The objective assessment of sleep in cluster headache: State of the art and future directions

Several lines of evidence suggest that cluster headache is related to chronobiology and sleep. Nevertheless, the nature of such a relationship is unclear. In this view, the objective evaluation of sleep in cluster headache has strong theoretical and clinical relevance. Here, we provide an in‐depth narrative review of the literature on objective sleep assessment in cluster headache. We found that only a small number of studies (N = 12) focused on this topic. The key research aims were directed to assess: (a) the relationship between cluster headache and sleep breathing disorders; (b) the temporal relationship between sleep stages/events and cluster headache attacks; (c) sleep macrostructure in patients with cluster headache. No studies considered sleep microstructure. The reviewed studies are heterogeneous, conducted by a few research groups, and often characterised by relevant methodological flaws. Results are substantially inconclusive considering the main hypothesis. We outline several methodological points that should be considered for future research, and suggest that evaluating sleep microstructure, local sleep electrophysiology and actigraphic measures may strongly increase knowledge on the relationship between sleep and cluster headache.


| INTRODUCTION
Cluster headache (CH) is the most common trigeminal autonomic cephalalgia, affecting up to 0.1% of the population (Fischera et al., 2008).This primary headache is characterised by recurrent attacks (15-180 min) of severe unilateral pain in the orbito-temporal area associated with ipsilateral autonomic symptoms.Attacks occur most frequently at specific times of the year, occurring in "cluster" or "bouts" that last weeks or months; their frequency can vary from once every 2 days to eight times per day.The International Classification of Headache Disorders, 3rd edition (ICHD-3), discriminates an episodic form of the disorder (eCH), with bouts from 7 days to 1 year separated by pain-free periods ≥ 3 months, and a chronic form (cCH), with bouts lasting 1 year or longer without remission or with out-of-bout periods lasting less than 3 months (Olesen, 2018).Such disorder represents a heavy burden for patients (Jensen et al., 2007) with relevant socio-economic costs (Gaul et al., 2011;Hoffmann & May, 2018;Olesen et al., 2012;Petersen et al., 2022).
The CH attacks occur regularly with circadian and seasonal rhythmicity (Barloese, Lund, et al., 2015;Rozen & Fishman, 2012), pointing to the existence of a link between chronobiological processes and CH.The relevance of the hypothalamus (where the main biological clock of the brain is located) in CH pathophysiology is suggested by clinical, neuroendocrinological and neuroimaging evidence (Ferraro et al., 2019;Silvestro et al., 2022).Crucially, sleep seems to have a role in CH (Barloese, 2021;Pergolizzi Jr et al., 2020;Pilati et al., 2023): up to 80% of patients report nocturnal sleep as a trigger for CH attacks (Barloese, Lund, et al., 2015), and nocturnal attacks are more severe than daytime ones (Hagedorn et al., 2019).Moreover, patients exhibit reduced subjective sleep quality both during and outside the bout (Barloese, Lund, et al., 2015), and frequent insomnia (Ofte et al., 2013;Sahota & Dexter, 1993).Nevertheless, the specific nature of the relationship between sleep and CH is unclear.
First, starting from the observation that CH attacks are more frequent during sleep, it is essential to understand if specific sleep-related processes or events represent predictors and/or triggers for CH attacks.Second, a clear description of sleep features in different conditions (e.g.nights with and without headache) and phases of the disorder (e.g. in bout and in remission) can help to determine if altered sleep represents a "stable" feature of patients with CH or a specific correlate of in-bout periods.Finally, it is important to clarify if sleep in CH is characterised by specific disorders or alterations, and whether they are merely correlates/ consequences of CH or have a role in its pathophysiology.Until now, several hypotheses have been proposed concerning: (a) a possible role of rapid eye movement (REM) sleep as a trigger of CH attacks; (b) an association between sleep disorder breathing and CH; (c) reduced arousal in CH as a marker of altered sleep in patients with CH (Barloese, 2021;Pergolizzi Jr et al., 2020;Pilati et al., 2023).Clearly, the objective evaluation of sleep in CH is necessary to answer these questions, but only a few studies provided this kind of assessment.
The present paper is aimed to provide a comprehensive narrative review of the state of the art on the objective evaluation of sleep in CH.Specifically, we will assess in depth the questions that guided research in this field, highlighting the methodological features and limitations of the available studies and the main results obtained.We will discuss the actual support of the present results to the proposed theories on the relationship between sleep and CH, also considering methodological facets.Finally, we will outline future directions for the research in this field, considering a theoretical and methodological framework provided by the current knowledge on sleep regulation processes and functions.

