Hallmarks of primary headache: part 3 - cluster headache.

Compared to migraine, CH is a relatively uncommon condition primarily affecting males. Consequently, there is less knowledge regarding its epidemiology during pregnancy and breastfeeding. Bouts of CH are rare during pregnancy. This is attributable to various potential factors. In the reproductive years, specifically under the age of 50, the male-to-female ratio is 7.2:1 for episodic CH and 11:1 for chronic CH [ 487 ]; these ratios become lower after the age of 50. Moreover, data indicates relative hypofertility in women with CH [ 29 , 488 ], particularly when the disease begins prior to the first pregnancy [ 489 ]. Women who suffer their initial attack prior to their first pregnancy typically have fewer offspring than those who were already mothers at the time of clinical onset [ 488 , 489 ]. In a particular study, almost 60% of female patients diagnosed with CH remained childless. Of these women, 8% expressed concern that the attacks would hinder their ability to care for a child, while 4% feared transmitting the disorder [ 37 ]. Burish et al. [ 490 ] conducted a retrospective study in which CH patients were interviewed about their experiences of similar pain, in order to compare them with CH attack pain. Labour pain (childbirth), pancreatitis and nephrolithiasis were then examined; the pain experienced during CH attacks was found to have an average value of 9.7 on a numerical rating scale from 0 to 10. This was significantly higher than the score of 7.2 found for labour pain [ 490 ]. Sharaf & Ali [ 491 ] reported two patients who experienced both CH and labour pain in rapid succession. Both patients articulated their preference for labour pain over the pain associated with CH. The data are inconclusive about whether CH improves throughout pregnancy or remains stable, and most research is constrained by restricted sample sizes and retrospective methodologies. A survey of 249 individuals with CH revealed that 34 were female, of whom only eight had experienced pregnancy since the onset of CH; six of these eight reported a remission of clusters during pregnancy [ 488 ]. A retrospective questionnaire-based investigation was undertaken involving 196 women with CH, compared to migraine sufferers and healthy controls. Among the women with CH, 143 patients who had undergone at least one pregnancy experienced their initial CH attacks postpartum; only 19 of these patients indicated experiencing CH attacks during pregnancy, and the majority did not report an escalation in the severity or frequency of these attacks relative to those occurring outside of pregnancy [ 489 ]. Out of 111 patients with episodic CH who had been pregnant, 23% stated that a “predicted” cluster phase did not manifest during pregnancy. In eight cases, the cluster phase commenced within one month of delivery. Bahra et al. [ 18 ] showed that merely 5% of women with CH experienced attacks during pregnancy, and their seasonal pattern altered at conception, meaning that anticipated attacks did not occur. Manzoni and colleagues [ 29 ] reported that CH was unaffected by menstruation, pregnancy, and the postpartum period in 82 patients. Certain therapeutic recommendations and clinical studies concentrate explicitly on TACs in the context of pregnancy and lactation [ 482 , 492 – 494 ]. In the context of pregnancy and lactation, high-flow oxygen therapy at 100% administered using a non-rebreather mask, subcutaneous sumatriptan or nasal spray, and zolmitriptan nasal spray should be employed with caution and under medical supervision as the primary acute interventions [ 492 ]. These medications do not seem to elevate the risk of congenital anomalies in babies, and sumatriptan appears to be safe for use during nursing [ 482 ]. Evidence supporting the efficacy of intranasal lidocaine and transcutaneous vagal nerve stimulation is limited. In the absence of comprehensive research, verapamil and short-term oral prednisone/occipital nerve block are regarded as likely safe during pregnancy and nursing, as they do not elevate the teratogenic risk to the foetus [ 495 , 496 ]. Steroids can elicit moderate and transitory reduction of adrenocortical function in the baby only when taken during late pregnancy [ 497 ]. Particular attention must be directed towards patients administered high dosages of verapamil, as it is detectable in breast milk; however, it is improbable to induce harmful effects in breastfed newborns [ 497 ]. Lithium medication is contraindicated during pregnancy and nursing due to potential foetal and maternal problems. Generally, the same contraindications are applicable to topiramate and valproate, although they are deemed compatible with breastfeeding healthy full-term infants post-neonatal period [ 498 , 499 ]. There is no conclusive evidence endorsing the use of galcanezumab during pregnancy and lactation. Coupled with the fact that immunoglobulins can cross the placental barrier, the technical drug data sheet states that galcanezumab administration is contraindicated during pregnancy [ 500 ]. In summary, CH attacks during pregnancy are infrequent; however, when they manifest, they typically retain their original characteristics and severity. Although some women may observe a modification in the seasonal rhythm of their CH during pregnancy, many do not. Options for treating CH during pregnancy are restricted due to safety concerns associated with drugs, making management of these headaches difficult.

One of the first potential drugs for CH treatment was histamine. Since the very first description by Bayard T. Horton in 1939 [ 268 , 269 ], he proposed that histamine could be a diagnostic and therapeutic aid in the syndrome, to the point that it still inadequately named “histaminic cephalalgia” by some authors [ 268 , 270 , 271 ]. Even before the recognition of “Horton syndrome” or “Histaminic cephalalgia”, in 1937 Wilfred Harris reported that subcutaneous injections of ergotamine tartrate alleviated “migrainous neuralgia” headache, nowadays considered to probably be CH patients [ 269 ]. Indeed, in 1941, Bayard T. Horton also communicated that ergotamine could alleviate “histaminic cephalalgia” [ 271 , 272 ]. Methysergide was explored as a possible migraine treatment in 1959 for the first time [ 273 ], and in 1961 Harris published the first study that evaluated the efficacy of the drug in migraine and CH prevention, showing a reduction in the frequency of CH attacks with respect with the baseline [ 274 ]. Since that first study, adverse effects have commonly been observed. Together with the occurrence of fibrotic complications, this ultimately led to the abandonment of methysergide [ 273 , 274 ]. Caffeine is a non-selective antagonist of A 1 and A 2 A adenosine receptors [ 275 ]. Within the effects of caffeine, vasoconstriction of cerebral blood vessels was thought to be an effective way to alleviate various headache disorders, including migraine, tension-type headache, and CH [ 275 ]. Evidence in CH treatment is weaker than with other drugs, and in many cases, it is used in various preparations, such as coffee, energy drinks or in combination with other drugs, such as aspirin, ergot derivates or butalbital. In a large survey published in 2019 that evaluated the employed drugs for CH treatment, 41/1604 (2.5%) participants reported to use caffeine as an acute treatment, and 17% of them reported high effectiveness of it [ 276 ]. In addition, 112/589 (19%) of patients had used combinations of ergots and caffeine too [ 276 ]. The use of the lithium salt as a remedy in CH is another example of a serendipitous finding which was first suggested by Karl Ekbom in 1974 [ 277 ]. The pharmacological effect of lithium for CH has been evaluated since the 1980´s and has indicated a good effect, especially in chronic CH, though the treatment comes with a myriad of adverse effects, and the dosage is notoriously difficult to titrate [ 278 ]. While lithium’s mechanism of action in CH is undetermined, several clues from its effect in bipolar disorder can be found. Lithium has been shown to promote inhibitory neurotransmission via facilitating the pre-synaptic release of GABA and upregulation of post-synaptic GABA A -receptors [ 279 ]. Interestingly, GABA has been noted as a dominant neurotransmitter in the hypothalamus [ 280 ]. Moreover, lithium can inhibit post-synaptic G-proteins and has downstream inhibitory effects on the adenylyl cyclase system [ 281 ]. Further, long-term lithium treatment has been shown to increase neuronal monoamine synthesis, particularly that of the indolamine 5-hydroxytryptamine (5-HT) [ 281 ]. This ties well together with the inhibitory effect 5-HT (and triptans) has on trigeminal neurons [ 282 ]. In 1975 Jammes communicated the results of the first double-blind, placebo-controlled, single crossover study that showed that prednisone decreased the CH attack frequency in 17/19 treated patients with prior failure to other therapies, such as ergotamine, methysergide, caffeine, phenobarbital and other analgesics [ 283 ]. Three years later Couch and Ziegler communicated the results of a retrospective study that showed that 14/19 CH patients treated with 10–80 mg of prednisone per day had at least a 50% relief in the headache, However, 79% of patients experienced a CH recurrence when the drug was tapered down [ 284 ]. In 1981, Lee Kudrow communicated for the first time that 100% oxygen at 7–10 L was able to abort CH attacks within minutes [ 285 ]. Subsequent studies showed that oxygen effectively aborted attacks when the inhalation is started at the onset of the attack (in 50–85% of patients [ 278 ]. While oxygen treatment has the advantage of no adverse effects, it requires patients to keep oxygen tanks at home which may be impractical. Oxygens mechanism of action has long been considered to be vasoconstriction [ 286 ], which is somewhat at odds with the current understanding of headache pathophysiology. While the mechanism requires further clarification, studies have shown that oxygen can inhibit neuropeptide release and neuronal activation in the trigeminocervical complex [ 81 , 287 ]. The first scientific evidence supporting the use of verapamil in CH was published in 1983 by Meyer and Hardenberg, in a series of five patients that received it [ 288 ]. Verapamil binds to a relatively large range of voltage-gated calcium channels (CaV). Among them Cav1.2, a subunit of the ‘long-lasting’ L-type calcium channels, has been noted as important as it is expressed by trigeminal neurons [ 289 , 290 ]. Further, verapamil is reported to inhibit potassium channels. Verapamil can easily cross the blood-brain barrier (BBB) and elicit a central effect. Conversely, the recommended doses for verapamil (160–720 mg/daily) are quite high, likely due to the fact that verapamil is a substrate for p-glycoprotein (an efflux protein situated at the BBB) [ 291 ]. Verapamil’s mechanism of action in CH is not yet determined. However, it has been suggested to modify CGRP-release via blocking of presynaptic calcium channels [ 290 ], and to modify circadian rhythm via interactions with calcium channels in the hypothalamus [ 288 ]. Another possible mode of action could be the prolongation of the action potential, elicited by calcium channel blockers (CCBs), reduces the frequency of neuron ‘firing’ across the trigeminal system. Interestingly, verapamil has been shown to dose-dependently inhibit amplitude and frequency of epileptic after-discharge, an estimate of local excitability, in male rats [ 292 ]. Thus, speculatively, verapamil could reduce the amount of pain signals sent from the trigeminal system to higher order pain centres. In addition, the initial hypotheses supported that CCBs could be effective in migraine and CH due to the vascular changes observed during headache episodes. It should be noted that though most available data on CCBs effect on CH comes from verapamil, other voltage-gated CCBs such as nifedipine (30–180 mg/daily) and nimodipine (60–120 mg/daily) have also demonstrated a high efficacy in the prophylactic treatment of CH [ 288 ]. Which can both cross the BBB to some extent. The first preliminary evidence of sumatriptan in CH acute treatment was published in 1989 [ 293 ], subsequently confirmed in a randomized, double-blind, placebo-controlled crossover study that included 49 CH patients, published in 1991 [ 294 ]. The rationale for their use in CH was also linked to their vasoconstrictive properties (via agonism of 5-HT 1B receptors) when migraine was considered to be of vascular origin. Contemporarily, triptans have been shown to inhibit the release of neuropeptides (e.g. CGRP) from trigeminal neurons (which express 5-HT 1B/D receptors) [ 295 , 296 ] and can inhibit neuronal firing in the trigeminal ganglion (TG) of rats [ 297 ]. Therefore, triptans likely have a triple effect on pain modulation relevant to both CH and migraine; (1) inhibition of neuropeptide release, (2) inhibition of neuronal firing, and (3) vasoconstriction. The balance between the 3 different mechanisms remains an enigma. Figure 3 summarizes the timeline of the first evidence of efficacy for older drugs in the treatment of cluster headaches. Fig. 3 Timeline of the first evidence of efficacy for older drugs in the treatment of cluster headaches Timeline of the first evidence of efficacy for older drugs in the treatment of cluster headaches CGRP is increased in plasma of CH patients during attacks when compared to out-of-bout periods [ 298 ]. This plausibly has a sensitizing effect on the facial sensory nerves relevant to CH and migraine. In support, sensory C-fibre projections from the maxillary nerve pass through the SPG [ 299 ]. The SPG is important to CH pathophysiology as a major efferent parasympathetic relay [ 300 ], and its neurons express CGRP receptors [ 301 ]. Thus, the release of CGRP from C-fibre boutons projected via the maxillary nerve may activate CGRP receptors expressed on SPG neurons. The CGRP release could sensitize the SPG neurons in a similar way as likely occurs in migraine pathophysiology [ 302 ]. Psilocybin and LSD have been suggested to abort CH attacks and increase remission periods between bouts [ 303 ]. While the exact mechanism of action is not elucidated for psilocybin or LSD in CH, the plausible target for them is a broad variety of 5-HT receptors in both the CNS and PNS. LSD is readily able to cross the BBB, while psilocybin first requires enzymatic dephosphorylation into its active, and more lipophilic, form psilocin to cross [ 304 ]. Ketamine is a NMDA receptor antagonist that inhibits glutamate-mediated excitatory neurotransmission. This may decrease periaqueductal gray substance, dorsal raphe nucleus and locus coeruleus activation, also decreasing nociceptive signals to the trigeminal nucleus caudalis and ultimately reducing the CGRP release [ 305 ]. Recent GWAS studies have revealed an upregulation of the MERTK receptor in CH patients [ 306 , 307 ]. Activation of the MERTK receptor induces the MAPK/ERK1/2 pathway and is further involved in phagocytotic and neuroinflammatory processes [ 72 ]. Speculatively, MERTK has been suggested to be involved in neuron-glia interactions relevant for the hypersensitization to pain in CH [ 72 ]. Though this work is in its infancy it is an interesting and novel target for CH.

Headache disability means limitations in daily life. This includes work, school, and social time. It is a major part of the total burden. Burden also includes cost, distress, and social impact. Global Burden of Disease studies use attack rate and severity to measure it. Migraine Disability Assessment Score Questionnaire (MIDAS) and Headache Impact Test (HIT-6) are tools commonly used in clinics to measure the disability caused by migraine. The burden of CH is difficult to measure [ 552 ]. It affects many aspects of life, from daily function and mental health to financial cost and even suicide risk. There are four main challenges in assessing this burden. First, there are no tools made specifically for CH. Most studies rely on migraine tools instead. Second, the timing and pattern of CH attacks, especially in episodic form, make it hard to assess symptoms clearly. Third, many studies involve small groups, Often those with the more severe chronic form. Fourth, there is a major lack of data from low- and middle-income countries, where nearly 90% of the world’s population lives [ 10 , 553 ]. Quality of life (QoL) in CH is low. Studies used SF-36, SF-12, and EuroQoL-VAS—all general tools to assess QoL. CH patients reported worse QoL than healthy subjects and even migraine patients. Invasive treatments may reduce attacks but often do not improve QoL [ 553 – 557 ]. CH affects QoL even between attacks. Worry about future attacks adds stress. Patients avoid triggers, which limits daily life. These emotional effects add to the total burden [ 10 , 558 , 559 ]. Disability in CH is often underestimated. Studies using MIDAS and HIT-6 tools showed high scores on HIT-6 (Often over 60) and MIDAS (over 21). Even after major treatments like nerve blocks or brain stimulation, disability stays high. This shows that the real impact of CH is deeper than what current tools can capture [ 10 , 553 , 560 , 561 ]. A recent study by Göbel et al. used the WHO Disability Assessment Schedule (WHODAS 2.0) in nearly 200 CH patients. They showed that 92.7% had severe limits in daily function and major impairment [ 562 ]. The socioeconomic burden of CH is heavy. Few studies cover this aspect, but findings are clear. Many patients lose workdays, miss chances for promotion, or leave jobs entirely. Daily life is limited, even outside attack periods. Some lose their jobs due to the headache impact [ 10 , 553 ] 563 – [ 565 ]. CH leads to high healthcare use and costs. CH patients see more doctors and use more treatments than the general population. A German 6-month study showed average costs of €5,963 per patient. Costs were higher for chronic CH (€10,985) than episodic CH (€2,583), mostly from acute treatments. A Danish study reported similar findings, with annual direct costs of €9,158 for chronic CH and €2,763 for episodic CH [ 10 , 553 , 563 – 565 ]. Suicide risk in CH is high. CH is often called the ‘suicide headache’ due to the extreme pain it causes, despite large studies on actual suicide rates in this group are still lacking [ 10 ]. Anyway, a U.S. survey found that 55% of CH patients had an increased risk of suicide. Another study showed high rates of suicidal thoughts, often linked to demoralization. This highlights the urgent need for better support and care [ 37 , 452 , 566 ]. A Korean study examined suicide risk in CH. During attacks, 64% reported passive suicidal thoughts, 36% had active thoughts, 6% made plans, and 2% attempted suicide. These risks dropped significantly between attacks. This shows suicide risk is closely tied to the pain. Effective treatment of attacks may lower this risk [ 451 ]. Other conditions like personality disorders, insomnia, and substance abuse may raise suicide risk in CH.

