Abstract
Purpose of review
Sleep and cluster headache (CH) are believed to be interconnected but the precise relation to the other trigeminal autonomic cephalalgias (TACs) is uncertain and complex. A better understanding of these relations may eventually lead to a clarification of the underlying mechanisms and eventually to more effective therapeutic regimens. This review aims to evaluate the existing literature on the subject of TACs and sleep.
An association between episodic CH and distinct macrostructural sleep phases, especially the relation to rapid eye movement (REM) sleep, has been described in some older studies but could not be confirmed in other, more recent studies. Investigations into the microstructure of sleep in these patients are lacking. Only a few case reports exist on the relation between sleep and other TACs.
Summary
Recent studies do not find an association between CH and REM sleep. One older study suggests chronic paroxysmal hemicranias may be locked to REM sleep but otherwise the relation is unknown. Reports indicate that CH and obstructive sleep apnoea are associated in some individuals but results are diverging. Single cases show improvement of CH upon treatment of sleep apnoea, but the causal relationship remains in question. Other TACs are probably not connected to sleep and strictly nocturnal attacks should prompt investigations for secondary causes. The relation between CH and sleep is, however, fascinating and detailed sleep studies in carefully diagnosed patients are warranted.
Introduction
Clinically, sleep and headache are very closely related but the nature of this relation is complex and multidirectional, as sleep can both serve as a trigger and cure for headache, a headache can be the first symptom of a sleep disorder, and lastly headache and sleep disorders can represent a common etiological pathway.
Cluster headache (CH) is the most prevalent of the group of primary headache syndromes known as trigeminal autonomic cephalalgias (TACs), characterised by strictly unilateral pain with accompanying autonomic symptoms. TACs include CH, paroxysmal hemicrania (PH) and short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing (SUNCT). These TACs are mainly separated by the duration and frequency of the pain attacks and by their diurnal rhythmicity. Of these syndromes, CH occurs in about 0.1% of the population, with a male to female ratio of around 4 (1) and is also by far the most studied, also in relation to sleep.
As mentioned, sleep has diverging properties which make it both a precipitating and an alleviating factor for headache. The International Classification of Sleep Disorders classifies hypnic headache, CH, chronic paroxysmal hemicrania (CPH) and migraine as so-called sleep-related headaches (2). A distinguishing feature of CH from other primary headaches is the circannual and especially the circadian rhythm with which the clusters and individual attacks occur. A high prevalence of nocturnal sleep-related attacks is reported in patients suffering both from chronic CH (CCH) and episodic CH (ECH) (3) in which up to 75% of attacks begin between 21:00 and 10:00 hours (4). The attacks may occur at all times of the day although there is a close association with sleep and times of physical relaxation (4). In many patients attacks also occur at specific time points night after night, which has led to the hypothesis of a specific relation to certain sleep stages. The classical clinical picture shows the cluster attack beginning 90 minutes after sleep onset, typically believed to coincide with the first rapid eye movement (REM) phase (5), and there is also some evidence that especially REM sleep may precipitate CH attacks. The stereotypical attacks often occur at predictable times of the day and clusters in specific circannual rhythms (4–8) but yet there are no large series of consecutive CH patients. The hypothalamus is highly involved in the generation of circannual and circadian rhythms, and neuroimaging shows hypothalamic activation during CH attacks (9–11), so there may be an underlying common neuro-anatomical pathway for TACs and sleep.
Overview of published results on TACs and sleep.
TACs: trigeminal autonomic cephalalgias; CH: cluster headache; SUNCT: short-lasting unilateral neuralgiform headache attacks with conjunctival injection and tearing; PSG: polysomnography; DBS: deep brain stimulation; ECH: episodic cluster headache; CCH: chronic cluster headache; REM: rapid eye movement; SDB: sleep disordered breathing; NREM: non-rapid eye movement; EEG: electroencephalogram; AHI: Apnoea Hypopnea Index; BcT: bolus clearance time; OSA: obstructive sleep apnoea; CPAP: continuous positive airway pressure; ECG: electrocardiogram; BiPAP: bilevel; positive airway pressure.
Of the other TACs, and in contrast to our clinical experience, CPH may be nocturnal (12), and was found to be strongly associated with REM sleep in a single and old study (13). In SUNCT, nocturnal attacks have been reported in some patients but there are no studies of this available (14).