| Inclusion/exclusion criteria
Books, abstracts, comments, reviews, meta-analyses, pre-prints, letters to editor, and single case studies were excluded.Instances and disagreements concerning study eligibility were resolved through careful consultation and discussion.The reference lists of the selected articles were further reviewed for other potential papers.If a reference included some of the search terms in the title, it was considered as a potential paper of interest and checked for the inclusion/exclusion criteria.
Considering the phase of the pathology in which patients were recorded, two studies did not clarify if recordings were performed during or outside the active cycle of the headache disorder (Kudrow et al., 1984;Nobre et al., 2005).In the study from Chervin et al.Kudrow et al., 1984;Zaremba et al., 2012) to 8 (Terzaghi et al., 2010), but one study did not perform EEG monitoring (Evers et al., 2014) and another one did not specify the EEG montage (Nobre et al., 2003).
F I G U R E 1 Flowchart of the literature search and study selection.

| Relationship between CH and sleep breathing disorders
Several observational studies suggested that sleep apnea was common in eCH attacks.In particular, an early uncontrolled work conducted on a small sample found that six out of 10 patients with eCH had sleep apnea, and most recorded nocturnal attacks were preceded by oxyhaemoglobin desaturation (Kudrow et al., 1984).Similarly, Chervin et al. (2000) found a high frequency (80%) of obstructive sleep apnea (OSA) in their sample of 25 patients with CH.Moreover, patients with active CH had greater maximum ends-tidal carbon dioxide levels compared with inactive CH, and more severe oxygen desaturation was related to a typical occurrence of attacks reported in the first half of the nocturnal sleep period (Chervin et al., 2000).In their  Barloese, Jennum, et al. (2015) found no difference in the prevalence of sleep apnea between patients with CH and HC.Moreover, the authors report the absence of a temporal association between apneas/desaturation events and spontaneous nocturnal attacks recorded during the study.Similarly, Lund, Snoer, Petersen, et al. (2019) found no difference in breathing parameters between patients with CH in bout, in remission, and HC, and detected CH attacks were unrelated to apnea episodes.
Overall, the results about the relationship between sleep breathing disorders and CH appear conflicting.Albeit several studies suggest the existence of such a relationship, many of them were uncontrolled, performed on small samples, or characterised by several methodological flaws (i.e.absence of information about age; absence of exclusion criteria).The studies with the higher quality level found no relationship between sleep apnea and CH (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019), but they were performed by a single Danish research group and need replication.

| Temporal relationship between CH attacks and sleep stages and events
In their descriptive observational study, Kudrow et al. (1984)  In summary, the present literature does not support the existence of a temporal relationship between specific sleep stages/events and the emergence of CH attacks.

| Sleep macrostructure in patients with CH
A first description of sleep macrostructure in patients with CH showed that headache nights were characterised by reduced frequency and length of REM periods, reduced sleep efficiency, and a higher number of awakenings and sleep stage shifts (Pfaffenrath et al., 1986).showed lower total wake time and arousal index and higher sleep efficiency during nights without headache, lower sleep latency in nights with headache, and overall lower total wake time and stage 4 sleep percentage (Zaremba et al., 2012).Nevertheless, these results have been not replicated.
In a first attempt to compare sleep architecture in patients with CH and HC, the authors found increased stage 1 sleep in patients with CH (Nobre et al., 2003).In a larger group, Barloese, Jennum, et al. (2015) showed that, compared with HC, patients with CH were characterised by lower REM density and latency, reduced sleep efficiency, longer sleep latency, and fewer arousals.Finally, a recent study from the same group compared for the first time sleep macrostructure in CH in bout, in remission, and HC (Lund, Snoer, Petersen, et al., 2019).No difference in sleep macrostructure and sleep events was found between PSG nights of patients in bout and in remission.
Compared with HC, patients in bout showed longer sleep latency and REM sleep latency, lower sleep efficiency and lower REM sleep density, while patients in remission exhibited only longer sleep latency.Among patients in bout, the authors found no difference between nights with and without attacks (Lund, Snoer, Petersen, et al., 2019).
Taken together, the literature on sleep macrostructure in patients with CH is quite heterogeneous.Only five studies assessed this topic, with different specific aims: comparison between nights with and without headache (Pfaffenrath et al., 1986;Zaremba et al., 2012); comparison between patients with eCH and cCH (Zaremba et al., 2012); comparison between patients with CH and HC (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019;Nobre et al., 2003); and comparison between patients in bout and in remission (Lund, Snoer, Petersen, et al., 2019).The most consistent findings provided by a couple of studies from a single research group point to longer sleep latency, reduced sleep efficiency and lower REM density in CH compared with HC (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019).