GONB is recommended for the preventive/transitional treatment of CH by the American Headache Society evidence-based guidelines (Level A: established as effective) as well as by the recently published European Academy of Neurology guidelines [ 342 , 428 ]. A consensus on the best procedure to perform GONB is lacking. However, the injections are frequently given ipsilateral to attacks at the medial third between the inion and mastoid process [ 429 ]. For the treatment of CH, GONB is usually performed, with a combination of corticosteroid and local anaesthetics. Most commonly the use of methylprednisolone (80 mg) and lidocaine (2 ml, 2%) have been reported in the literature [ 430 ]. The mechanism of actions of these agents and the block are not fully understood. Preclinical and clinical evidence suggests that stimulation or inhibition of the greater occipital nerve modulates the neuronal activity of the TCC, a pivotal area of the trigemino-vascular system [ 431 , 432 ]. Similarly, the specific mechanism of action of steroids in CH remains unclear besides its general anti-inflammatory effect. It has been suggested that nociceptive C-fibers in substantia gelatinosa are reversibly blocked by corticosteroids [ 433 ]. Steroids may also prolong the effects of local anaesthetic [ 434 ]. GONB proved effective as a transitional treatment for CH in three RCTs [ 332 – 334 ] and several open-label observational studies [ 430 ]. Ambrosini et al. [ 333 ] assessed the efficacy of unilateral GONB performed with a mixture of rapid-acting and long-acting betamethasone plus xylocaine in 16 patients with episodic CH and 7 patients with chronic CH [ 433 ]. Four subjects were on a stable oral preventive medication at the time of enrolment; in addition, verapamil was started after 1 week from GONB in participants with persistent CH attacks. Eighty-five and 61% of CH subjects who received GONB were pain-free after 1 and 4 weeks, respectively, compared to 0% in the placebo group [ 333 ]. In the trial by Leroux et al. [ 332 ], 28 patients with episodic CH and 15 with chronic CH were randomized to receive three unilateral suboccipital injections of cortivazol (2–3 days apart) or placebo. GONB was performed as add-on treatment to verapamil [ 434 ]. The percentage of patients reporting less than two CH attacks per day within four days after the third GONB was 95% for CH subjects who received cortivazol compared to 55% of participants in the placebo group [ 434 ]. After 15 days from GONB, the number of CH attacks per day was lower in individuals who received cortivazol, although the percentage of subjects with a 50% reduction in CH attack frequency was comparable between active con placebo groups [ 332 ]. More recently, the ANODYNE study was designed in accordance with the guidelines of the International Headache Society for controlled trials in CH [ 334 , 435 ]. Thirty-nine episodic CH subjects were randomized to receive GONB (methylprednisolone plus lidocaine) or placebo. A stable oral preventive medication was allowed at baseline; in addition, verapamil was started or titrated in subjects still reporting CH attacks one week after GONB. At the end of the 4-week follow-up period, a more pronounced reduction in weekly attack frequency was reported after GONB (−7.7 attacks) compared to placebo (−3.4 attacks). In addition, a higher proportion of participants achieved pain remission after the first and second weeks after GONB. By contrast, the rate of complete remission after 3 and 4 weeks was comparable between GONB and placebo groups, with this finding possibly explained by the planned titration of verapamil, or by the occurrence of spontaneous CH remissions. At the end of the 4-week follow-up, 29 participants (12 in the active GONB group, and 17 in the placebo group) were on verapamil treatment, without difference in daily dose. The ANODYNE trial included an additional 8-week open label phase at the end of the double-blind treatment period. Noteworthy, 5 participants from the placebo group required an additional GONB during the open-label phase, while no subjects from the active GONB performed a second treatment. Few data are available regarding the duration of pain-freedom and the consistency of response across subsequent GONB [ 430 ]. In the open-label study performed by Lambru et al. [ 436 ], 83 chronic CH subjects underwent unilateral GONB with methylprednisolone and lidocaine [ 436 ]. Noteworthy, a sub-group of participants received up to four quarterly GONB procedures. After the first GONB, a positive response (complete remission, or 50% reduction in attack frequency) was described in 57% of participants, with a median duration of the therapeutic response of 21 days [ 436 ]. In addition, around 70% of patients were pain-free after the third and fourth GONB. These findings support a consistent and reproducible effect of GONB as a preventive treatment for chronic CH. Regarding safety and tolerability, several evidence suggests that GONB is a safe, well-tolerated, and minimally invasive procedure. The reported adverse events were described as mild and transient, with injection site pain being the most prevalent one. Less frequent adverse events were local numbness, local bleeding, neck stiffness, dizziness, syncope, local alopecia [ 332 – 334 , 430 , 436 ]. Besides GONB, oral prednisolone as a transitional treatment in episodic CH has been shown to be effective in an RCT [ 331 ]. In the absence of any head-to-head RCT comparing oral prednisolone with GONB, it is uncertain which one is superior, or whether both are equally effective. GONB offers the advantage of smaller dose of steroids, minimal systemic side effects and drug interactions. Contrarily, its invasiveness (though minimal) and lack of a standardized protocol, especially defining the timing and frequency of repeat injections in episodic CH and chronic CH patients are considered as disadvantages. Sphenopalatine ganglion block (SPGb) has been tested as acute and preventive treatment for cluster headache. Several technical procedures have been described, ranging from an easy-to-perform local intranasal application of anaesthetic to endoscopically guided drug injection in the pterygopalatine fossa [ 437 ]. At state of the art, SPGb is considered by available guidelines as a second-line acute treatment for CH attacks. Specifically, the American Headache Society rated SPGb with a Level C (possible effective) recommendation, while in the European Academy of Neurology guidelines it is considered only for patients with contraindications for triptans [ 342 , 428 ]. It is worth noting that these recommendations are mainly based on open label studies, case reports, or treatment of experimentally nitroglycerin triggered CH attacks [ 320 , 321 , 438 ]. SPGb has been tested as a preventive treatment for CH as well. In the study by Felisati et al. [ 439 ], 20 subjects with drug-resistant chronic CH underwent SPGb with local anaesthetic and steroid under endoscopic control. Eight subjects became pain-free, with the duration of the clinical improvement lasting from 1 to 24 months. Nasal epistaxis and transient diplopia were reported as SPGb-related adverse events. In 2010, the same group performed an uncontrolled trial in 15 drug-resistant CH individuals [ 440 ]. SPGb was performed with triamcinolone, bupivacaine, and mepivacaine with adrenaline under endoscopic control. A complete remission was achieved in 8 out 15 participants (54%), with a duration of pain-freedom ranging from 1 to 28 months. SPGb-related adverse events were two cases of severe nasal epistaxis and one case of transient reduced buccal opening [ 440 ]. A novel approach is represented by injection of onabotulinumtoxinA toward the sphenopalatine ganglion [ 441 – 443 ]. In the pilot study by Bratbak et al. [ 441 ], a single injection of 25 or 50 IU of onabotulinumtoxinA towards the SPG was tested in 10 participants with drug-resistant chronic CH. The procedure was performed via a transnasal approach aided by surgical navigation under general anaesthesia. One serious adverse event was reported, namely a severe epistaxis; in addition, three subjects suffered a transient accommodation disturbance in the ipsilateral eye, while another subject reported reduced buccal opening and masticatory strength. Regarding efficacy, a reduction in CH attack frequency per week was reported up to 6 months after treatment, with the most prominent benefit achieved by 3 months post-treatment. Seven participants were later enrolled in long-term (18 to 24 months) extension study [ 442 ]. Five patients repeated the treatment (up to 6 times during the 24-month follow-up) using a novel percutaneous infrazygomatic approach under local anaesthesia. Weekly CH attacks reduced from 14.3 ± 8.9 to 3.1 ± 3.8 and to 6.1 ± 4.8 after 18 and 24 months, respectively. Transient accommodation difficulties, jaw pain, reduced buccal opening and moderate epistaxis were reported by two participants [ 442 ]. In the open-label series by Simmonds et al. 31 drug-resistant chronic CH subjects were treated with 25 IU of onabotulinumtoxinA towards the SPG according to the previously described procedure [ 443 ]. A total of 206 procedures were considered for a safety analysis. Overall, 63 adverse events were described, with one serious adverse event reported (Hospitalization for facial asymmetry who resolved within 12 weeks). The most prevalent adverse event was intermittent visual disturbance, accounting for 23 out of 63 adverse events, followed by pain at incision site ( n  = 11) and jaw problems ( n  = 11). Two-month effectiveness data were available for 14 chronic CH subjects, for a total of 92 SPGb procedures. Bearing in mind the limitation related to the high lost to follow-up rate, SPGb with onabotulinumtoxinA induced a reduction of 11.1 in weekly CH attacks after the first treatment, with a rate of 50% responder ranging between 53 and 69% during the 4-month follow period.

Despite the established role of the hypothalamus in the circadian rhythmicity and circannual rhythmicity of CH attacks [ 172 , 173 ], growing evidence indicates that other brain structures and network-level alterations may also contribute to the underlying pathophysiological mechanisms [ 174 ]. The brainstem structures are considered to play a pivotal role in CH pathophysiology [ 175 , 176 ], particularly through descending pain modulation and autonomic regulation pathways [ 177 ]. The periaqueductal gray, the locus coeruleus, and the dorsal raphe nuclei are key components of these circuits and are known to modulate the nociceptive trigeminal inputs, possibly contributing to CH attacks [ 174 , 178 ]. Interestingly, a recent study demonstrated that the most discriminative MRI patterns for distinguishing migraine and CH patients from healthy controls comprised the functional connectivity of the periaqueductal gray and the hypothalamus [ 179 ]. Notably, a study conducted on a small sample of healthy individuals showed a limited circadian modulation of the functional connectivity of brainstem nuclei involved in pain processing, and no fluctuations in pain threshold [ 180 ]. While preliminary and limited by the small sample size, this evidence opens interesting questions regarding the extent to which brainstem nuclei contribute to the circadian modulation of pain processing in CH. Recently, the ventral tegmental area (VTA)—a midbrain structure and central hub of the dopaminergic mesocorticolimbic system—has emerged, along with the broader mesocorticolimbic network traditionally associated with reward and motivation processes (including also the prefrontal cortex, nucleus accumbens, hippocampus, and amygdala), as a crucial component in the pathophysiology of CH. Multiple lines of evidence—including findings from several neuropsychiatric and chronic pain conditions, the effectiveness of VTA deep brain stimulation (DBS) in chronic CH, and CH-specific neuroimaging results—converge to support the involvement of this dopaminergic circuit in the disorder. Abnormalities within the mesocorticolimbic system are thought to contribute to pain chronification [ 181 ] and have been identified as a transdiagnostic marker across various neuropsychiatric disorders [ 182 ], which are frequently comorbid with CH [ 183 ]. Although not constituting direct evidence, these associations point toward a possible involvement of this network in the disorder. More direct and robust evidence comes from studies on VTA-DBS, which has demonstrated greater efficacy than posterior hypothalamic stimulation—the first region historically targeted in DBS interventions for CH [ 184 ]—in reducing both the frequency and the intensity of attacks in severe chronic cases [ 185 ]. Moreover, providing more direct evidence of the involvement of the mesocorticolimbic circuit, a series of functional neuroimaging studies in patients with chronic and episodic CH have progressively highlighted altered functional connectivity within this network. In particular, resting-state functional connectivity abnormalities have been observed in chronic CH between the posterior hypothalamus and diencephalic regions, including VTA [ 186 ], supporting the hypothesis of an integrated dysregulation between homeostatic and motivational networks in the pathophysiology of the disorder. Along the same line, reduced resting-state functional connectivity between the prefrontal cortex and amygdala and structural alterations in diencephalic areas (encompassing the VTA, nucleus accumbens, hippocampus, and prefrontal cortex) have been observed [ 187 ]. Notably, these mesocorticolimbic regions exhibited an unexpected volume increase, unlike the volume reductions typically reported in chronic pain conditions [ 188 , 189 ], suggesting CH-specific features in the chronic form. A recent study used the Monetary Incentive Delay (MID) fMRI task [ 190 ] to robustly trigger the activity of the mesocorticolimbic system across CH patients showing middle prefrontal cortex dysfunction in both chronic and episodic CH, but VTA alterations only in the chronic group, supporting a possible role of this region in pain chronification. Notably these alterations were unrelated to mood disturbances, nicotine use and CH phase (in-bout and out-of-bout). Another recent study [ 191 ] conducted on a large sample of episodic CH patients provided further evidence showing anatomical and functional alterations within the mesocorticolimbic system, in patients studied during the bout period but outside of attacks. Specifically, structural changes were observed in subregions of the amygdala and hippocampus, accompanied by widespread functional alterations involving the hypothalamus and thalamus as well. Although the precise functional significance of these results remains to be fully understood, overall, all these findings point to a general mesocorticolimbic imbalance as a core feature of CH pathophysiology. Another network thought to contribute to CH is the salience network. This important functional circuit mainly comprises the dorsal anterior cingulate cortex, anterior insula, and frontal operculum and is involved in attentional, affective, and autonomic modulation and coordination. Functional connectivity changes in salience network, even during headache remission [ 192 ] suggested that this network may be persistently affected in CH patients, possibly contributing to their susceptibility to pain and autonomic dysregulation. Importantly, several studies have highlighted a disrupted interplay between the salience network and other brain networks and regions in patients with episodic CH. Specifically, reduced functional connectivity between the salience network and the left executive control network has been observed, suggesting potential difficulties in switching between internally and externally oriented cognitive states during in-bout period [ 193 ]. Moreover, altered interactions between the salience network and the hypothalamus [ 159 ], as well as between the salience network and the thalamus in the in-bout phase [ 194 ], were observed further supporting the role of salience circuit dysregulation in the pathophysiology of CH. The thalamus is another key brain structure significantly involved in CH pathophysiology due to its role—among many others—in the processing and integration of nociceptive stimuli. Evidence suggests that the thalamus contributes to headache sensitization, through mechanisms similar to those observed in migraine patients, where thalamic activity correlates with hyperalgesia and allodynia [ 195 ]. Direct evidence for the involvement of the thalamus in CH was provided by the observation of altered fractional anisotropy of this structure [ 193 ], as well as by findings showing that CH patients can be distinguished from migraine patients based on decreased functional connectivity between the left thalamus and parietal regions—including the precuneus and angular gyrus—which are part of the default mode network (DMN). Although further studies are needed to clarify the involvement of DMN in CH pathophysiology, increased functional connectivity between this network and the executive control network was recently observed in patients with CH [ 193 ]. The DMN, known for its role in self-referential thought processes and typically antagonistic to networks such as the salience network during pain processing [ 196 ], may underlie some of the cognitive-affective alterations associated with CH. Importantly, recent functional imaging studies have revealed the existence of a complex network of interconnected regions within the cerebellum, including the prefrontal and insula cortices, as well as the amygdala. These regions have been shown to be involved in the processing of pain and emotional responses, suggesting a potential influence on the clinical symptoms of CH. The precise mechanisms by which these regions may modulate pain perception, emotional responses, and behavioural performance remain to be fully elucidated. Furthermore, structural imaging studies have demonstrated that patients suffering from chronic CH exhibit a reduced thickness in the anterior lobe of the cerebellum when compared to healthy individuals [ 197 ], thereby further substantiating the notion that the cerebellum plays a pivotal role in CH pathophysiology. The pituitary gland may also be implicated in a subset of patients. The presence of functional abnormalities or microadenomas has been demonstrated to exacerbate symptoms, particularly in the presence of hormonal imbalances such as hyperprolactinemia [ 198 ]. Altogether, the evidence reviewed highlights that CH is not the result of a localized brain dysfunction but rather reflects widespread disturbances involving both subcortical structures—such as the brainstem, thalamus, and hypothalamus—and large-scale network dysfunctions, particularly within the mesocorticolimbic, salience, and the default mode networks. Although this supports the conceptualisation of CH as a multisystem disorder, a comprehensive understanding of the functional meaning and temporal dynamics of these alterations remains elusive. The lack of a robust unifying interpretative framework underscores the need for longitudinal studies across the different phases of the disorder in large sample of CH patients, aimed at disentangling the role of brain networks reorganization in the pathophysiology of CH and their relationship with the clinical and neuropsychiatric data.