Factors known to precipitate clusters of CH attacks but probably not other TACs include histamine, nitro-glycerine, alcohol, sleep, environmental alterations and changes in the level of physical, emotional or mental activity (15). Subjectively the structure of sleep affects CH patients (16,17). During bouts the patients report reduced total duration of nocturnal sleep, increased time to fall asleep, high frequency of nocturnal awakenings and poor sleep quality. More objectively there are reports of a reduced total duration of sleep, increased time to fall asleep, increased latent periods of all sleep stages, increased percentage representation of superficial sleep stages (1 and 2), reduced durations of REM stage, increased number of awakening episodes during the night and increased movement activity (16).
CH attacks show periodicity
Reflecting the chronobiological nature of the headache, CH attacks often occur with remarkable cyclicity. Even though this is not mandatory in the present diagnostic International Headache Society (IHS) criteria (18), where CH is dichotomised into the episodic and chronic form based on the duration of attack freedom, the annual clustering is still a key feature which most sufferers can recognise (7). Table 1 summarises research conducted in TAC and CH rhythmicity and their association with sleep.
Rhythmicity in CH is well recognised and Manzoni et al. were among the first to describe cyclicity in CH (8). Focusing on the daily occurrence, they found that onsets of the attacks peaked around 1 and 2 a.m. and 1, 3 and 9 p.m. (1 and 3 p.m. incidentally being the time when most of the patients stopped for lunch break). Some of the first papers published on CH also describe that CH exhibits a relationship with the sleep-wake cycle and there is evidence that attacks are influenced both by activity and relaxation (19–21). In a study of 77 spontaneous attacks, Russell found that 75% of these began between 9 p.m. and 10 a.m., indicating a relation to relaxation whereas the relation to sleep was more uncertain. Of the daytime attacks, 71% occurred when the patients were physically relaxed (a few while napping) (4). In the other TACs there are no such reports of clear diurnal relations.
Are TACs associated with REM sleep?
CH
The specific and precise rhythmicity with nightly attacks occurring one to 1.5 hours after falling asleep has led to the hypothesis of a relationship between CH and REM sleep, i.e. that REM sleep triggers CH which was confirmed in early sleep studies (22). More than 40 years ago Dexter and Weitzman (5) (Table 1) showed that in three patients with ECH all nine of the recorded nocturnal attacks occurred during REM sleep or within nine minutes of this sleep stage terminating. Further strength is added to this observation by the finding that many attacks occur between 4 and 7 a.m., the time when REM sleep is increased (4). In one series of recorded attacks, almost 60% of CH attacks followed REM sleep, even though REM sleep comprised only 20% of total sleeping time (23). Two recent studies of limited size (seven and five), failed, however, to find any temporal association between ECH attacks and REM sleep (24,25). Consequently, there is still debate on the association of CH attacks and REM sleep, reflecting the difficulty in studying sleep and periodic pain disorders as CH.
Whether nocturnal CH attacks co-occur or are triggered by REM sleep is uncertain; however, it seems that REM sleep is somehow affected in CH. In one CH patient in active bout, the PSG showed altered sleep architecture, including prolonged REM latency and increased REM fragmentation (6). The reasons for these observations are obscure and underline the complexity in CH and perhaps headache pathophysiology in general, but the findings may indicate involvement of the hypothalamus (26).
It is reported in some cases that sleep deprivation can temporarily prevent or delay the onset of CH (27) but whether sleep deprivation can be used systematically in the management of CH is not clarified. In our clinic, we often meet CH patients who describe hypnophobia, a real fear of going to bed – the patients pursue their daily activities and sports just to avoid or postpone sleep and thereby their nocturnal attacks. This insomnia seems to be solely associated with the bout (17).
It has been proposed that switching between different states of relative arousal may precipitate the attack, specifically the transfer of dominance of the parasympathetic cholinergic system (mediators of REM sleep) to a primary sympathetic serotonergic system (mediators of non-REM sleep) (28,29). It has also been suggested that CH attacks comprise a condition of hypoarousability based on a case report of a single patient by the use of actigraphy (a method for measuring gross motor activity) (6) Hypoarousability is a possible factor in the pathogenesis of migraine (30,31) but the role in CH is still unclear. The previously mentioned implication of the posterior hypothalamus during CH attacks is especially interesting in this context. This central nervous system (CNS) structure houses cells producing hypocretin, which plays a significant role in pain and arousal mechanisms (32).