| Actigraphic findings
Only one study collected actigraphic sleep measures in eCH and HC during a continuative period (Lund, Snoer, Jennum, et al., 2019).The authors recorded actigraphic measures for 2 weeks in 23 patients with eCH and 15 HC.Among eCH, 17 eCH actograms were recorded in bout and 11 eCH actograms in remission.Five patients participated in both disease phases.Results showed that patients with eCH in bout had a longer time in bed and TST compared with HC, while actigraphic measures were comparable.On the other hand, no difference was observed between HC and patients in remission, between eCH in the two disease phases, and between nights with and without attacks.
The authors also described that, in five out of the 10 patients that

| DISCUSSION
The number of studies focused on the objective sleep assessment in CH is small, including 12 papers in a period of 33 years  if we exclude single case reports.Despite the scientific interest on this topic (Barloese, 2021;Pergolizzi Jr et al., 2020;Pilati et al., 2023), no studies have been performed during the last 3 years.The reviewed literature is heterogeneous, characterised by several methodological issues, and conducted by a few research groups.Taken together, these points most likely highlight the intrinsic difficulties of performing research studies on objective sleep measures in a rare condition like CH (Barloese, 2021).Consistently, the main hypotheses proposed until now to describe the relationship between sleep and CH are not supported by striking evidence at present.
One frequently discussed hypothesis in this field concerns the existence of a causal relationship between OSAS and CH.Indeed, early (mainly uncontrolled) studies found a large prevalence of OSAS in patients with CH (Chervin et al., 2000;Graff-Radford & Newman, 2004;Kudrow et al., 1984).More recent studies that compared patients with CH with HC found conflicting results about the prevalence of sleep breathing disorders in this population (Barloese, Jennum, et al., 2015;Evers et al., 2014;Lund, Snoer, Petersen, et al., 2019;Nobre et al., 2003Nobre et al., , 2005)).Two well-conducted Danish studies found no temporal relationship between CH attacks and apnea episodes or desaturation events (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019).Therefore, the literature on this issue is inconclusive.It has been proposed that sleep apnea and CH are not causally associated, but may represent parallel hypothalamus-related processes (Graff-Radford & Teruel, 2009).
Another theory claims a role of REM sleep as a trigger of CH attacks, suggested by the observed trend of arising attacks about 90 min after sleep onset (Barloese, Lund, et al., 2015;Manzoni et al., 1981Manzoni et al., , 1983)), which should coincide with the first REM episode of the night.However, the reviewed literature does not support the notion of a specific temporal relationship between REM sleep (or any other sleep stages/events) and CH attacks.Sleep fragmentation has been recently reported before the attacks (Lund, Snoer, Petersen, et al., 2019), but it has been interpreted as a secondary phenomenon caused by pain instead of an attack trigger.An intriguing observation came by the recent 2-week assessment performed by the only actigraphic study reviewed (Lund, Snoer, Jennum, et al., 2019): in half of the patients in which nocturnal attacks were recorded, CH attacks and awakenings were detected around the same hour several nights in a row, with an individual timing for each patient not related to previous sleep duration.On the other hand, this pattern of stable timing for nocturnal awakening has not been observed in patients in remission and HC (Lund, Snoer, Jennum, et al., 2019).If confirmed, this finding may suggest that nocturnal CH attacks are not triggered by specific sleep macrostructural features per se but can be explained by involvement of the circadian system.In particular, the possibility that the nocturnal recurrence of CH is due to an intrinsic circadian rhythm independent of the sleep-wake cycle cannot be excluded.Further studies, and actigraphic studies in particular, are needed to clarify this point.
The most consistent findings have been recently reported by two studies from a Danish group, and pointed to signs of disrupted sleep in patients with CH in bout compared with HC, including longer sleep latency, reduced sleep efficiency and lower REM density (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019), with lower differences between HC and patients in remission and no differences between patients in bout and in remission (Lund, Snoer, Petersen, et al., 2019).At present, no conclusion can be drawn as these results need adequate replication.
Starting from these findings, it is clear that further effort is needed to elucidate the relationship between sleep and CH.Beyond increasing the number of studies in this field providing the necessary replication, several methodological and theoretical points should be considered.As previously observed, some early uncontrolled studies have been performed on small and heterogenous samples without a control group.
On the other hand, the most recent studies provided evidence on larger samples with narrower exclusion criteria and a comparable HC group (Barloese, Jennum, et al., 2015;Lund, Snoer, Petersen, et al., 2019).