The epidemiology of CH in children and adolescents is limited by the lack of large population studies. Changeableness of headache phenotypes across the developmental age and verbal limitations of children to describe subtle clinical features required for diagnosis represent other obstacles [ 465 ]. Furthermore, a variable number of cases exhibit migraine-like features, which can pose a diagnostic challenge for those lacking experience [ 466 , 467 ]. In a nationwide sample of 5,671 children aged from 5 to 12 years old, the prevalence of children with headache attacks characterized by severe unilateral fronto-orbital pain, lasting from 15 to 180 min and not fulfilling criteria for migraine and/or tension-type headache, was 2.1%. Among them, there were no cases with parents reporting prominent cranial parasympathetic autonomic features accompanying the attacks [ 468 ]. The prevalence CH in paediatric age is only available from clinical samples of tertiary care centers specialized in headache in youth and retrospective clinical studies of adults with CH, both potentially influenced by selection and recall bias. In a multicenter study with 6,629 children and adolescents attending 27 headache centers in Italy, only two cases of CH were diagnosed in one year (0.03% of the total cases) [ 469 ]. In retrospective clinical studies of adults with CH, the prevalence rate of childhood-onset disease ranges from 5 to 27.5% [ 465 , 470 , 471 ]. The peak age of onset of the episodic CH is from 15 to 18 years of age [ 465 , 472 ], while, curiously, the chronic subtype can start as early as 1 year of age [ 472 ]. According to a recent systematic review and meta-analysis, the 1.8 sex ratio of childhood-onset CH [ 467 ] is lower than that (4.3) of adult-onset CH [ 5 ], although a decreasing trend of the male predominance has been reported over the last ten years [ 37 , 276 ]. The study of headaches peculiarities at developmental age is a unique opportunity to deeper understand neurobiological, clinical, and evolutionary aspects of headache disorders with potentially impacts on therapeutic interventions. Due to its rarity in childhood, identifying peculiarities of CH is challenging. However, a number of peculiarities of CH in children has been reported in case series, systematic reviews and meta-analysis, including: (a) less numerous attacks during a given bout [ 467 , 473 , 474 ]; (b) less frequent and shorter cluster periods [ 467 , 473 , 474 ]; (c) the presence of pain-free intervals lasting several days within the cluster period [ 474 , 475 ]; (d) less pronounced ipsilateral autonomic manifestations and associated restlessness behaviour during the attacks [ 467 ]; (e) more prototypical migrainous features (photophobia, phonophobia, pain aggravated by physical activity, nausea, and vomiting) [ 1 ], and co-occurrent diagnosis of migraine [ 470 ]; and (f) a favourable response to indomethacin treatment [ 466 , 474 , 476 – 478 ]. Conflicting findings predominantly come from retrospective studies of CH in adults reporting more similarities than possible peculiarities of CH in childhood. Among these, the circannual periodicity of attacks, the circadian rhythm of attacks, the equal proportion of migraine-like features, and the complete ipsilateral cranial autonomic signs deserve to be mentioned [ 467 , 472 ]. Despite the overlapping characteristics of CH and migraine in children, as well as some similar features of CH and other trigeminal autonomic cephalalgias, the idiosyncratic headache pattern of CH tends to emerge and helps the clinician in the correct diagnosis. However, secondary causes must be carefully ruled out by appropriate investigation, especially intracranial neoplasms (e.g., prolactinoma, hypothalamic pilocytic astrocytoma) and rhinosinusitis, in addition to other less reported causes as cerebrovascular diseases, head and neck trauma, optic neuritis, and Graves’ disease [ 470 ]. As for other primary headaches, the therapy of CH in youth derives from what is currently used in adult CH patients. Usually, CH treatments are divided in three types (acute, preventive, and transitional) based on the intervention timing. Inhalation of 100% oxygen at a dose of 7–12 L/min for 15 to 30 min through a non-breathing mask, commonly considered as the most effective acute treatment, is particularly suitable for children and adolescents due to the lack of adverse effects [ 342 , 479 , 480 ]. Subcutaneous sumatriptan proved scarcely useful when used in paediatric patients [ 470 , 474 ]. On the contrary, triptans administered as nasal spray (sumatriptan and zolmitriptan) have some evidence of efficacy [ 477 ]. Some retrospective studies reported the use of non-recommended drugs, such as NSAIDs, acetaminophen, rizatriptan, and oral sumatriptan [ 470 , 477 ]. Notably, the results with these therapies are generally poor. Verapamil, topiramate, and melatonin may be useful for CH prevention in youth [ 18 ]. Verapamil can be titrated until the target dose of 3–10 mg/kg/day, monitoring the electrocardiogram P-R interval. It is considered as the most effective drug for CH prevention in children and adolescents [ 470 ]. Hypotension, fatigue, bradycardia, and atrio-ventricular block are the main side effects. Topiramate has lower evidence of efficacy, compared to verapamil [ 470 ]. In paediatric age, it is used at the dose of 1–2 mg/kg/day. Decreased serum bicarbonate, hyperammonaemia, dizziness, and Weight loss are the main side effects. In the literature, lithium was used only in 3 Young patients, among whom only one 13-year-old boy had a positive outcome. Melatonin, whose use in youth should be considered due to the lack of serious adverse events, has no solid evidence of efficacy in children and adolescents [ 478 ]. The dose to be used is of 0.1–0.2 mg/kg/day. The use of valproate, gabapentin, indomethacin, propranolol, methysergide, and pizotifen has been anecdotally reported in the literature [ 470 , 477 , 478 , 481 – 483 ]. Transitional therapies are started along with preventive medications and should lead to a quick improvement, before the preventive drugs may become effective, which may take days or weeks [ 342 ]. The typical transitional medication is represented by corticosteroids whose use for short periods has been suggested in children [ 484 ]. Indeed, there are case series showing both efficacy and safety of oral and intravenous prednisone [ 478 , 482 , 485 ]. The commonly used dose of oral prednisone is 1–2 mg/Kg/day. GONB has not been studied adequately in Young CH patients. Unilateral injection with 30 mg of 1% Lidocaine and 40 mg of methylprednisolone has been shown effective on both intensity and frequency of the attacks in 3 adolescents with CH [ 486 ]. CH has been found associated to tobacco exposure also in paediatric patients [ 372 , 472 ]. Therefore, smoking cessation in the household is recommendable. Alcohol and recreational drugs might trigger the attacks; thus, they should be avoided. In conclusion, educational therapy regarding alcohol, tobacco and substance abuse should be an important part of CH treatment in adolescents.

Traditional preventive treatment of CH mostly includes several non-specific pharmacological agents with limited evidence (Table  2 ). Verapamil is considered a first line preventive option (medication of choice) for episodic and chronic CH based on consensus opinion, open label studies, and a couple of small randomized controlled trials [ 342 , 343 ]. The mechanism of action of verapamil has not been fully elucidated; however, it is thought to modulate circadian rhythms and calcitonin gene-related peptide [ 289 ]. Table 2 Non-specific therapies for cluster headache Medication Mechanism of Action Dose Evidence Adverse Events/Monitoring Comments Verapamil Calcium channel blocker; inhibits cortical spreading depression and vascular reactivity 240–960 mg/day (divided doses) Gold standard for prophylaxis Bradycardia, hypotension, constipation, ECG monitoring (AV block, QT prolongation) High doses often required; regular ECGs essential Lithium Modulates neurotransmitters (serotonin, dopamine); affects second messenger systems 600–1500 mg/day (target serum: 0.4–0.8 mmol/L) Good evidence, especially for chronic cluster Renal and thyroid function, lithium levels, tremor, nausea, polyuria Narrow therapeutic range; frequent monitoring needed Topiramate Blocks voltage-gated sodium channels; enhances GABA activity 50–200 mg/day Limited evidence, off-label use Cognitive dysfunction, nephrolithiasis, paresthesia, weight loss Off-label; reserved for refractory cases Melatonin Regulates circadian rhythms; antioxidant properties 9–12 mg at bedtime Moderate evidence, particularly for episodic cluster Drowsiness, vivid dreams, hormonal effects Useful for nocturnal attacks; well-tolerated Non-specific therapies for cluster headache The recommendation is supported by small RCTs and open label studies including a multi-center, randomized placebo-controlled, crossover study [ 344 ]. After a 5-day run-in, 30 participants were randomized to verapamil (120 mg tid) or placebo (tid) for 2 weeks. The primary endpoint was reduced attack frequency; intensity and duration of attacks and use of abortive agents, mostly subcutaneous sumatriptan, was captured with a headache diary. In the verapamil arm, 80% were responders versus in the placebo arm there were non-responders. In a multicenter, double blind, cross-over trial, verapamil was also compared to lithium carbonate for chronic CH prevention; lithium carbonate and verapamil were both effective, however, participants on verapamil had fewer adverse effects and the latency period was shorter than for lithium carbonate [ 345 ]. Higher doses of verapamil may be more effective; however, adverse events may be limiting including constipation, gingival hyperplasia, and serial electrocardiograms are needed to evaluate for electrocardiac abnormalities [ 346 ]. The mechanism of action of lithium for the preventive treatment of CH is not fully elucidated; however, lithium is thought to act on several neurotransmitter systems and signal transduction pathways [ 347 ]. According to the European Academy of Neurology guidelines [ 342 ], lithium carbonate 600–1500 mg should be considered for both episodic and chronic CH when there is a lack of response to verapamil. The recommendation is supported by some randomized controlled trials [ 348 , 349 ] with mixed results, and by positive open label studies [ 350 – 352 ]. Due to narrow therapeutic windows there is a risk of toxicity, drug levels should be checked routinely for safety but not efficacy. In addition, chemistry, liver, renal and thyroid function need to be monitored. Lithium should be avoided with concurrent indomethacin use; concurrent use of verapamil may also enhance excretion of lithium [ 342 ]. Among newer treatments, monoclonal antibodies targeting the CGRP pathway have shown promise, though with subtype-specific limitations (Table  3 ). Galcanezumab, an anti-CGRP ligand monoclonal antibody, at the 300 mg dose, significantly reduced attack frequency in episodic cluster headache in a randomized controlled trial during weeks 1 to 3 [ 86 ]. However, this benefit could not be replicated in chronic CH [ 87 ]. Similarly, erenumab, a CGRP receptor antibody, failed to show efficacy in a randomized, double-blind, placebo-controlled phase 2 trial in chronic CH (CHERUB01) [ 353 ]. In interim analyses for primary endpoint of mean change from baseline in the weekly average number of CH attacks at week 4, fremanezumab showed futility in Phase 3 trials for both episodic and chronic CH (ENFORCE) [ 354 , 355 ]. Eptinezumab, a monoclonal antibody to CGRP administered intravenously, failed to show efficacy on the primary outcome measure in a phase 3, parallel-group, double-blind, placebo-controlled trial (ALLEVIATE) for prevention of episodic CH [ 90 ]. However, eptinezumab treatment was associated with numerically higher responder rates and improvements in average daily pain and patient-reported outcomes In an open label study longer term safety and efficacy for prevention of chronic CH (CHRONICLE trial), the participants reported improvements in attack frequency, pain severity and patient-reported outcomes [ 91 ]. Table 3 CGRP targeted therapies for cluster headache Medication Mechanism of Action Dose Evidence Adverse Events/Monitoring Comments Galcanezumab Monoclonal antibody against CGRP ligand 300 mg monthly (higher than for migraine) Positive evidence only for episodic cluster Injection site reactions, constipation Only approved therapy for episodic cluster headache ; not for chronic cluster Fremanezumab Monoclonal antibody against CGRP ligand No established dose for cluster Insufficient evidence Injection site reactions Not approved for cluster headache Erenumab Monoclonal antibody against CGRP receptor Not approved for cluster Negative trials Constipation, hypertension, injection site reactions No approval; trials failed to show efficacy Eptinezumab Monoclonal antibody against CGRP ligand 400 mg every 12 month Controversy results Nasopharyngitis, hypersensitivity reactions IV administration only; not approved for cluster e Rimegepant Oral CGRP receptor antagonist (Gepant) No approval for cluster No evidence Nausea, urinary tract infections Approved only for migraine (acute and prevention) Atogepant Oral CGRP receptor antagonist (Gepant) No approval for cluster No evidence Nausea, constipation, fatigue Approved only for migraine prevention CGRP targeted therapies for cluster headache Topiramate has shown some benefit in open-label studies and small trials, although its efficacy is inconsistent and limited by tolerability concerns [ 356 , 357 ]. Sodium valproate has also been evaluated in small series with modest effect, but the evidence base remains insufficient to support routine use [ 358 ]. Melatonin is a hormone produced by the pineal gland, regulated by the suprachiasmatic nucleus, and is associated with shifts in circadian rhythms, and the sleep-wake cycle. The circannual urinary melatonin concentrations is lowered in individuals with episodic CH [ 359 ]. The European Academy of Neurology recommends melatonin when standard therapies fail [ 342 ], although evidence remains limited and heterogenous. In a double-blind placebo-controlled pilot study with parallel group consisting of 20 participants with CH, (2 chronic, 18 episodic), melatonin 10 mg taken in the evening was found to be more effective at reducing headache frequency than placebo [ 360 ]. Specifically, 5 out of 10 participants had a reduction in attack frequency 3–5 days after treatments; participants with chronic CH and those that received placebo did not response. Adverse events were not reported for either group, suggesting a potential utility for adjunctive therapy despite the low response rate. Naratriptan, a long-acting oral triptan, has been used in selected cases with predictably nocturnal CH, administered in the evening to pre-empt nighttime attacks. While evidence is limited to case series and clinical experience, some reports suggest that naratriptan 2.5 mg given prophylactically may reduce the frequency or intensity of nocturnal attacks [ 361 ]. This approach may be considered in patients for whom standard prophylaxis is insufficient or contraindicated. Ergotamine, though now rarely used, has demonstrated efficacy in nocturnal CH when administered in sustained-release formulations or intravenously during inpatient settings [ 362 , 363 ]. OnabotulinumtoxinA has been evaluated in small case series using an adapted PREEMPT injection protocol, originally developed for chronic migraine or sphenopalatine injections. Although some patients with chronic CH have reported benefit, controlled trials are lacking, and its application remains experimental [ 364 – 366 ]. In practice, combination therapy is frequently employed, particularly in chronic or refractory cases. Clinicians may combine verapamil with lithium, corticosteroids, or nerve blocks, often tailoring regimens to the patient’s comorbidities and response. While such multimodal strategies are considered clinically reasonable, controlled data on specific combinations are lacking. Finally, several experimental and lesser-studied therapies have been reported. Intranasal capsaicin and its synthetic analog civamide may reduce attack frequency via desensitization of trigeminal afferents [ 367 ]. Clomiphene citrate, a selective estrogenic receptor modulator, has been proposed in individual cases, presumably via hypothalamic modulation [ 368 ]. Additionally, ketogenic diets have been associated with reduction in attack frequency in select patients, possibly through metabolic and anti-inflammatory mechanisms [ 369 ].