In CPH only case reports on the relation with sleep are available and the results are inconsistent. However, in one review the authors go so far as to call it a
In SUNCT as in the very similar trigeminal neuralgia (TN) the majority of attacks are diurnal although the latter lacks the ictal autonomic symptoms. Nocturnal attacks of SUNCT or TN should always raise suspicion of a symptomatic lesion and we recommend that brain magnetic resonance imaging (MRI) scans are conducted – mainly in the search for pituitary adenomas, either growth hormone-producing tumours, as described by Rozen et al. (34), or prolactin (PRL)-producing tumours inducing strict nocturnal attacks in parallel with increased nocturnal prolactin levels, as described by Bosco et al. (35). It has also been suggested that there is a close temporal relationship between REM sleep and high serum PRL levels but there are no data to support this hypothesis yet. However, the neurohumoral interface between sleep and headache is another fascinating field which deserves much more attention.
CH and the hypothalamus
The current knowledge on the function of the hypothalamus in both CH and sleep points to an unknown, underlying pathway for trigeminal pain perception and sleep (and REM) regulation (26,36). With this background we also aimed to evaluate the literature on the subject of CH, other TACs, pain and sleep.
Body homeostasis is in part regulated by the secretion of several hormones in a cyclic fashion, and CH patients have been investigated in detail for specific variations (37). These findings include melatonin, cortisol, testosterone, luteinising hormone, follicle-stimulating hormone, prolactin, growth hormone, thyrotropin and beta-lipotropin. In CH, some of these change only during the bout, returning to normal once this ends, suggesting a periodic and reversible hypothalamic dysfunction (37). Furthermore, studies in functional neuroimaging (positron-emission tomography (PET) and functional MR) report neuronal activation in the ipsilateral posterior inferior hypothalamic grey during the pain state (9–11). Neurons secreting hypocretin are located almost exclusively in the posterior inferior hypothalamus and project widely to CNS structures involved in pain processing, sleep and arousal (38–41). Several studies have found a polymorphism in the hypocretin-2 receptor (1246G > A) to be associated with cluster headache (42–44) but the clinical significance of this finding is yet unclear. Interestingly, the pathogenesis of narcolepsy-cataplexy revolves around a sometimes total deficiency in the hypocretin system (45) and sufferers exhibit more migraine over time (46,47). TACs are not systematically reported in patients with narcolepsy although one case report is available (48). A recent study of the hypocretin concentration in the spinal fluid of ECH patients did not report abnormal findings (49) but subsequent larger studies are underway.
Sleep-disordered breathing (SDB)
The association between CH and SDB has been studied for several decades but many questions remain unanswered. The most common form of SDB is obstructive sleep apnoea (OSA). In one study approximately 4% of middle-aged males were estimated to have OSA with associated daytime hypersomnolence and up to 24% of males may have sleep apnoea (defined as Apnoea Hypopnea Index (AHI) ≥ 5) (50). Although OSA seems to be a very common finding, there may be a positive association with CH as the OSA prevalence in these patients may be as high as six in 10 (23). Another study found that CH patients are 8.4 times more likely to exhibit OSA than normal individuals (51) but it is unclear whether results are controlled for lifestyle factors relevant for SDB such as obesity and/or smoking (particularly relevant in CH, as these patients frequently are smokers).
The tendency for attacks of CH to occur during the night time has led to speculation that the changes invoked by SDB, in particular OSA, might somehow trigger the cluster attack. OSA may lead to recurrent hypoxemia, hypercapnia, excessive negative intrathoracic pressure and increased intracranial pressure, amongst other findings. It may also be accompanied by abrupt changes in autonomic tone. All of these changes could theoretically serve as triggers of attacks. Further strengthening the association between the two disorders is the fact that OSA severity is higher during REM sleep (52), the sleep stage which may be associated with an increased likelihood of CH attacks.