Albeit the recruitment of patients with rare disorders like CH for objective sleep studies is challenging, the research in this field should continue in this last direction.On one hand, the exclusion of patients with medical, neurological and psychiatric conditions that may affect sleep is crucial to understand the actual relationship between sleep and CH.On the other hand, the possible interplay between CH, sleep problems and the frequently associated psychiatric disorders (Robbins, 2013) (Gorgoni et al., 2023).In this view, there are strong theoretical reasons to promote their assessment in the research field on CH.Several studies found alterations of the cyclic alternating pattern (CAP), a microstructural measure of sleep instability (Parrino et al., 2012), in patients with different migraine disorders (Della Marca, Vollono, Rubino, Di Trapani, et al., 2006;Nayak et al., 2016;Zhou et al., 2023).The assessment of CAP in CH may help to understand the relationship between nocturnal attacks and the arousal modulation system in these patients.Moreover, according to a growing body of in vitro, animal and human studies in the theoretical framework of the local sleep theory (Krueger et al., 2019) (Ferrara & De Gennaro, 2011).Local sleep EEG peculiarities have been observed in several neurological, psychiatric and sleep disorders (Castelnovo et al., 2018;D'Rozario et al., 2017;Gibbs et al., 2016;Gorgoni et al., 2020;Gorgoni & Galbiati, 2022;Kaskie & Ferrarelli, 2020;Mander, 2020;Poryazova et al., 2015).Moreover, the role of local sleep electrophysiology in rehabilitation and functional recovery has been highlighted (Gorgoni et al., 2013), and regional sleep features are under evaluation as possible therapeutic targets for neurodegenerative processes (Cordone et al., 2021).Starting from these notions, the assessment of local sleep in CH could represent a relevant scientific and clinical research field.Considering the role of regional electrophysiology in sleep protection, arousal and homeostasis, a specific local EEG pattern may trigger, or at least be predictive, of the CH attacks.Moreover, bearing in mind the fingerprint-like nature of the topographic EEG distribution during sleep (De Gennaro et al., 2005) and the genetic regulation of many electrophysiological sleep features (Adamczyk et al., 2015;Ambrosius et al., 2008;De Gennaro et al., 2008;Gorgoni et al., 2019;Markovic et al., 2018), it would be interesting to understand if patients with CH, and specifically those with prevalent nocturnal attacks, are characterised by a distinctive sleep EEG profile.Also, the assessment of the main sleep EEG hallmarks may allow a more in-depth understanding of the sleep modulation between different phases of the pathology, helping to determine if sleep alterations in this population are stable features (e.g.observable in both bout and remission periods) or strictly link to cluster periods and/or headache nights.
Actigraphy may also represent a useful tool to better understand the role of sleep in CH.Indeed, it allows a continuative, ecological, noninvasive and cost-effective evaluation of sleep and circadian rhythms for long periods.Actigraphic recordings have been performed to assess the relationship between sleep and several types of headaches (Bruni et al., 2004;Bursztein et al., 2006;Kikuchi et al., 2007Kikuchi et al., , 2011;;Smitherman et al., 2016).At present, with the exception of a single case report (Della Marca, Vollono, Rubino, Capuano, et al., 2006), only one recent study used actigraphic recordings to assess sleep in CH (Lund, Snoer, Jennum, et al., 2019), showing longer time in bed and TST in patients with eCH in bout compared with HC, interpreted as the effect of greater tiredness in patients, and absence of differences between HC and patients in remission, between eCH in bout and in remission, and between nights with and without attacks.Results are promising, but they clearly need replication.Moreover, the observation that the recorded CH attacks and awakenings were often observed around the same hour several nights in a row without a relationship to previous sleep duration suggests the need of a systematic assessment of this phenomenon through continuative actigraphic recordings.Indeed, as previously observed, it may point to an intrinsic circadian rhythm underlying the nocturnal CH recurrence, without a specific role of the sleep-wake cycle.A longitudinal evaluation of in bout and remission periods should be provided, and the possible relationship between sleep alterations and the so-called "ghost attacks" should be clarified (Giuliani et al., 2023).Finally, to better characterise the specific sleep and circadian rhythms features in CH, a comparison with other types of primary headache disorders should be considered.
Overall, the objective sleep assessment in CH is challenging: the reviewed literature is limited and characterised by several methodological issues.Nevertheless, a better understanding of the role of sleep in CH has strong scientific and clinical relevance (Barloese, 2021;Pilati et al., 2023;Pergolizzi Jr et al., 2020).We suggest that the assessment of sleep microstructure, local sleep electrophysiology and actigraphic measures have the potential to strongly increase our knowledge on the relationship between sleep and CH.
Two authors (MG, GG) screened titles, abstracts and keywords independently to meet the following criteria: (a) English language; (b) peer-reviewed paper; (c) empirical study; (d) objective assessment of sleep features; who performed two consecutive sleep recordings, Zaremba et al. (2012) who performed four consecutive PSG recordings, and Evers et al. (