CH is recognized as a complex disease, due to interactions between genetic and environmental factors and to date, various molecular genetic clues have been identified. First degree relatives of patients with CH have a 5–18 increased risk of developing the disease compared to the general population while second-degree relatives of patients have an 1–3 increased CH risk [ 48 , 49 ]. Inheritance is likely to be autosomal dominant with low penetrance in some families, although there may also be autosomal recessive or multifactorial inheritance in others [ 50 ]. CH has also been reported to be significantly increased in monozygotic twins than in dizygotic twins [ 49 , 51 ]. Results from Linkage studies in CH have so far been negative. Calcium channel alpha 1 subunit (CACNA1A) micro satellites have been screened in CH families from the Netherlands and in a Swedish cohort, both studies were reported negative [ 52 , 53 ]. A genome wide scan of 400 micro satellites markers in CH families from Denmark, Italy, Sweden and UK was additionally also reported negative [ 54 ]. Several genetic association studies have been performed in CH patients using the candidate gene strategy. The putative pathogenetic role of posterior hypothalamus in CH pathophysiology as well as the chronobiological features of the disease suggested a role for genes related to hypothalamic and chronobiological functions in the disease. Multiple studies investigated the association between CH and the Hypocretin Receptor 2 gene ( HCRTR2 ), that is expressed in the posterior hypothalamus. In an Italian cohort, a significant association between allele G of the G1246A polymorphism (rs2653349) was found [ 55 ]. Haplotype analyses have further supported a role for HCRTR2 in CH and proposed that rs2653349 can lead to minor changes in the mRNA structure, affecting both mRNA stability and the dimerization process of the HCRTR2 gene [ 56 , 57 ]. Nonetheless, the suggested association with rs2653349 could not confirmed by a recent meta-analysis although showed a weak association with rs9357855 [ 58 ]. Variance in genotype and allele frequencies among the different populations examined may explain these discrepancies. At present, therefore, the HCRTR2 gene remains an interesting candidate gene for involvement in the pathophysiology of CH. Considering the periodical occurrence of CH bouts, genes that regulate chronobiological function have been investigated. Several studies genotyped CH patients and controls for the rs1801260 polymorphism of the CLOCK gene. So far, however, no association has been identified between CH and this SNP [ 59 – 62 ]. A more recent study also explored the association with two other SNPs (rs11932595 and rs12649507) of the CLOCK gene and reported a significant association of the disease with rs12649507 [ 63 ]. CRY1 and CRY2 are genes deeply involved in the regulation of the circadian clock. In a Swedish CH cohort, a strong association was found between rs8192449 and CH [ 64 ]. The PER1 , PER2 , and PER3 genes are light sensitive clock genes and have been investigated in relation to CH in Swedish and Norwegian cohorts [ 65 , 66 ]. The results indicated no involvement of these genetic variants in the disease. According to clinical studies, more than 50% of CH patients have attacks triggered by alcohol consumption, suggesting a causative role for genes related to alcohol metabolism and alcohol dependence in the genetic predisposition to the disease. The alcohol dehydrogenase 4 ( ADH4) gene encode the pi subunit of the alcohol dehydrogenase (ADH) enzyme and significantly modulates alcohol metabolism. An Italian study found a significant genetic association between ADH4 SNP rs1126671 and the disease [ 67 ]. Another Italian study explored the association between the rs1126671 and rs1800759 polymorphisms of the ADH4 gene in CH patients versus controls finding significantly different allele and genotype frequency between sporadic CH and controls [ 61 ]. The attempt to further replicate the association between polymorphisms of the ADH4 gene and CH failed in case-control cohort studies performed in Swedish, Chinese, and Greek cohorts [ 62 , 68 , 69 ]. So far there has been limited focus on the genetic aspects of calcitonin gene-related peptide (CGRP) and its receptors in CH, but one genetic variant in the receptor activity-modifying protein 1 gene ( RAMP1 ), part of the CGRP receptor complex, has been linked to CH in a Swedish cohort [ 70 ]. In a recent meta genome wide association study (GWAS) including 10 European cohorts and one East Asian eight loci were identified [ 15 ]. Three of these loci interestingly overlap with identified migraine loci; FHL5 , PLCE1 and LRP1 [ 71 ]. Mendelian randomization analysis further implicated smoking as causal risk factor. The GWAS top hit on chromosome 2 include MERTK (Mer tyrosine kinase protooncogene), a gene crucial for regulating microglia mediated inflammation. MERKT is primarily active in the brain’s supporting cells, microglia, and has an important role in inflammation and phagocytosis of harmful molecules and cell debris in the brain. MERTK gene expression has been studied in patient-specific tissue, which shows increased levels of MERTK mRNA and its ligand Gal-3 in serum from study participants with CH [ 72 ]. Both MERTK and Gal-3 are further expressed in the trigeminal ganglion, involved in pain signaling in the facial region, in tissue from rodents. These findings, strengthens the hypothesis that the MERTK gene may play an important role in the pathophysiology of CH. A recent study from France, using the whole genome sequencing (WGS) strategy, searched for candidate genes and new genetic variants of the disease investigating a multigenerational CH pedigree. The WGS on four members of the pedigree found that two family members showing the same phenotypic circadian pattern (familial periodicity) of symptoms had two genetic risk loci in the HCRTR2 and in the CLOCK genes [ 73 ]. Thus, the risk of CH appears to be significantly increased by the concomitant presence of these polymorphisms. Pharmacogenetic studies investigating the role of gene polymorphisms in drug responses in CH have been performed. In an analysis of 184 CH patients, no association between the HCRTR2 G1246A polymorphism and treatment response to triptans, oxygen, verapamil, or corticosteroids was found [ 74 ]. Contrariwise, a polymorphism of the GNB3 significantly influenced the chance of responding to treatment with triptans in CH patients [ 75 ]. A similar trend was observed in a Greek CH cohort; however, the association did not reach significance [ 69 ]. A Swedish study shows that genetic variants such as rs1024905 located on chromosome 12 and previously linked to triptan usage in migraine, can influence triptan usage in CH in Sweden. The cumulative effector score of five variants also indicates a complex genetic contribution to triptan usage [ 76 ]. Genotypes of the 5-HT transporter gene-linked polymorphic region ( 5-HTTLPR ), and CYP3A4 genes do not influence treatment response to triptans or verapamil in CH patients [ 77 , 78 ]. Increased knowledge in the genetics field of CH can improve diagnosis and treatment by identifying biomarkers and novel drug targets, as well as developing future treatment strategies. In the past years there have been a significant process in the genetics of CH. Several candidate genes have been screened in CH cohorts worldwide, but with conflicting results. This is mainly due to few replication studies, small cohorts and geographical differences. Genome wide association studies have on the other hand identified eight significant loci, where three of them are shared with migraine, and the prioritized genes show enrichment in arterial and brain tissues.

In contrast to other more common forms of headache, such as migraine, this condition, precisely because of its lower prevalence, is little known among laypeople and often among physicians too, leading to frustrating clinical and diagnostic experiences for patients [ 567 , 568 ]. To achieve significant improvements in care systems, foster research, diminish stigma, and promote health equity, advocacy for patients with CH is essential (Fig.  4 ). Below are several key issues that should be addressed through advocacy. Fig. 4 Factors involved in advocacy for cluster headache patients Factors involved in advocacy for cluster headache patients As mentioned above, one of the major challenges of CH is its rarity, which results in limited knowledge and awareness of the disease [ 569 ]. Even in countries with well-developed health care systems, patients with CH often receive a delayed diagnosis or face misdiagnosis and clinical mismanagement. Delays in diagnosis may be due to the patient delaying seeking medical care or the physician struggling to make the correct diagnosis [ 568 ]. Therefore, advocacy should focus on educating and raising awareness among laypeople and physicians by providing precise and accessible information about the condition, its severity, and the need for timely professional intervention [ 570 ]. The creation and dissemination of educational materials, involvement in awareness campaigns, and collaboration with patient advocacy groups can help bridge this knowledge gap [ 570 ]. The highly disabling nature of CH attacks is another critical aspect that requires advocacy [ 571 ]. Attacks are so intense that they are often described as the most excruciating pain imaginable, and they have a profound impact on patients’ daily lives and their ability to work, maintain social relationships, and carry out everyday activities [ 569 ]. Advocacy should aim to make individuals, society, and policymakers aware of the real disabling effects of CH [ 572 ], in order to gain social recognition of the condition and concrete responses, such as the promotion of policies that ensure patients’ access to adequate sick leave, workplace accommodations, and support for disability claims [ 573 , 574 ]. The lack of formal recognition of CH as a disabling condition in disability assessment systems is a shortcoming that needs to be overcome at global level [ 575 ], in line with what has already happened in Italy [ 576 ], thanks in part to the efforts of local patient associations. In Italy, this recognition has made it possible to reach a consensus statement between health professionals, forensic experts, and representatives of patient associations, precisely to help physicians ensure recognition of their patients’ social rights as people living with disabilities [ 577 ]. As mentioned, this achievement, enabling patients with chronic headaches (including chronic CH) to lead more dignified and sustainable lives, needs to be extended globally. Although the stigma associated with headache has been extensively studied in migraine [ 578 , 579 ], it does not spare patients with CH [ 557 , 580 ]. Lack of knowledge about the disorder among the general public can lead to skepticism and minimization of patients’ suffering [ 557 ]. This stigma can be internalized by patients themselves, who may develop feelings of shame, isolation, and reluctance to speak openly about their condition [ 579 ]. Advocacy must actively counteract this stigma through awareness campaigns that promote accurate and non-stigmatizing portrayals of headache [ 567 ], including CH. This can entail the use of patient-friendly language, for example, where the focus is on the disorder rather than the person (“person with cluster headache” rather than “cluster headache patient”) [ 581 ]. The process of destigmatization [ 570 ] must also extend to the workplace, which must be rethought so that it becomes an inclusive environment where the invisible nature of the disorder is recognized, environmental triggers are reduced, awareness among colleagues is encouraged, and organizational policies based on the principles of diversity, inclusion, and accessibility are adopted, thus helping to improve the quality of life and productivity of patients with the disorder [ 582 ]. In terms of access to care, patients with CH often face challenges due to a lack of specialists with experience in managing this rare condition [ 570 ]. The concentration of the relevant expertise in specialized centers can create geographic and time barriers for patients, especially in rural areas or areas with limited health care resources [ 583 , 584 ]. Advocacy can promote the development and strengthening of networks of centers specializing in headache care to ensure a more equitable distribution of services and facilitate access to specialized consultations through tools such as telemedicine [ 567 , 580 ]. In addition, it is crucial to support the training of primary care physicians to recognize the signs and symptoms of CH at an early stage. This area may also include the use of artificial intelligence to provide specific tools to support the clinician [ 585 ] and ensure that patients are promptly referred to specialists [ 570 , 586 ]. Collaboration among patient advocacy groups, scientific societies, and healthcare institutions is essential to define and implement structured care models that address the specific needs of patients with CH [ 580 ]. Although research on CH is progressing, more scientific effort and more financial support are needed to fully understand the pathophysiological mechanisms of the disease, identify new therapeutic targets, and develop more effective and targeted treatments [ 11 , 587 ]. The funding of migraine research has been considered disproportionately low compared with the burden of the disease [ 575 ], and it is likely that CH research, due to the rarity of the condition, will receive even less attention. Therefore, advocacy groups must approach national and international funding agencies to urge increased investment in CH research [ 570 , 584 ]. This includes support for epidemiologic studies, basic research into disease mechanisms, and clinical trials to evaluate the efficacy and safety of new therapies. Notably, CH treatment has not seen the level of research activity that has been devoted to new therapies for migraine, and there is no consensus on whether new therapies developed for migraine can be applied to CH, despite a considerable pathophysiological overlap between the conditions [ 11 ]. As a result, many patients are driven by frustration at the limited range of treatment options and difficulty accessing treatment to seek alternative, not always licit, routes [ 372 , 375 ], sometimes with the support of associations [ 588 ]. These problems are also found in countries with greater healthcare resources. Patient empowerment is a central element of advocacy in CH: providing patients with accurate and understandable information about their condition, available treatment options, and their rights can help them to become active participants in treatment decisions and advocates for themselves [ 570 ]. Patient associations are essential in creating supportive communities where patients can share experiences, obtain information, and feel less isolated [ 570 ]. Therefore, it is crucial to support and strengthen these organizations by providing them with resources and platforms to make their voices heard [ 572 ]. Finally, there are new challenges on the horizon that should be considered when advocating for patients with CH. Given the increasing pathophysiological knowledge of the disease and the growing possibility of monitoring and modulating brain activity [ 589 ], patients with CH are exposed to the risk of manipulation and control of this activity. This aspect is being studied within the field of “neurorights” [ 590 ] with the aim of ensuring preservation of patients’ mental integrity and cognitive freedom, crucial for safeguarding their mental autonomy and protecting them against non-consensual interference with their brain activity. This ethical issue, still little discussed, is highly relevant to patients suffering from CH, and represents an evolution of the similar one already raised in psychiatry concerning the impact of drugs on the autonomy, privacy, and integrity of patients receiving neuropsychopharmacological treatments [ 591 ]. Moreover, as artificial intelligence becomes more widespread in health care and the field of headache [ 585 ], it will be crucial to monitor its ethical aspects to ensure fairness, transparency, and protection of the rights of patients, especially those who, like individuals with CH, are at risk of being marginalized by algorithmic systems that are opaque or inadequately calibrated to the specific needs of those affected by rare diseases [ 592 ]. In conclusion, advocacy in CH is an ethical obligation and a practical necessity. The objectives of addressing the unmet needs of patients, combating stigma, improving access to care, and promoting research are attainable goals that can be achieved through the collaborative efforts of patients, family members, physicians, researchers, associations, and policymakers. By following the advocacy models developed for migraine and adapting them to the specificities of CH, it is possible to bring about change and significantly improve the lives of those living with this painful and disabling condition [ 575 , 584 ]. All efforts must be directed towards addressing the hidden disabilities of patients with CH, whose quality of life is severely impacted by the severity of their acute attacks and the discomfort experienced during the intercritical phases. Like those with migraine [ 593 ], patients with CH have to cope with anticipatory anxiety, social and work restrictions, and a severe emotional impact of their condition, which is completely underestimated by those around them and by those who should be caring for them.