More than two decades ago Kudrow et al. suggested hypoxemia to be a potential trigger of CH attacks (23). Hypoxemia has been observed prior to CH attacks in wakefulness and if induced by nitro-glycerine it can trigger attacks if the patient is in an active bout (53). The fact that inhaling supplementary oxygen is a very effective treatment for an acute attack also lends support to this theory (54). However, in one study induced hypoxemia failed to provoke attacks at all (55) but large series are lacking and there is probably also great variability within individual CH patients and especially between CH patients. With nasal continuous positive airway pressure (CPAP), there is now very effective treatment available for OSA and the accompanying hypoxemia, and case reports are available where effective treatment of OSA resulted in reduced severity and frequency of otherwise refractory CH attacks (56–58) but obviously larger controlled series are needed to confirm the observations.
One could speculate that there may be an overlap between the underlying factors that lead to OSA and CH. However, it seems clear that in some patients OSA can affect the severity of CH, as evidenced by the case reports mentioned above, and maybe also trigger a latent CH in susceptible individuals.
Deep brain stimulation (DBS) affects sleep in CH and PH patients
Two small studies on the effect on sleep of DBS of the posterior hypothalamus, a fairly invasive therapy for treating refractory CCH, has been conducted recently (59,60). The first study (three patients) seems to confirm existing theories. In patients receiving DBS nocturnal CH attacks disappeared, and sleep architecture and sleep quality were normalised as PSG showed increased total sleep time, sleep efficiency, slow-wave sleep stages and decreased PSG indexes of fragmented sleep compared to the situation before DBS. The authors also found that DBS decreased periodic limb movement during sleep (PLMS). They state that this cannot be ascribed solely to decreased light sleep stages (where PLMS is more frequent) and it suggests that the areas affected by DBS also influence the relative state of arousal. In contrast with these findings, posterior DBS in two male patients with pharmacoresistant and concomitant CH and SUNCT had a dramatic effect on their sleep patterns with disrupted sleep architecture, and poor sleep efficiency after DBS so the relation is far from simple (60). Stimulation of the subthalamic nucleus in Parkinson's disease has not been reported to curtail PLMS, rather it seems to provoke restless legs syndrome (61,62) so other mechanisms must be sought.
These findings support the known fact that the posterior hypothalamus plays an important part in regulating sleep-wake states and relative arousal. Furthermore it is important in pain modulation as evidenced by its connections to certain brainstem regions (dorsal raphe nuclei, the locus coeruleus, periaqueductal grey). An H215O PET study performed in CCH patients with implanted DBS devices showed increased blood flow in the ipsilateral posterior inferior hypothalamic grey (site of the stimulator tip) and the ipsilateral trigeminal system in the brainstem (63). It is important to note that activation of the trigeminal system in this case did not provoke pain or accompanying autonomic symptoms.
Conclusion
The prospect of understanding the pathology behind TAC attacks occurring during sleep is tantalising but only CH seems to have a consistent relation to arousal and sleep. Nocturnal CPH and especially SUNCT should prompt investigations for a secondary cause – primarily pituitary pathology.
Not only could a better understanding of these complex interactions bring potentially more effective treatment, it may also offer a chance of better understanding the regulation of sleep- and wake-promoting mechanisms. A role by which the hypothalamus serves as a gating or switching mechanism has been suggested and a disorder of the hypothalamic system may result in destabilisation of pro- and antinociceptive input and the activation of the trigeminal-autonomic reflex also affecting sleep. The gradual progression from ECH to CCH that occurs in some patients may be associated with such a change resulting in a loss of association with REM sleep.
Despite the clinical similarities between the TACs, there is no specific relation between sleep and TACs other than CH. The posterior hypothalamus plays a significant role in arousal mechanisms, since it projects to the cerebral cortex. Its stimulation results in cortical activation, motor activity and sympathetic responses (the last two accompanying symptoms in CH) and impaired arousal in CH has been considered (17). However, the connection between these arousal mechanisms and the chronobiological nature of CH, including the interaction with sleep, still offers much to be elucidated. Investigation of a possible association with the hypocretin system is very challenging and should be investigated in larger groups of TAC patients. The true role of the hypothalamus in TACs is still unknown. There is still a possibility that the symptoms and findings pointing towards its involvement are simply epiphenomena but further research is needed.
Clinical implications
Only CH seems to have a consistent relationship with sleep and treatment of concurrent sleep disorders may lead to improvement of headache. Nocturnal attacks of CPH and SUNCT should raise concerns of pituitary pathology. Treatment of refractory patients with invasive therapies may both improve and worsen sleep, which should be considered before a decision is made.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of interest
None declared.