(
2000), only eight subjects out of 25 were recorded during the cluster period.Lund, Snoer, Petersen, et al. (2019) compared patients with CH in bout with patients with CH in remission and HC.Only the MESAM study compared PSG measures collected during and outside the episode in the same patients with eCH(Evers et al., 2014).In the actigraphic study, 17 eCH actograms were collected in bout and 11 in remission(Lund, Snoer, Jennum,   et al., 2019).All PSG studies described the sensors used for the recordings with only one exception (Graff-Radford & Newman, 2004).Concerning the electroencephalogram (EEG) montage, the scalp derivation goes from 2 ( Zaremba et al. (2012) recorded four consecutive PSG nights in a small group of patients with CH (N = 5; 2 eCH and 3 cCH), showing that nights with headache were characterised by lower total wake time and higher stage 1 sleep and sleep efficiency compared with nights without headache.Considering only cCH, nights with headache showed lower total wake time and Stage 3 sleep amount and higher sleep efficiency compared with nights without headache.Moreover, eCH showed lower total wake time, TST, stage 2 sleep percentage, and higher stage 1 sleep percentage, arousal index and sleep efficiency during nights with headache.Compared with eCH, cCH exhibited nocturnal attacks during the study, CH attacks/awakenings were detected around the same hour several nights in a row.The time was individual for each patient and appeared unrelated to previous sleep duration.Patients in remission and HC did not exhibit awakenings at the same hour several nights in a row.
Studies focused on objective sleep assessment in CH.
T A B L E 1 Patients with CH had 8.8 times more chance of OSA than HC The risk increases to 26 in patients with BMI > 25 kg m À2 , and increases to 14.25 in patients > 40 years old The risk diminishes sharply in patients with BMI < 25 kg m À2 sleep: four during NREM sleep, one during REM sleep One patient was on verapamil 360 mg daily All patients were in active cluster cycle Horizontal bipolar EOG montage Three NREM sleep-related episodes arose during stable stage 2 sleep, and the other one, towards morning, during a mixed state of unstable stage 2 sleep and wakefulness Two patients: triptans/other: O2 EKG Higher percentage of total wake time and lower stage 1 NREM T A B L E 1 (Continued) Key findings sleep and SE in nights without headache compared with those with headache in all patients EOG Higher total wake time and NREM stage 3 amount and lower SE during nights without headache than nights with headache in patients with cCH Chin EMG Higher total wake time, TST, NREM stage 2 percentage, and lower NREM stage 1 sleep percentage, arousal index and SE during nights without headache than nights with headache in eCH Thoracic and abdominal respiratory efforts (impedance plethysmography) Total wake time and arousal index were lower and SE was higher in cCH compared with eCH during nights without headache SL was lower in cCH during nights with headache Overall, total wake time and NREM stage 4 sleep percentage were lower in cCH compared with eCH Note: Main methodological features and key findings are reported.Abbreviations: BMI, body mass index; cCH, chronic cluster headache; CH, cluster headache; eCH, episodic cluster headache; EEG, electroencephalography; EKG, electrocardiography; EMG, electromyography; EOG, electrooculography; HC, healthy controls; ICHD-II, International Classification of Headache Disorders, 2nd edition; IHS, International Headache Society; LMs, limb movements; M, males; min, minutes; NREM, non-rapid eye movement; OSA, obstructive sleep apnea; OSAS, obstructive sleep apnea syndrome; PLMs, periodic limb movements; PSG, polysomnography; REM, rapid eye movement; SDB, sleep-disordered breathing; SE, sleep efficiency; SL, sleep latency; TST, total sleep time.