CH is a debilitating disorder characterized by severe, unilateral pain typically located around the eye. As one of the most common TACs, autonomic symptoms are typical during CH attacks. CH can be divided into episodic and chronic subtypes. Those with episodic CH have bouts that last from 7 days to one year, separated by pain-free periods for at least 3 months, when untreated. Most cluster periods usually last between two weeks and three months [ 1 ]. On the other hand, those with chronic CH present with persistent attacks for more than one year without any remission periods or with less than three months of remission [ 9 ]. The clinical features of these subtypes reveal significant differences in attack frequency, bout duration, and associated symptoms. Individuals suffering from episodic CH experience well-defined cycles of recurrent attacks, often referred to as “bouts”, which can last for weeks to months. These bouts are commonly followed by extended periods of remission, sometimes lasting months or even years [ 10 ]. In contrast, patients with chronic CH experience attacks without substantial remission periods, leading to frequent bouts that can persist indefinitely [ 11 ]. CH episodes typically manifest as short but intense attacks, with each attack lasting between 15 min and three hours [ 9 ] on average of 55 to 130 min without any significant difference between episodic and chronic CH [ 12 , 13 ]. During an active bout, patients may experience varying frequencies of attacks, ranging from one attack every other day to as many as eight attacks within a single day. The average frequency reported within a bout can vary, with many patients reporting two to three daily attacks [ 10 ]. The episodic nature might also be influenced by factors such as smoking, with approximately 70–90% of patients with CH having a history of smoking, which correlates with an increase in attack frequency and duration of the bouts [ 14 ]. In fact, a causality between smoking and CH was recently demonstrated by a Mendelian randomization study [ 15 ]. In contrast, chronic CH is characterized by a more relentless pattern of attacks. Patients with chronic CH experience a higher frequency of attacks, with some individuals enduring as many as 10 or more headache episodes each day during active bouts, which can last for years without significant relief [ 11 ]. Patients with chronic CH often report similar attack durations to those with the episodic form. However, the frequency and lack of remission can exacerbate the individuals’ experience of pain and incapacitation. In very rare cases, bilateral CH has been reported [ 16 ]. However, most patients with CH presented strictly side-locked headache, meaning that a patient’s headache episodes were mostly confined to one side within a bout. The intensity of the headache is usually excruciating. In fact, some considered CH one of the most intensely painful human conditions reported [ 17 ]. However, a proportion of patients with CH demonstrates “side shifting” either within or between bouts. Side shifting between bouts has been reported at around 13–18% [ 18 ]. Within individual bouts, a study [ 18 ] indicated that approximately 33% of patients with chronic CH patients experienced side shifts, compared to 14% of their episodic CH counterparts while another study on Nordic cohort observed similar prevalence of 13–15% between episodic and chronic CH [ 19 ]. A recent study reported that the presence of side-shifting was associated with a significantly increased odds (OR = 2.24) of chronic CH compared to episodic CH [ 20 ]. Patients with CH frequently exhibit restless and agitated behaviours, presenting with continuously moving in an unsuccessful effort to alleviate their pain. Approximately 80–90% of individuals become restless in the Western cohorts [ 21 ] but lower at around 50% in the Eastern cohorts [ 22 , 23 ]. The associated symptoms of CH are notable for their specificity to the attack’s side. The most common cranial autonomic symptoms (CAS) are lacrimation/conjunctival injection (reported by > 90% of patients). However, rhinorrhea, miosis, ptosis, and eyelid oedema are also common [ 9 ], and rhinorrhea has been reported less in chronic CH [ 24 ]. These features distinguish CH from other headache disorders, such as migraine, where similar CAS may not present unilaterally [ 25 ]. A careful distinction from trigeminal neuralgia involving the first division of the trigeminal nerve may be required in selected cases. However, the neuralgiform pain character, the sporadic and mild degree of intensity of CAS and the absence of restlessness during the painful attacks of trigeminal neuralgia, make the clinical distinction with CH reasonably straightforward [ 26 ]. Notably, visual disturbances such as unilateral visual sensitivity can also occur, primarily affecting the same side as the headache [ 27 ]. Specific data differentiating the prevalence of individual CAS between episodic CH and chronic CH are limited. Both subtypes exhibit similar autonomic features, but comprehensive studies directly comparing the frequency of each CAS are lacking. However, results from an extensive survey suggest that the highest pain intensity (numeric rating scale = 10/10) is linked to an increased number of CAS, indicating a proportional relationship between pain severity and autonomic symptom expression [ 17 ]. In regard to the sex-related differences in clinical features, studies from large cohorts have reported older onset age [ 4 , 28 , 29 ] and a higher proportion of smokers and alcohol consumers [ 4 , 28 ] in male patients. Male patients also demonstrate a higher proportion of conjunctival injection [ 4 , 30 ] and a lower proportion of eyelid oedema [ 4 , 30 ], ptosis [ 30 , 31 ], restlessness [ 30 , 31 ], pacing [ 4 , 28 ], and nausea [ 4 , 28 , 29 ] during attacks compared to female patients. Two studies [ 30 , 31 ] showed that among patients with CH, female patients were diagnosed with chronic CH more. In contrast, others [ 4 , 28 , 29 ] found no significant sex difference. Female patients with CH also reported higher comorbidity with migraine [ 4 , 28 , 31 ], which is reasonable considering the female predominance of migraine per se. For episodic CH, there is a characteristic circannual pattern, often peaking in the spring and fall, possibly aligning with temperature changes [ 32 ] and daylight changes [ 33 ], environmental stressors or lifestyle changes during these times [ 12 , 13 ]. Conversely, due to the incessant nature of chronic CH, the triggers for the bout period may be less distinct [ 11 ]. However, such circannual fluctuations of attack frequency have still been reported in patients with chronic CH who demonstrate higher attack frequency during certain times of the year than the rest [ 34 ]. The circadian pattern is also a signature of CH, with a clear peak between 21:00 and 03:00 [ 35 ]. The individual pattern is exact and varies only by minute in many patients on a day-to-day basis. Triggers for attacks have also been reported. Alcohol consumption has been identified as a significant trigger reported by up to 50–80% of patients with CH, with a slightly higher rate (65%) in episodic CH compared to chronic CH (54%) [ 18 ]. Sleep has also been reported as a trigger in 80% of a 275-subject cohort [ 36 ]. Other common triggers include smells, bright/flashlights, hot showers/heat, nitroglycerin intake, etc [ 37 , 38 ]. The sustained hypoxemic event cause by either sleep apnea, poor ventilation/stagnant air, or change of altitude, could provoke the attacks as well [ 39 ]. Interestingly, these triggers only trigger attacks during the bout period for episodic CH and usually have no effect during the out-of-bout period. There is a lack of research comparing the differences between episodic and chronic CH regarding their triggers. Additionally, the pre-cluster symptoms, frequently presenting as localized pain and sensations, may alert patients to the impending onset of a bout and tend to be more recognized in individuals with a history of multiple bouts [ 40 ]. Patients who experience frequent cluster bouts may become more familiar with recognizing these signs. However, the related symptoms can vary significantly between individual experiences [ 41 , 42 ]. Research shows that patients with episodic CH report more symptoms before a headache attack compared to those with chronic CH [ 5 ]. Patients with chronic CH tend to report less predictability in their symptomatology, as the continuous nature of their condition makes patterns harder to discern. Neurobiological mechanisms underpinning these clinical features highlight the role of the hypothalamus in CH pathophysiology [ 43 ]. Activation of the hypothalamus during an attack may explain the circadian patterns and the subsequent discharge of pain and CAS [ 44 ]. A recent study also found that the sphenopalatine ganglion (SPG) volume is larger on the painful side, and the volume of ipsilateral SPG is associated with the number of CAS in episodic CH [ 45 ]. In chronic cases, repeated activations of the hypothalamus due to constant attacks [ 46 ] may lead to neurotransmitter exhaustion [ 47 ], disturbing the circadian rhythmicity of hormones, such as testosterone and cortisol, and further complicating the management options. Such insights remind healthcare providers of the importance of considering long-term disease management strategies for those who suffer from chronic CH. In summary, both chronic and episodic CH presents with a set of distinctive clinical features characterized by severe unilateral pain and ipsilateral autonomic symptoms. The differing attack frequencies, bout durations, and symptomatology distinguish the episodic and chronic CH, highlight the complexity of CH. The necessity for individualized treatment approaches to manage the debilitating nature of CH effectively should be underlined.

There are currently three invasive neurostimulation methods that have been tested in CH. T: posterior hypothalamic deep brain stimulation, occipital nerve stimulation and spheno-palatine ganglion stimulation. The evidence of efficacy is limited for the first two, moderate for spheno-palatine ganglion stimulation [ 342 ]. Altogether, their high cost, access limitations and potential side effects make them indicated only for rCCH. Given its implication in autonomic function and chronobiology, as well as attack-related hypermetabolism demonstrated in the posterior hypothalamus ipsilaterally to the site of CH attacks, the hypothalamus emerged as a central target for potential neuromodulation in CH. Nearly 25 years ago, Leone et al. reported for the first time the effectiveness of posterior hypothalamic DBS used compassionately in a patient suffering from a very severe form of rCCH [ 410 ]. Then, the same team published the first case series with 5 rCCH patients treated with DBS who were pain-free after 2 to 22 months of follow-up monitoring [ 411 ]. In the aftermath, others, in different experienced centers, replicated the same procedure of DBS in rCCH. A recent systematic review on preventive treatment of rCCH reported more than a hundred published cases of patients treated by DBS and a meta-analysis of the reported data resulted in a pooled response rate of 77% (OR 0.770 [95% CI 0.594–0.957]) [ 412 ]. This response rate must be interpreted with caution as it results from data obtained under uncontrolled conditions, especially as a French randomized control crossover study including 11 patients did not show significant differences among the sham and active periods (both lasting one month) regarding weekly attack frequency [ 413 ]. Nevertheless, the effectiveness observed in the case reports and case series does not seem to be related to a placebo effect since a blind-to-the-patient interruption of DBS has been achieved in many cases, with a relapse on each interruption and an improvement when the stimulator has been turned back on [ 414 ]. DBS appears to induce a long-term therapeutic effect as reported by Leone et al. in 70% of patients with a median follow-up of more than 8 years [ 415 ]. To date, the cost-effectiveness of DBS in rCCH has not been evaluated. Given the changes in the stimulation target over time and between the different teams, a meta-analysis aimed to construct a probabilistic stimulation map of effective DBS and identified 2 hotspots of stimulation covering the midbrain ventral and retrorubral tegmentum [ 415 ]. Identifying the optimal stimulation target is essential to ensure the best therapeutic effect, but also to understand the DBS mechanism of action in CH. Studies using brain PET [ 170 ] and quantitative sensory testing [ 168 ] argue against an unspecific antinociceptive effect or a pure inhibition of an overactive central generator but suggest functional modulation of a pain-processing network. In addition, the efficacy of DBS in CH is supposed to depend on stimulated volumes that results in slow modulation of this network, in line with the long latency to obtain a preventive effect (minimum of 1-3months and even longer) and evidence that acute DBS is not useful to stop CH attack [ 416 ]. Major complications have been reported in 17% of subjects, including one death of day 3 of surgery due to intracerebral hemorrhage [ 417 ]. Software-related adverse events most commonly are gaze disturbances, dizziness, in general reversible with adjusting stimulation parameters. Hardware-related adverse events are electrode migration or breakage, wound dehiscence, skin erosion, and battery failure, that may need reoperation. Surgical-related adverse-event are lead misplacement, infection and intracerebral haemorrhage [ 418 ]. The principle of ONS is to deliver continuous stimulation to the greater occipital nerve and/or to the lesser occipital nerve, via a subcutaneous implanted electrode adjacent to the nerve, producing paraesthesia that should cover the occipital region. Originally described by Weiner [ 419 ] to control occipital nerve neuralgia, this technique has been proposed to treat primary headaches, including chronic CH, ONS was initially considered to be acting through a ‘gate control theory-like’ mechanism involving a modulation of convergent nociceptive inputs in the TCC [ 420 ]. Twelve studies involving a total of more than four hundred patients were published and the meta-analysis of the reported data estimated the pooled response rate of 57.8% (OR 0.573 [95% CI 0.481–0.665]) [ 412 ]. The data included in this meta-analysis are mainly derived from open studies due to the paraesthesia induced by the stimulation, which makes it difficult to conduct controlled studies. Notwithstanding, these challenging hindrances have been overcome by the ICON Study, based on the hypothesis of a dose-response efficacy, in which a very low-intensity but still sufficient to induce paraesthesia have been used as sham versus a full intensity stimulation [ 421 ]. Surprisingly both 100% ONS intensity and 30% ONS intensity resulted substantially with the same rate of success in reducing attack frequency [ 422 ]. ONS appears to induce a long-term therapeutic effect as reported by Brandt et al. with a follow-up of 2–8 year [ 423 ] and by Leplus et al. with a mean follow-up of 43.8 months [ 424 ]. Regarding its cost-effectiveness, ONS is also dominant in a long-term perspective [ 425 ]. A recent study performed in rCCH patients treated by ONS showed that voltage tuning may cease and/or terminate CH attacks [ 426 ]. Despite its uncontrolled conditions, this study raises interests in the eventual use of ONS for the acute treatment of attacks and supports the ‘gate control theory-like’ mechanism of action of ONS. Software-related adverse events are local pain and neck stiffness. Surgical-related adverse-events are infection that is local but requires removal of the device. Hardware-related adverse events are mainly electrode migration (due to neck movements) and battery exhaustion (due to high level of power consumption) that need reoperation. The risk of both hardware-related adverse events could be limited by use of a dedicated lead (Ankerstim ® developed by Medtronic with a specific CE mark for CH treatment) and rechargeable implantable pulse generator respectively. Hypothalamic dysfunction is assumed in CH pathophysiology to cause secondary parasympathetic activation involving paraventricular nucleus, superior salivatory nucleus and sphenopalatine ganglion. Given its easy access in the pterygopalatine fossa, the SPG was considered a potential target for stimulation to induce an inhibition of postganglionic parasympathetic outflow and stop CH attacks. With this aim, a chronically implantable neuromodulation device (Pulsante ® ), specifically designed for acute SPG stimulation, has been developed by the company Autonomic Technologies (ATI) and had been the subject of clinical development in the treatment of rCCH. A multiple CH attack was conducted in 28 rCCH patients who randomly treated 566 attacks with full, sub-perception or sham stimulation and the proportion of treated attacks with pain relief at 15 min following SPG stimulation was significantly greater with full than with sub-perception and sham stimulation (67.1%, 7.3% and 7.4% respectively, p  < 0.0001) [ 138 ]. In a long-term study lasting 24 months and involving 33 rCCH patients, SPG stimulation was an effective acute therapy for 45% of participants, and 33% had decreased frequency of attacks [ 300 ]. In a real-world study involving 88 CH patients considered as difficult to treat, 68% of patients had improvement, with 55% having a 50% reduction in attack frequency at 1 year and 32% of patients experiencing acute pain relief in at least 50% of the treated attacks [ 427 ]. SPG stimulation is often presented as the invasive neuromodulatory technique with the highest level of evidence for efficacy in the treatment of CH, but it is important to emphasize that this high level only applies to its use as an attack treatment. Similarly, SPG stimulation is often presented as a minimally invasive neuromodulation technique, but it is important to remember the high rate of reported adverse effects observed during the clinical development of the Pulsante ® device given 81% with almost facial sensory disturbance, although reversible, and near 20% of reoperation related to lead misplacement or lead migration [ 138 ]. Finally, SPG stimulation has limitations in its high cost (without cost-effectiveness data available) and the specialized surgical facilities required.

