
Predictability of upcoming CH bout
When excluding the seven patients that had only experienced one CH bout, 57% were able to predict an upcoming bout whereas 35% did report pre-cluster symptoms but were unable to predict an upcoming bout. Moreover, this prediction was associated with high uncertainty. Based on our findings, several factors may influence the predictability of a CH bout. First, patients able to predict an upcoming bout had experienced more bouts than patients not able to predict an upcoming bout. It is most likely, and not unexpected, that experiencing many bouts increases disease awareness and thus the ability to predict an upcoming cluster. Secondly, patients that experience shadow attacks have three times higher odds of predicting a CH bout, adjusted for gender and number of bouts experienced. Thus, shadow attacks may be interpreted as a predictor for an upcoming CH bout. However, 23% of the 93 patients with ≥ 2 CH bouts were unable to predict an upcoming bout even when having shadow attacks. This might be explained by the unexpected report of infrequent symptoms and shadow attacks in remission periods.
Figure 6 shows the prevalence of pre-cluster symptoms according to the ability to predict an upcoming bout, and it is clear the patients able to predict an upcoming bout had experienced more symptoms than the patients not able to predict an upoming bout. However, we found no significant difference in any symptom prevalence between the groups after correction for multiple testing.
Out-of-bout symptoms
The majority of the eCH patients in this study experienced symptoms without relation to a CH bout (58%). This indicates that even in out-of-bout periods the CH pathophysiology is present in eCH patients but does not always escalate into attacks. This is in line with CH studies on sleep disturbances, which find that even in remission periods CH patients have decreased sleep quality, suggesting that sleep disturbances are unrelated to CH attacks (21,22) and might be due to an ongoing underlying mechanism. However, some of these symptoms are unspecific and general and may be experienced by all people. This may complicate the prediction and partly explain the high prevalence of out-of-bout symptoms. When looking at the 17 patients with > 95% certainty in predictability, we find that these patients had fewer symptoms in remission periods and were more able to distinguish which symptoms will develop to a bout compared to the patients with 25–95% certainty. However, due to power issues, a formal test was not completed.
The occurrence of symptoms, infrequent shadow attacks and single attacks in remission periods increase the burden of CH beyond the bout-period. Some studies suggest that eCH patients have impaired quality of life also in out-of-bout periods (23,24), which might be due to an ongoing symptomatology.
Pathophysiology
The interesting finding of single symptoms in remission periods and the marked presence of multiple pre-cluster symptoms suggest an underlying neurological pathophysiology that is ongoing beyond the eCH bouts. The involvement of the hypothalamus in the pathophysiology of CH has been suggested several times in the literature and is supported by the symptomatology, especially the chronobiological features (25–27), neuroendocrine alterations (28), and imaging studies (29). Neck pain is experienced by 41% in the pre-cluster period and has been suggested to be a result of hypothalamic dysfunction (30) involving the trigeminocervical complex (TCC) (31,32). Thus, it is hypothesised that central changes in pain modulation involving the hypothalamus and possibly other structures could be a part of the underlying ongoing pathophysiology of cluster headache. However, it seems that even though symptoms in remission periods are common to a certain extent, they are not always enough to escalate to an actual bout, which could indicate that a threshold must be exceeded for regular attacks to occur. This is in line with imaging studies of migraine that suggest thresholds as an explanation behind pain perceptions (33). Paraclinical studies on patients with pre-cluster symptoms could contribute to the understanding of the onset of CH bouts.
Strengths and limitations
This study has several strengths and limitations. The retrospective design introduces the possibility of recall bias. However, the questionnaire investigates the appearance of specific named symptoms, which may increase the patient’s ability to remember the pre-cluster symptoms and thus reduces the recall bias. To further reduce the recall bias, the patients had to be in bout within the last year or have experienced more than five CH bouts. Thus, patients with more than a year since last bout were accepted, which may increase the risk of recall bias. However, to draw firm conclusions on pre-cluster prevalence and predictability of CH bouts a prospective design is needed.
Furthermore, this study did not investigate the relationship between CH bouts and the use of a well-known trigger factor such as alcohol before the bouts. It is possible that alcohol could trigger attacks when experiencing pre-cluster symptoms due to the indicated lowered threshold, as discussed above. However, it is unlikely that alcohol alone could cause a CH bout to develop. Nonetheless, this is unknown and should be examined in future studies.
A strength of this study is the fact that only one physician and/or one specially trained medical student interviewed the patients. Thus, the interviewer bias was minimal. A further strength is that all patients were recruited from the Danish Headache Center, a tertiary referral centre, and consequently were well characterised. Moreover, most patients have had the disorder for a long duration, which means they were able to answer the questions with high certainty. However, patients at a tertiary clinic might also suffer from a more severe condition than the average CH patient. Consequently, a selection bias may by present and generalisability may be limited to patients from specialised headache centres.
Lastly, this study included patients using low doses of preventive treatment in out-of-bout periods and eventually when experiencing pre-cluster symptoms. The definition of shadow attacks (i.e. not requiring abortive treatment) was carefully explained to patients, but we do not know whether the patients used abortive treatment or began using preventive treatment without consulting a doctor. Such patients might experience symptoms that do not develop into a bout or the pre-cluster symptoms might be masked by the preventive treatment.
Conclusion
This study reveals for the first time that the majority of eCH patients experience pre-cluster symptoms in days to weeks before their bout. More than half of the patients can predict an upcoming bout, and the experience of shadow attacks is associated with predictability. The time-period of these symptoms was within 2 weeks of the first attack, which could inspire patients and clinicians to consider and initiate early treatment intervention on an individual basis. Furthermore, the experience of mild symptoms and single shadow attacks in remission periods are also identified. These findings suggest an underlying pathophysiology that goes beyond the CH bouts and may be present even in remission periods.
Clinical implications
Based on pre-cluster symptoms, more than half of the episodic cluster headache (eCH) patients can predict an upcoming bout in the days or weeks preceding the bout. Thus, there is a potential for early initiation of preventive treatment. Neck pain, a dull/aching sensation and shadow attacks are very common pre-cluster symptoms, and the experience of shadow attacks increase the predictability of an upcoming bout. The experience of mild symptoms and single shadow attacks in remission periods are very common and suggest an underlying pathophysiology that goes beyond the CH bouts but is not always able to escalate to a regular bout.
Supplemental Material
sj-pdf-1-cep-10.1177_0333102421989255 - Supplemental material for Prevalence of pre-cluster symptoms in episodic cluster headache: Is it possible to predict an upcoming bout?
Supplemental material, sj-pdf-1-cep-10.1177_0333102421989255 for Prevalence of pre-cluster symptoms in episodic cluster headache: Is it possible to predict an upcoming bout? by Adam Sebastian Pedersen, Agneta Snoer, Mads Barloese, Anja Petersen and Rigmor Højland Jensen in Cephalalgia
Footnotes
Acknowledgements
The authors would like to thank the study participants and the staff at the Danish Headache Center, especially study nurse Mette Bisgaard for helping with recruitment of patients.
Declaration of conflicting interests
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Rigmor Højland Jensen has been principal investigator in studies sponsored by Eli Lilly and ATI and has given lectures for Novartis, Lundbeck, Allergan, TEVA, and ATI. Mads Barloese has been a speaker for ATI.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by grants from Tryg Foundation, Lundbeck Foundation and research foundation of Rigshospitalet.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