The trigeminovascular and cranial parasympathetic systems are activated during CH attacks, leading to the expression and release of vasoactive neuropeptides, including CGRP, PACAP, VIP and NO. This section will review the role of these molecules in CH and will highlight the presence of other potential modulators of CH pathophysiology. CGRP is a neuropeptide that is expressed in the areas involved in the pathophysiology of CH, including the trigeminal ganglia (TG), the trigeminal nucleus caudalis (TNC) and the hypothalamus [ 79 ]. CGRP activates the canonical CGRP receptor that is formed by the calcitonin-like receptor (CLR) coupled to the RAMP1 and receptor component protein (RCP). CGRP also has lower affinity for other receptors, such as the amylin subtype 1 receptor (AMY1), albeit less so in primates compared to rodents [ 80 ]. CGRP binding to its receptor results in the activation of complex intracellular pathways mostly mediated via cAMP signalling [ 80 ]. CGRP has been linked to CH pathophysiology in a variety of studies. CGRP is elevated in plasma in acute cluster headache in both spontaneous [ 81 ] and triggered attacks [ 82 ], and its levels are normalised by oxygen and sumatriptan. CGRP levels are elevated in saliva [ 83 ] and tears [ 84 ] during CH attacks when compared to healthy controls. CGRP provokes CH attacks in patients who are in the active, but not in the remission phase of the disease [ 85 ]. Importantly, treatment with galcanezumab 300 mg s/c effectively reduced the frequency of attacks in episodic CH [ 86 ], although it was not effective in the chronic variant [ 87 ]. Fremanezumab was ineffective in a placebo-controlled trial in episodic cluster headache with a four-week endpoint [ 88 ]. Fremanezumab for chronic CH was also discontinued after an interim analysis showed no efficacy [ 89 ]. Most recently, it has been shown that eptinezumab 400 mg intravenously while not reducing attack frequency against placebo did have a greater 50% responder rate in episodic CH [ 90 ]. In an open-label study in chronic CH, eptinezumab 400 mg was well tolerated and apparently clinically useful over twelve months [ 91 ]. The neuropeptide PACAP is expressed in CH relevant areas, including the trigeminal ganglia, the sphenopalatine ganglia, the otic ganglia, the trigeminal nucleus caudalis and the hypothalamus, among other regions [ 92 , 93 ]. PACAP is found in two different isoforms, PACAP27 and PACAP38, the latter being the predominant mammalian form, that bind to three G-protein-coupled receptors: PAC1, VPAC1 and VPAC2. Binding of PACAP to its receptors activates the adenylate cyclase, inducing the increase of intracellular cAMP [ 94 ]. Binding to the PAC1 receptor increases the levels of intracellular calcium and activates protein kinase C signalling [ 93 ]. PACAP38 has also been found to promote pain behaviour in mice by activating the Mas-related G-protein coupled receptor member B2 (MrgprB2) in meningeal mast cells, suggesting the potential relevance of this receptor for PACAP’s actions [ 95 ]. Like CGRP, PACAP is released during CH attacks. An exploratory study ( n  = 5) found that levels of PACAP38 in plasma are higher during CH attacks when compared to the remission phase, but that PACAP38 levels are more elevated in healthy controls when compared to CH participants in the remission phase [ 96 ]. However, further studies with larger sample sizes are needed to understand the expression patterns of PACAP in the different phases of CH. Other key studies highlight the relevance of PACAP for CH biology. The infusion of PACAP38 provokes CH-like attacks in episodic CH participants in the active, but not in the remission phase, and in chronic CH participants [ 97 ]. Interestingly, PACAP38-induced CH attacks are not linked to changes of plasma CGRP, VIP, tryptase or histamine [ 98 , 99 ]. Currently, there are no treatments targeting PACAP for CH. However, a proof-of-concept trial has shown the efficacy of an anti-PACAP antibody for migraine [ 100 ]. These encouraging results highlight the need to perform further trials focusing on CH patients to increase the therapeutic options for them. VIP is from the same peptide family, secretin-glucagon, as PACAP, with which it shares nearly 70% sequence identity [ 101 ]. Due to this similarity, VIP and PACAP bind to the same receptors: VPAC1 and VPAC2 receptors with a similar affinity, whereas PAC1 with 1000-fold lower affinity for VIP. VIP is expressed in areas that are relevant for CH biology, including the parasympathetic ganglia, specifically the sphenopalatine ganglion [ 92 ], and hypothalamus [ 102 ]. The studies that assess the role of VIP in CH are few, although they indicate a potential implication for its pathophysiology. Like CGRP and PACAP, VIP levels are elevated during CH attacks [ 81 ] and infusion of VIP induces CH-like attacks in episodic CH active phase and in chronic CH participants [ 97 ]. However, VIP-induced CH attacks did not alter the levels of plasma CGRP, tryptase or histamine [ 98 ] and PACAP- or VIP-induced CH attacks are not associated with changes in plasma levels of VIP [ 99 ]. Nitric oxide (NO) is a signalling molecule widely expressed throughout the body that plays a role in several functions that are linked to CH pathophysiology, such as processing sensory information, pain sensitization and circadian regulation [ 103 ]. It is synthesized by three isoforms of nitric oxide synthases (NOS) – endothelial, neuronal and inducible NOS. The neuronal NOS (nNOS) is expressed in areas relevant to CH biology, including the hypothalamus and trigeminal and parasympathetic ganglia [ 104 – 106 ]. Although it is not possible to measure the levels of NO due to its short half-life, its levels have been assessed indirectly by measuring the products of its oxidation, nitrite and nitrate. Two studies have shown higher levels of nitrite and nitrate in CSF and plasma of CH patients in their active period when compared to remission and healthy controls [ 107 , 108 ], indicating the relevance of NO for CH biology. Another argument in this favour is that nitroglycerin, an NO donor, induces CH attacks [ 109 , 110 ]. Adenosine triphosphate (ATP)-sensitive potassium (K ATP ) channels have been linked to the pathophysiology of primary headaches, including CH. K ATP channels are expressed in several tissues, including the CNS, and are key regulators of the vascular tone, among other functions. Studies suggest that K ATP opening increases vasodilation which may activate and sensitize perivascular nociceptors [ 111 ]. This was tested in a study were the infusion of levcromakalim, a K ATP channel opener, induced CH attacks in episodic CH patients in their active phase and in chronic CH patients. However, further studies with larger sample sizes are needed to validate these results and to better understand the role of K ATP channels in CH biology. Melatonin is a hypothalamic hormone that is only secreted in darkness. Its role in CH biology has been suggested due to its anti-nociceptive and anti-inflammatory properties and its implication in circadian and circannual regulation [ 112 ]. Different studies support this potential role, as melatonin levels are lower in CH patients when compared to healthy controls and treatment with melatonin reduced attack frequency in a small randomized placebo-controlled study [ 113 ]. However, larger studies would clarify its involvement in the disease. Orexin A and B are hypothalamic neuropeptides involved in the promotion of arousal and in the regulation of nociception, body temperature, feeding behaviour and neuroendocrine system, among other functions [ 112 ]. Lower levels of orexin A were found in episodic and chronic CH patients, suggesting the potential role of orexins in CH [ 114 ]. In summary, studies support the role of CGRP, PACAP, VIP and NO in the pathophysiology of CH. Further research should focus on understanding the pathways regulated by these and emerging molecules to discover new therapeutic targets and to optimise current therapeutic strategies for CH.

rCCH refers to chronic CH that remains disabling and frequent despite adequate trials of evidence-based, standard recommendation at recommended dose and duration. The burden of refractory cases is markedly high, rCCH patients frequently experience significant impairments, including restrictions in daily functioning, sleep disruption, and comorbid depression [ 453 ]. The exact prevalence of rCCH remains unclear. A recent multicenter study found that over two-thirds (68.2%) of chronic CH patients met criteria for refractoriness, with 60.2% experiencing high disease activity despite treatment [ 8 ]. Similar to refractory migraine, refractory cluster headache is not formally defined in the International Classification of Headache Disorders, 3rd edition (ICHD-3) [ 1 ], emphasizing a gap in standardized diagnostic framework for treatment-resistant cases. The definition of rCCHis pivotal for both clinical management and research, ensuring that patients receiving advanced or invasive therapies are genuinely resistant to standard treatments. Standard preventives like verapamil, lithium, and topiramate often failed, leading many patients to rely on off-label therapies [ 412 ]. Among advanced treatments, onabotulinumtoxinA and ONS showed the most consistent benefit, offering hope for this highly disabled population. Early recognition and specialized care may help improve outcome. Although the ICHD-3, does not formally recognize rCCH as a distinct diagnostic entity, it provides diagnostic criteria for chronic CH, which is defined by attacks fulfilling the criteria for cluster headache that occur for at least one year without remission, or with remissions lasting less than three months [ 1 , 454 ]. Thus, in the case of CCH the characteristic bouts of episodic cluster headache are missing. In 2014, the consensus statement by the European Headache Federation provides the first formal clinical definition of rCCH, bring attention to a critical gap in the classification of patients with chronic CH who fail to respond to standard preventive therapies [ 455 ]. These criteria require at least three severe CH attacks per week despite at least three consecutive trials of adequate preventive treatments have been tested (Table  5 ). To note, treatments that showed efficacy over placebo in RCTs for chronic CH include verapamil, lithium, oral or iv steroids, GONB, topiramate, methysergide, ergots, civamide and long acting triptans. Among them verapamil has better documentation. Interestingly, some medicines may be not available across all European countries. However, the criteria, particularly regarding treatment resistance, must remain adaptable to emerging therapeutic advances, especially when the CGRP monoclonal antibodies gain validation for cluster headache prevention in the market. Table 5 European headache federation (EHF) diagnostic criteria for rCCH A Headache attacks fulfilling the ICHD-3 beta criteria for chronic CH, or probable CH and B-E criteria. B At least three severe CH attacks per week that impact patients’ quality of life despite preventive or symptomatic treatment C Failed consecutive prophylactic treatment trials with at least three agents that showed efficacy over placebo in randomized controlled studies, used at the maximum tolerated dose over a sufficient period of time D Symptomatic chronic CH is ruled out by negative investigation with brain MRI and MRA, eventually supplemented with carotid CT angiograms or triplex carotid ultrasound E Not better accounted for by another ICHD-3 beta diagnosis European headache federation (EHF) diagnostic criteria for rCCH Once a patient is diagnosed with rCCH—clinicians must consider alternative therapeutic options beyond standard pharmacological therapy [ 342 ]. These include warfarin, psilocybin, intravenous dihydroergotamine, galcanezumab and onabotulinumtoxin A, along other options. However, most patients with rCCH will need to try neuromodulation approaches. Warfarin showed good efficacy (number needed to treat of 2.6) in a randomized, crossover, placebo-controlled trial [ 456 ]. Warfarin also significantly reduced attack frequency, duration, and intensity. Although the exact mechanism remains unclear, proposed actions include modulation of neurogenic inflammation and hypothalamic pathways. While the only RCT to date showed a non-significant reduction in attack frequency [ 386 ], small open-label and extension studies reported significant improvements with psilocybin, with 30–50% reductions in attacks and prolonged remission [ 387 ]. Though early evidence is promising, larger trials are needed to confirm efficacy and guide safe, targeted use in this difficult-to-treat population. Retrospective data indicate that intravenous DHE may be effective for rCCH, particularly the episodic form [ 362 ]. DHE acted rapidly and was generally well tolerated, with few adverse events. Galcanezumab, a CGRP-targeting monoclonal antibody, has shown efficacy in one RCT for episodic CH but not conclusively in chronic CH [ 457 ]. However, real-world data suggest potential benefit of high doses of galcanezumab (240-360 mg monthly) in refractory CH cases and should be included in the armamentarium for rCCH [ 458 – 460 ]. Observational, real-world studies have provided evidence for the efficacy of onabotulinumtoxinA in rCCH, which was administered according to the PREEMPT [ 461 ]. Whether cannabinoids are helpful or not in rCCH remains debatable [ 462 ]. Because PACAP is expressed with the parasympathetic system, treatments targeting the PACAP signaling may be effective in cases of rCCH, but this remain to be proven in clinical tests [ 463 ]. Ketogenic diet has been reported to help patients with rCCH as well [ 412 ]. The field of neuromodulation is emerging as a promising and alternative therapeutical option for rCCH, offering relief when other treatments have failed. Neuromodulation includes invasive and non-invasive techniques targeting either central or peripheral parts of the cephalic nervous system including autonomic circuits that are involved in CH, particularly abnormal activity in the posterior hypothalamus and trigeminal-autonomic pathways [ 412 ]. A recent meta-analysis found that among invasive neuromodulation techniques for rCCH, ONS was the most studied with a pooled response rate of approximately ∼57%. The highest pooled response (77%) was obtained by hypothalamic DBS which was the second most frequent technique tested for rCCH. DBS data were more heterogeneous than ONS, since the stimulation targets differ in the studies included in the analysis. Furthermore, DBS studies reported more serious adverse events than in ONS studies [ 412 ]. In a 9-year follow-up study invasive occipital nerve stimulation (iONS) provided sustained relief in rCCH with a 70% average reduction in attack frequency and 40% of patients shifting to an episodic form. However, 33% required device removal due to complications, and all patients needed additional surgeries [ 464 ]. Other neuromodulation techniques include serial GONB, sphenopalatine ganglion radiofrequency or stimulation, vagus nerve stimulation, percutaneous bioelectric current stimulation and upper cervical cord stimulation [ 412 ]. Overall, noninvasive and peripheral neuromodulation (e.g. vagus nerve or sphenopalatine ganglion stimulation) have lower-level evidence for rCCH.

Despite pharmacological advancements, many CH patients remain refractory to conventional treatments. As a result, there has been growing patient-driven interest in alternative medical approaches, including herbal therapies, dietary supplements, psychedelics and compounds such as sodium oxybate. We evaluate the current evidence and therapeutic potential of these alternative approaches in the management of CH (Table  4 ). Table 4 Alternative treatment options in cluster headache: an overview abbreviations: CH: cluster headache; IV: intravenous; IN: intranasal; GI: gastrointestinal; RCT: randomized controlled trial; mg: milligram; kg: kilogram; IU: international units; LSD: lysergic acid diethylamide Substance Administration and dosing Preventive or abortive CH subtype Level of evidence Summary of outcomes Herbal therapies Kudzu Oral, tablets or decoction of dried root. Typically 1–3 times daily, dose range: 613–1500 mg Preventive Chronic, episodic Case series May reduce attack frequency, intensity and duration in some. Well tolerated, mild GI side effects Cannabis Inhaled or oral (e.g. tablets dronabinol 5 mg) Primarily abortive Chronic, episodic Case report, survey studies Highly variable. Some report quick pain relief (~ 26–29%), others worsening of attacks (~ 22–25%) or no benefit Dietary supplementals Vitamin D Oral, tablets 10,000 IU/day Preventive Chronic, episodic Survey study May reduce attack frequency, severity, and duration. Toxicity risks with high-dose regimens Magnesium IV, 1000 mg (repeated in responders) Preventive Mainly episodic Case series Possible benefit primarily in cases with low ionized magnesium. Small sample size, unclear outcome measures and study design limit generalizability Melatonin Oral, tablets 9–10 mg nightly (case reports use 3–15 mg) Preventive Mainly episodic RCT, single-blind crossover, case series and reports Reduces attack frequency in some, particularly in episodic CH. Well tolerated Sodium Oxybate Oral, tablets 3–9 g/night in 2 doses. Titration required. Preventive Chronic, episodic Exploratory RCT, open label trial, case series and report Possible reduction of (nocturnal) attacks and improved sleep; acceptable tolerability. Small RCT showed benefit at 1 week but not at 2, likely due to dropout Psychedelic compounds Psilocybin Oral, typically as capsules or dried mushrooms, ‘pulse dosing’ 3 doses (e.g. 0.143 mg/kg psilocybin) every 5 days Preventive Chronic, episodic Exploratory RCT, surveys, case series and report Possible attack frequency reduction. Larger effect in chronic CH. In some, effects are prolonged. Well-tolerated in trial settings LSD Oral, typically as liquid or blotter. Both psychedelic and non-psychedelic doses. Dosing intervals range from once every 5–7 days to monthly. Preventive Chronic, episodic Case series, survey studies Indications of attack cessation, prolonged remissions, also at sub-hallucinogenic doses Ketamine IV: 0.25-0.5 mg/kg; single or up to 4 biweekly infusions IN (spray): 15 mg every 6 min up to 5 times (max. 75 mg) IV: preventive IN: abortive Mainly chronic Open label trials, case-series Preventive: ≥50% reduction in 54–76% of chronic patients, effects lasting weeks to months reported Abortive: no effect at 15 min, but 59% mean pain reduction at 30 min. No serious adverse events reported BOL-148 Oral, tablet, 3 mg every 5 days Preventive Mainly chronic Open label trial Non-hallucinogenic LSD analogue; reduces attack frequency and/or intensity in preliminary data. Well tolerated Alternative treatment options in cluster headache: an overview abbreviations: CH: cluster headache; IV: intravenous; IN: intranasal; GI: gastrointestinal; RCT: randomized controlled trial; mg: milligram; kg: kilogram; IU: international units; LSD: lysergic acid diethylamide IV: 0.25-0.5 mg/kg; single or up to 4 biweekly infusions IN (spray): 15 mg every 6 min up to 5 times (max. 75 mg) IV: preventive IN: abortive Preventive: ≥50% reduction in 54–76% of chronic patients, effects lasting weeks to months reported Abortive: no effect at 15 min, but 59% mean pain reduction at 30 min. No serious adverse events reported Evidence for plant-based compounds in CH is observational, with cannabis and kudzu ( Pueraria lobata ) being the most studied. Kudzu root is traditionally used in Chinese medicine. In a survey of 235 CH patients, 16 self-administered kudzu. Among these, 69% reported reduced attack intensity and 56% a decrease in attack frequency [ 370 ]. Notably, a dose-response relationship was observed​. Side effects were minimal, mostly mild gastrointestinal discomfort. Cannabis use is more prevalent among CH patients compared to the general population [ 371 , 372 ]. Research on its therapeutic potential is scarce and mainly focused on cannabis as an abortive treatment [ 373 – 375 ]. One case experienced symptom alleviation within 5 min after inhaling marijuana and 15 min after an oral dose of dronabinol 5 mg [ 373 ]. In survey studies, 25.9 and 29.4% of users experienced attack relief, while most reported either no benefit or worsening of attacks [ 374 , 375 ]. These findings underscore the variability in reported efficacy and cannabis’ potential to worsen symptoms [ 371 ]. Evidence for dietary supplements in CH management remains limited compared to more established data in migraine prophylaxis. Some studies have explored potential roles for vitamin D and magnesium. Vitamin D deficiency is notably prevalent in CH patients [ 376 ]. High-dose supplementation (10,000 IU/day) with Omega-3 fish oil was associated with reductions in attack frequency, severity and duration in a presented abstract [ 377 ]. However, clinical trials are lacking and high-dose vitamin D treatment should include medical monitoring and personalized dosage adjustments due to toxicity risks [ 378 ]. Intravenous magnesium sulphate provided relief in 41% of 22 patients, primarily those with low baseline ionized magnesium levels [ 379 ]. Yet methodological limitations, including the open-label design, unclear outcomes and small sample size, limit interpretability. Oral magnesium supplementation, although researched in migraine, lacks evidence for CH. Intravenous magnesium combined with ketamine is addressed under ‘ Ketamine’ . The sodium salt of gamma-hydroxybutyrate (GHB) exhibits GABAergic activity and is primarily used in narcolepsy. Its sleep-modulating effects have prompted exploration in CH [ 236 , 380 , 381 ]. A case report in episodic CH and an open-label study in four chronic patients reported reduction in attack frequency, mainly nocturnal, alongside improved sleep parameters [ 236 , 380 ]. Side effects such as dizziness, vomiting, and weight loss were reported​. A small randomized, double-blind trial including eight episodic and chronic patients also found a significant reduction in nocturnal attack frequency after one week of treatment [ 381 ]. However, effects lost statistical significance after two weeks, possibly due to patient dropout (predominantly among placebo-treated individuals with increased headache attacks). Larger controlled studies are necessary and planned to validate these preliminary findings ( NCT06950281 ). Psychedelic compounds are potent neuroactive substances that influence perception, mood, and cognition [ 382 ]. Classic psychedelics, such as lLSD and psilocybin, mainly act through serotonin receptor agonism. They are distinguished from non-classical compounds like ketamine, which share clinical similarities, but have different mechanisms of action. Anecdotal reports and patient advocacy groups have long suggested that psychedelics may abort CH attacks, prolong remission periods, and reduce attack frequency [ 383 ]. Most data originate from uncontrolled case studies or surveys, which notably describe beneficial effects from certain psychedelics, particularly psilocybin and LSD [ 303 , 371 , 375 ,  383 – 385 ]. Sub-hallucinogenic doses appear to retain therapeutic benefits while minimizing psychoactive effects [ 303 ]. Moreover, the beneficial effect of these psychedelics may be prolonged, lasting longer than would be expected based on their pharmacokinetics. Clinical trials are sparse, but the concept of ‘pulse dosing’ involving repeated low doses spaced over days, has been investigated for psilocybin. In one study of 14 CH patients, a low-dose psilocybin regimen of three doses five days apart led to non-significant attack frequency reduction compared to placebo (−3.2 vs. −0.03 attacks per week) [ 386 ]. A larger effect was observed in participants with chronic CH. A blinded extension demonstrated approximately 50% reduction in attack frequency following repeat dosing, even in initial non-responders [ 387 ]. An open label trial including 10 patients with chronic CH and employing a similar dose of psilocybin, dosed every week for 3 weeks, showed a significant but variable reduction in attack frequency (~ 31%) compared to baseline [ 388 ]. Across these studies psilocybin was well-tolerated, without serious adverse effects. While no controlled LSD trials have been published, two randomized studies are currently ongoing, employing hallucinogenic and sub-hallucinogenic doses ( NCT03781128 and NCT05477459 . respectively). Other serotonergic psychedelics, such as Lysergic acid amide (LSA) and N, N-Dimethyltryptamine (DMT) are used by CH patients [ 375 , 389 ]. LSA as a prophylactic, DMT is also employed as abortive treatment. Both have been reported as beneficial, although controlled studies are lacking. This NMDA receptor antagonist’ therapeutic potential has primarily been evaluated in case series among refractory chronic CH patients, with varying dosing regimens [ 390 – 393 ]. Two series combining intravenous ketamine (0.5 mg/kg over 2 h) with magnesium sulphate showed ≥ 50% attack frequency reduction in over 75% of patients, with remissions lasting up to several weeks​ [ 392 ]. Another series using intravenous ketamine alone (0.5 mg/kg every 2 weeks, up to four sessions) reported termination of episodes in all episodic patients and temporary remission of attacks in 54% of chronic patients, with effects lasting 3–18 months [ 391 ]. A recent case series utilizing intravenous S-ketamine (0.25 mg/kg over 1 h) showed complete remission in 80% of participants, although remission duration varied significantly (1–26 months) [ 393 ]. Intranasal ketamine (15 mg every 6 min, max. 75 mg) as acute treatment in chronic CH did not meet its primary endpoint of ≥  50% pain reduction after 15 min. However, among 16 patients who refrained from using conventional rescue medication after 15 min, the mean pain reduction reached 59% at 30 min​ [ 390 ]. Across these studies, ketamine was generally well-tolerated, with mild and transient side effects. To our knowledge, no controlled studies have been published. However, a placebo-controlled trial investigating ketamine and magnesium infusion in chronic CH is now recruiting ( NCT04814381 ). This non-hallucinogenic structural analogue of LSD, demonstrated efficacy in a small case series in five CH patients, four of whom chronic [ 394 ]. BOL-148 substantially reduced attack frequency in four out of five patients. The patient who did not experience a significant reduction in attack frequency reported a decrease in attack intensity. BOL-148 was well-tolerated and not associated with psychoactive side effects. The accumulating evidence suggests that certain psychedelics, BOL-148, selected herbal and dietary supplements, melatonin and sodium oxybate, may offer a promising alternative or adjunctive treatment for CH. However, robust clinical validation is lacking and significant barriers, including regulatory challenges, persist. Future research should focus on clarifying the precise mechanisms of these compounds and developing safe, legally accessible treatment options for CH patients. Neuromodulation has emerged as a promising alternative to pharmacological treatments in primary headache disorders with some applications also in CH. Invasive modalities are usually offered for medically refractory patients [ 395 ]. Non-invasive modalities may be used earlier in the patients’ treatment pathway as an alternative or addonto pharmacological therapies [ 396 ]. Barriers to the wider use of these therapies have historically been the lack of robust evidence, largely dictated by the difficulty in producing a reliable sham in clinical trials and the high costs of the devices, which have discouraged some health authorities to cover the cost of these therapies, thereby limiting their wider use [ 397 ]. Stimulation of the cervical branches of the vagus nerve to modulate pain and other symptoms in CH has been attempted by using the gammaCore device. This is small, portable device containing two electrodes that transcutaneously stimulate the vagus nerve with a 1 ms burst of five kHz sine wave (24 V peak, 60 mA) stimulation repeated at 25 Hz for 90 s [ 398 ]. The vagus nerve provides parasympathetic innervation to the autonomic nervous system, involved in a variety of autonomic functions including the respiratory, cardiovascular, and nociceptive systems. This stimulation is thought to activate lowthreshold myelinated Afibers, producing an antinociceptive effect on the secondorder neurons of the spinothalamic and spinoreticular tracts within the TCC [ 399 ]. Transcutaneous vagus nerve stimulation (t-VNS) in animals has been shown to be able to modulate trigeminal activity and to inhibit the response of TCC neurons to trigeminal stimulation [ 400 ]. An fMRI study that investigated the neuromodulatory role of vagus nerve stimulation to the trigeminal autonomic reflex, showed that t-VNS may modulate functional connections between important areas for head pain and CH namely the pons, the spinal trigeminal nuclei and the hypothalamus [ 135 ]. The therapeutic effect of t-VNS in CH has been evaluated in two randomised-controlled trials, an open-label prospective study and real-world studies. Two randomized doubleblind shamcontrolled trials (ACT1 and ACT2 studies) were conducted to evaluate the efficacy of t-VNS using gammaCore device in CH [ 401 , 402 ]. The primary endpoint of ACT1 was the response rate, defined as the proportion of subjects who achieved pain relief (pain intensity of 0 or 1) at 15 min after treatment initiation for the first CH attack without rescue medication use through 60 min. The study did not meet the primary endpoint. Indeed, a response was achieved in 26.7% of t-VNS treated subjects vs. 15.1% of sham treated subjects ( P  = 0.1). ACT2 was a smaller study compared to the ACT1 but with a similar methodological design. The primary efficacy endpoint was the proportion of all treated attacks that achieved pain-free status within 15 min after treatment initiation, without rescue treatment. Even in this trial, t-VNS (14%) was not superior compared to sham (12%) treatment in the total cohort. In both trials, t-VNS was superior to sham in the episodic CH group compared to chronic CH group. Given that the current pathophysiological hypotheses for CH suggest that the two subtypes of CH follow the same biology, these results are difficult to explain. When evaluated as a prophylactic treatment in an open label small study in chronic CH, t-VNS plus standard of care (SoC) was superior to SoC alone at least for short term prevention (40% vs. 8.3% responders) [ 403 ]. A company funded small retrospective audit conducted in three headache services in the United Kingdom (UK), found positive results in chronic CH patients treated with gammaCore. However, these patients were known responders to t-VNS. Furthermore, the GONB, which is the preventive treatment that has amongst the best evidence of efficacy and tolerability in episodic and chronic CH, was not tried before using t-VNS in this study [ 404 ], suggesting that the patients evaluated in this study were non-difficult-to-treat. When used the refractory chronic CH patients, an independent realworld prospective analysis coming from a single headache centre in the UK suggested the poor effectiveness of t-VNS in their patients, with only one patient out of 12, responding for up to a year to gammaCore therapy [ 405 ]. A subsequent open-label study in a similar population showed a 50% response rate in 43% of patients. However, the follow-up duration of this study was only 3 months, which does not allow to draw any conclusion on the sustained effect of this therapy overtime [ 406 ]. In small open label studies, also other non-invasive stimulation techniques have been applied. Transcranial magnetic stimulation (TMS) has been tested in an open-label observational study including 19 patients suffering from chronic CH. A beneficial outcome of repetitive (10 Hz) TMS over the M1 area was reported. Paroxysmal pain and number of attacks were reduced when comparing baseline with 15 days poststimulation measures [ 407 ]. Although in this report permanent pain was reported in eight of the CH patients, it remains unclear if those fulfilled the definition of chronic CH; therefore, no clear data are available on the use of TMS in chronic CH. Transcranial direct current stimulation (tDCS) (anode at Fz, cathode over C7) for refractory chronic CH patients has only been studied in a proof-of-concept study [ 408 ] with a mean attack frequency reduction by 35%. The method was well tolerated. There are no more data available on this method in chronic CH at the moment. Thirty-six patients with chronic CH were treated by transcutaneous electrical nerve stimulation (TENS) of the GON for 8–12 consecutive weeks [ 409 ]. Weekly attack frequency decreased from 15.7 at baseline to 11.0 with TENS. 13 of the 36 (36%) patients had a minimum 30% reduction in attack frequency. Not all patients benefitted from TENS, but the treatment responders had a substantial improvement in their CH. In conclusion, t-VNS is currently the only non-invasive neuromodulation therapy with some controlled evidence in CH. The available body of evidence suggests that t-VNS may be effective as an acute attack therapy in patients with episodic CH. However, it is unclear whether t-VNS would be effective in patients who fail to respond to both subcutaneous sumatriptan and oxygen inhalation, which are very effective and more cost-effective than gammaCore. Open-label evidence on prevention of chronic CH suggest that t-VNS might be beneficial in some patients, but not in those with refractory chronic CH, at least long-term. Further methodologically robust randomised controlled trials are warranted to clarify its role in the arsenal of treatments for CH.

CH is characterized by several key clinical factors, including the severe unilateral pain associated with autonomic symptoms and notable agitation, the higher prevalence in males, and specific circadian and circannual patterns of recurrence. Despite the extensive previous efforts, our therapeutic arsenal remains constrained to medications targeting non-specific pathways. This is attributable to both the rarity of the disease and the stagnation in pathophysiological knowledge. Indeed, the pathophysiological evidence for CH closely mirrors that of migraine, including significant involvement of the trigeminal-autonomic system and hypothalamic activation. In particular, hypothalamic dysfunction, likely in the suprachiasmatic nucleus, seems to play a major role in determining the circadian recurrence of attacks, influencing the temporal aspects of the attacks. Several other cerebral regions, including the brainstem and mesocorticolimbic networks, play an additional role, particularly in cases of chronic CH. Nonetheless, when these neurolimbic structures are suppressed using invasive techniques like DBS, the positive effects do not manifest instantaneously, but rather with a delay of several months, suggesting that they are not the locus of the primary dysfunction. Several molecules with a pivotal role in migraine pathophysiology, including CGRP, PACAP, VIP, and nitric oxide, seem to influence the modulation of pain and autonomic symptoms also in CH. However, the evidence that their plasma levels do not correlate with attack frequency and the discouraging outcomes obtained with the administration of CGRP-blocking antibodies suggest that some crucial biological differences exist between these two primary headache disorders. While CH and other primary headaches, especially migraine, exhibit shared genetic and pathophysiological pathways, they possess distinct clinical features and treatment responses, which makes it impossible to consider them as a continuum. Consequently, we should change our perspective by refraining from directly translating the insights gained from the research of other main types of primary headaches into CH pathophysiology. Additional research is required to ascertain not only which brain structure(s) initiates the attack, but also which one(s) is in charge of stopping it, and what the relationships are with the hypothalamic master clock and the subordinate clocks distributed throughout the human body. Then, further efforts should be made by all the researchers worldwide, focusing also on the distinct nature of pain associated with a CH attack compared to other primary headaches. This includes understanding the pathophysiological reasons for its severity, as well as its strict lateralisation, the biology under the higher male prevalence, or the loss of certain clinical and pathophysiological characteristics when CH becomes chronic. Comprehending these pathophysiological elements may facilitate the identification of novel molecular and structural targets, which may fuel the development of drugs that operate more rapidly and effectively. In conclusion, CH is characterised by a complex, multifactorial, and yet inadequately understood pathophysiology. Precise diagnosis, additional research studies, and robust psychosocial and institutional support are necessary to improve the quality of life for individuals affected by this debilitating condition. And here, where the Hallmarks of Primary Headache Disorders trilogy appears to reach its conclusion with Cluster Headache, the project evolves instead, leaving a deliberate opening for the completion into a five-act masterpiece. We’re all set to appreciate the final two acts.

Increasing attention is being paid to the potential role of psychopathological factors, including personality traits, in determining both the expression and progression of CH. This shift reflects the principles of the biopsychosocial model of health, which posits that disease arises from the complex, multidirectional interactions among biological, psychological, and social factors [ 247 ]. In this context, examining personality traits [ 248 ] in CH may offer valuable insights into behavioural tendencies and lifestyle patterns commonly observed in this population. Certain traits could influence pain perception, treatment adherence, and overall outcomes. Moreover, identifying personality profiles might support the development of more tailored interventions, particularly in chronic sufferers. In this section, we review the current literature on personality features and related psychopathological aspects in CH. Although depression is one of the most common psychiatric comorbidities in CH [ 249 ], recent studies have shifted attention toward more stable personality features that may underlie or interact with emotional vulnerability in this population. CH patients have been found to score significantly higher on alexithymia [ 250 ], a personality construct defined by difficulties in identifying and expressing emotions [ 251 ]. This impaired emotional insight may contribute both to emotional vulnerability and to maladaptive coping strategies in the context of chronic pain. Further supporting this view, a Rorschach-based study [ 252 ] confirmed the presence of alexithymic tendencies in CH, highlighting features such as concrete and stereotyped thought, reduced emotional responsiveness, and limited relational adaptability, patterns commonly observed in psychosomatic profiles [ 251 , 253 ]. These emotional and relational patterns have led researchers to explore whether CH might also be linked to more enduring personality configurations that go beyond episodic distress. In line with this perspective, Mongini et al. [ 254 ] investigated the impact of personality traits on pain perception in primary headache patients using the Minnesota Multiphasic Personality Inventory – Second Edition. Those patients identified as having an “emotionally overwhelmed” profile reported higher affective pain scores, regardless of headache severity. This suggests that certain personality characteristics may intensify the emotional experience of pain, possibly through mechanisms like poor affect regulation or heightened attention to bodily sensations. These traits may not simply coexist with CH but actively shape its subjective burden. While no single personality profile appears to define CH, certain traits recur across studies. An historical work by Cuypers [ 255 ], using the Freiburg Personality Inventory, suggested elevated nervousness and diminished masculinity in CH patients compared to migraineurs. Using the Karolinska Scales of Personality, Levi et al. [ 256 ], highlighted that patients with episodic CH showed elevated levels of anxiety, reduced socialization, and greater hostility, in line with subsequent research. More recently, Piacentini [ 257 ], using the Millon Clinical Multiaxial Inventory-III, reported that over 90% of CH patients showed clinically significant personality traits, with a predominance of obsessive-compulsive, histrionic, narcissistic, and avoidant features. These same findings were recently confirmed by Telesca [ 258 ] highlighting that such personality features represented the most pervasive traits in CH. Interestingly, the authors also identified three distinct psychological profiles based on disease characteristics and personality measures. These profiles reflected marked heterogeneity within the CH population, and included, among others, a subgroup characterized by emotional dysregulation and low social engagement. This variability suggests that CH patients cannot be considered as a psychologically homogeneous group, but rather require individualized clinical attention based on their specific psychological features. This heterogeneity was also evident in the comparative study by Munoz [ 259 ] who analysed the personality traits of 80 CH patients and 164 migraine patients using the Salamanca screening test. CH patients most frequently exhibited traits from Cluster A (odd or eccentric), while Cluster C traits (anxious or fearful) were more common among migraineurs. For instance, 53% of CH patients had anankastic (obsessive-compulsive) traits and 43% had schizoid tendencies. This pattern suggests a paranoid-schizoid position in many CH patients, which may influence their response to pain and determine inadequate copying strategies. Importantly, these findings support the idea that emotional reactivity, difficulty in emotional disclosure, and rigidity in interpersonal dynamics may be central features in the personality profile of CH sufferers. One particularly interesting aspect of personality functioning in CH concerns the direction in which emotional suffering is expressed. While it has often been assumed that patients with CH might be more prone to externalized anger or irritability, the empirical evidence offers a different pattern. A multicentre controlled study [ 260 ] found no increase in overt or outward-directed aggression among CH patients. Instead, what emerged was a significant rise in self-directed aggression, especially in individuals with chronic CH. This internal hostility was closely linked to greater headache-related disability. These data suggest a reciprocal relationship between emotional suffering and disease burden, where psychological distress both reflects and potentially exacerbates the severity of the condition. These emotional dynamics may also contribute to changes in social behaviour and interpersonal functioning. From a social point of view, many CH patients appear to experience a reduction in social support and relational integration. As noted by Blomkvist [ 261 ] they tend to have fewer close connections and lower expectations for future social involvement, suggesting a more withdrawn and self-reliant coping style. Interestingly, this is sometimes accompanied by an optimistic outlook, which may serve as a compensatory mechanism to preserve a sense of control. This psychological profile often coexists with behavioural patterns that may further impact disease management, including unhealthy lifestyle habits and risk-prone behaviours [ 262 ]. Manzoni [ 263 ], in a sample of 374 CH patients, found high rates of these behaviours, especially among those in self-employed or high-responsibility occupations. Similarly, the Danish Cluster Headache Survey [ 264 ] confirmed elevated substance use among CH patients compared to controls, suggesting a possible propensity toward impulsivity or compulsivity. Lambru and Matharu [ 265 ] explored the link between head trauma and CH, proposing that distinctive personality traits, such as sensation-seeking or emotional dysregulation, might increase the risk of injuries or substance abuse. Though speculative, these interpretations could be linked to earlier findings by Saper [ 266 ] and Graham [ 267 ], who described CH patients as emotionally intense, anxiety-prone, and prone to “type-A” behavioural patterns. An extensive genome wide association study in 4,777 cluster headache patients identified genetic risk variants, showing that cigarette smoking, risk-taking behaviour, attention deficit hyperactivity disorder (ADHD), depression, and musculoskeletal pain were significantly associated to CH [ 15 ]. Impulse behaviours may be associated with the underlying biological basis found in ADHD, further studies are needed to clarify this issue. Taken together, the evidence reviewed highlights a complex and nuanced psychological profile in CH patients, where specific personality traits may interact with emotional regulation, behavioural tendencies, and social functioning. These patterns, while heterogeneous, could share common features that deserve meaningful clinical implications—particularly in terms of diagnosis, disease burden, and treatment. Although there is no single personality profile specific to CH, several traits—such as obsessive-compulsive, histrionic, paranoid, and schizoid features—appear more frequently in this population than in other headache disorders. Patterns of emotional detachment, impulsivity, and low social engagement are also common and may contribute to increased disability and treatment complexity. While current evidence provides valuable insights, methodological limitations—such as small samples and reliance on self-report tools—warrant caution. Future research should incorporate longitudinal designs and clinician-administered assessments to better capture the role of personality in CH. In very conclusion, what is evident is that a clearer understanding of the most recurrent psychological features in CH —particularly emotional instability, internalized hostility, and social withdrawal —could offer a useful framework for interpreting patient behaviour and guiding more personalized therapeutic approaches.

The hypothalamus belongs to the limbic system and is a small yet highly complex and functionally diverse brain region that is highly interconnected with other parts of the central nervous system and maintains direct connections with the vasculature. Serving as a critical brain-hormonal interface, it regulates and modulates a wide array of physiological systems, many of which may be dysregulated in the pathophysiology of primary headache disorders. These include amongst others body temperature regulation, sleep, food and water intake, autonomic nervous system control, circadian and circannual rhythms, and pain-modulating systems [ 140 , 141 ]. It is noteworthy that the 24-hour biological clock is primarily regulated by the suprachiasmatic nucleus (SCN) in the anterior hypothalamus. CH is known to follow strict circadian rhythm in over 70% of patients, suggesting a possible involvement of this area in the generation of cluster attacks [ 142 ]. Several clinical features of CH, including cranial autonomic dysfunction and the characteristic temporal pattern of the disease, strongly implicate the hypothalamus as a key structure involved in both prodromal manifestations and the initiation of headache attacks [ 141 ]. This hypothesis is further supported by anatomical connections between the hypothalamus and both the trigeminovascular and parasympathetic nervous systems, both of which have been implicated in CH pathophysiology [ 9 ]. Moreover, numerous neuroimaging studies have identified the hypothalamus as a key player in CH pathogenesis, particularly with regard to attack generation and disease chronification, which will be discussed in the subsequent paragraphs. May and colleagues demonstrated increased regional cerebral blood flow in the posterior hypothalamic grey matter ipsilateral to the attack side during nitroglycerin-induced CH-like attacks, as observed using positron emission tomography (PET) [ 143 ]. A similar finding has been reported in spontaneous CH attacks [ 144 ]. It needs to be pointed out that this finding, along with the below mentioned findings linking the cluster headaches with hypothalamic activation, are not in the hypothalamus proper but in the posterior inferior border between the hypothalamus and the tegmentum (Fig.  1 ). This finding has been replicated using arterial spin labeling functional magnetic resonance imaging (fMRI), which demonstrated that CH patients exhibit increased blood flow of the posterior hypothalamus during attacks when compared to either the (in-bout) interictal period or to healthy controls [ 145 ]. Furthermore, in a state dependent (in-bout ictal vs. interictal) resting-state fMRI study, Morelli and colleagues found posterior hypothalamic activation ipsilateral to the headache [ 146 ]. Another study utilizing PET and an opioidergic radioligand demonstrates an inverse relationship between the duration of CH and opioid receptor binding in the ipsilateral hypothalamus, suggesting that descending opioidergic mechanisms within the hypothalamus may contribute to the maintenance of cluster attacks [ 147 ]. Fig. 1 Posterior hypothalamic activation ipsilateral to the side of headache during an attack in comparison to the headache-free state as demonstrated through the use of a positron emission tomography scan. It is noteworthy that functional imaging studies have repeatedly demonstrated activation in the posterior hypothalamic gray matter at the level of the junction with the midbrain. This activation (of this part of the hypothalamus) is not observed in cases of migraine or other forms of experimental headache. The Journal of Headache and Pain needs to secure permission to reproduce the figure originally published in Neurology (10.1212/WNL.52.7.1522) Posterior hypothalamic activation ipsilateral to the side of headache during an attack in comparison to the headache-free state as demonstrated through the use of a positron emission tomography scan. It is noteworthy that functional imaging studies have repeatedly demonstrated activation in the posterior hypothalamic gray matter at the level of the junction with the midbrain. This activation (of this part of the hypothalamus) is not observed in cases of migraine or other forms of experimental headache. The Journal of Headache and Pain needs to secure permission to reproduce the figure originally published in Neurology (10.1212/WNL.52.7.1522) In addition to the role of the hypothalamus in attack generation, other studies sought to examine the long-term structural changes of the hypothalamus in CH. A pivotal study combining functional and structural data found a co-localization of functional activity in the individual attack and an increase in grey matter of the same area [ 148 ]. Voxel-based morphometric studies comparing the in-bout and out-of-bout phases in patients with episodic CH [ 149 , 150 ], as well as the comparison between CH patients and healthy controls [ 151 ], found no significant changes in hypothalamic volume. However, in a larger study utilizing the same algorithm for automatic segmentation of hypothalamic subunits, Ferraro and colleagues identified an increased volume in the ipsilateral paraventricular nucleus and preoptic area of the hypothalamus in patients with chronic CH compared to those with episodic CH [ 152 ]. Their findings are consistent with those reported in the study by May and colleagues [ 148 ] and those of Arkink and colleagues, who also reported an increased anterior hypothalamic volume ipsilateral to the headache side in chronic CH patients compared to healthy controls [ 153 ]. Additionally, bilateral hypothalamic volumes were found to be larger in CH patients compared to both healthy controls and migraine patients, suggesting distinct structural alterations specific to CH disorder [ 153 ]. Furthermore, a diffusion tensor imaging study revealed that CH patients exhibited lower fractional anisotropy and higher mean diffusivity metrics in the hypothalamus relative to healthy subjects [ 154 ]. Magnetic resonance spectroscopy studies in both episodic (in- and out-of-bout) and chronic CH patients demonstrated a reduction in N-acetyl aspartate levels and a decreased choline/creatine metabolite ratio in the hypothalamus compared to healthy controls [ 155 , 156 ]. Several studies have identified aberrant resting-state functional connectivity in patients with episodic CH, both during and between attacks, particularly between the hypothalamus and key regions of the pain-processing network, such as the prefrontal cortex, anterior cingulate cortex, contralateral thalamus, ipsilateral basal ganglia, insular cortex, inferior parietal lobule, parahippocampal gyrus, and bilateral cerebellar hemispheres [ 43 , 146 ,  157 – 161 ]. Moreover, a study examining in-bout interictal episodic CH patients demonstrated reduced functional connectivity between the hypothalamus and the salience network compared to healthy controls [ 160 ]. In contrast to the functional and structural studies, findings in hypothesis-free resting-state imaging have, however, not yet been reproduced. In chronic CH patients, functional connectivity alterations have been observed in the posterior hypothalamus, as well as in diencephalic-mesencephalic regions ipsilateral to the headache side [ 162 ]. In addition to the resting-state fMRI, a task-specific fMRI study investigated how the hypothalamus responds to trigeminonociceptive stimuli and found that the posterior hypothalamus in episodic CH has higher activation out of bout, but decreased activation in bout [ 46 ]. This finding contrasts the previous prevailing hypothesis based on PET studies that during a CH bout, the posterior hypothalamus becomes hyperactive. Instead, it shows that the CH patients have a hyperactive hypothalamus at baseline, and that activation is counterintuitively reduced in the in-bout state, possibly due to neurotransmitters depletion caused by frequency attacks [ 46 ]. The results of task-specific fMRI and PET [ 46 , 143 ] should not be considered as contradictory, as both methods have different temporal resolution and may suggest a disproportionate blood flow vs. oxygen metabolism change [ 163 ]. Nevertheless, both approaches documented functional changes in the hypothalamus during the cluster bouts. Evidence of the hypothalamic involvement in the generation of attacks led to the introduction of deep brain stimulation of the posterior hypothalamic grey matter as a treatment option for patients with intractable CH, particularly those with chronic forms of the disorder without remission periods, where disability is extremely high and alternative treatment options have been exhausted [ 164 ]. This invasive treatment showed a significant long-term effect in about 50–82% of CH patients [ 165 – 167 ]. The mechanism by which hypothalamic stimulation aborts or even prevents CH attacks remains incompletely understood. Psychophysical measurements in patients with hypothalamic stimulation showed a trigeminal dermatome-specific pain modulatory effect [ 168 ], reflecting the close anatomical connection between the hypothalamus and the spinal trigeminal nucleus [ 169 ]. Nevertheless, May and colleagues used PET to study the acute effects of hypothalamic stimulation and found that stimulation of the posterior hypothalamic area induced bidirectional (both excitatory and inhibitory) modulatory effects in brain regions known to be involved in pain processing [ 170 ]. These findings argue against a nonspecific antinociceptive effect or simple inhibition of hypothalamic activity. Instead, they suggest that hypothalamic deep brain stimulation exerts its effects through complex functional modulation of the pain-processing network, underscoring its role as a therapeutic target in cluster headache [ 170 ]. Finally, although the cluster attacks are most likely generated in the hypothalamus, the participation of the peripheral nociceptor is indispensable for the headache. This is evidenced by the neurostimulatory treatment options, such as sphenopalatine ganglion stimulation, which have been shown to prevent clinical cluster headache without a known mechanism of modulating the hypothalamus [ 171 ].

Cluster headache (CH) is a primary headache disorder classified under the trigeminal autonomic cephalalgias (TACs) and it is considered one of the most intense and debilitating pain condition in humans. Clinically, CH is characterised by severe and strictly unilateral pain, predominantly involving the territory of the first division of the trigeminal nerve, accompanied by autonomic symptoms. Attacks of CH generally last between 15 and 180 min, exhibit circadian and circannual recurrence, and may be associated with agitation or restlessness [ 1 ]. Although generally the disease is characterized by episodes of bout – the period in which the patient experiences the recurrence of attacks daily or every other day – and periods without attacks, approximately 15–20% of patients experience a chronic variant of CH with any periods without attacks or remissions that last less than three months. Among them, an indeterminate percentage of patients may develop a drug-resistant type of chronic CH (rCCH), suggesting that other strategies than conventional therapy, such as non-invasive or invasive neuromodulation, are required for some individuals. The historical prominence of CH in males has diminished over the decades, however a distinct male predominance of the illness persists [ 2 – 5 ]. Certain experts ascribe this declining tendency to changes in women’s lifestyles, which more resemble those of men, and to advancements in the diagnostic accuracy for female patients [ 6 ]. Notwithstanding its distinctive clinical presentation, CH frequently remains underdiagnosed and unacknowledged in clinical practice, resulting in a long lag between the onset of the disease and obtaining an accurate diagnosis and appropriate treatment [ 7 ]. Indeed, a recent meta-analysis reported a diagnostic delay of approximately 10 years for CH diagnosis, with markedly prolonged delays in patients with rCCH relative to those with non-rCCH. However, they noted a decrease in this delay over time since the 1960s, with a continuous reduction observed every decade after 2000, suggesting that awareness of this condition has improved among clinicians over the years [ 8 ]. Clinical, biochemical, and electrophysiological studies have recognised the crucial role of the trigeminal-parasympathetic system and the hypothalamus in the aetiology and recurrence of CH pain; nevertheless, the sequence of activation of these separate systems remains unclear. Additionally, while the hypothalamus is primarily recognised for regulating the seasonal and circadian patterns of attacks, additional structures seem to play a role, including the brainstem monoaminergic systems in the midbrain – particularly the dopaminergic system in chronic cases – and the descending pain control systems. However, the exact pathophysiological cascade and structures involved are still not completely understood. This lack of understanding is also witnessed by the restricted therapeutic alternatives, comprising solely sumatriptan and oxygen for acute management, steroids for bridging therapy, and a scarcity of genuinely effective prophylactic agents aside from verapamil. Additionally, despite multiple attempts to use alternative pharmacological and non-pharmacological methods, their use is limited by the lack of controlled clinical trials and their efficacy being restricted to a specific subset of patients. To delineate the hallmarks of CH across several domains, we provide a comprehensive description of several aspects of CH. Through different subsections, each designed to present the aspects that characterize CH-specific traits, this review aims to summarize the most thorough and accurate research available on this topic. These included CH clinical presentation, diagnostic challenges, pathophysiology, and current and novel therapeutic targets. Finally, it concludes by describing the disease burden and advocating for significant improvements in healthcare systems, reducing stigma, and promoting health equity.

Three core features of cluster headache—excruciating pain, autonomic disturbances, and a tendency to occur during sleep—have been investigated through various electrophysiological approaches. These studies have provided objective insights into the pathophysiological processes underlying the clinical presentation of the disorder. Among these features, pain processing has been the most extensively studied, using a range of electrophysiological techniques including the nociceptive blink reflex (nBR), laser-evoked potentials (LEPs), trigeminal evoked responses, and pain-related evoked potentials (PREPs) (Table  1 ). These methods have revealed alterations in sensory and pain processing involving both trigeminal and extra-trigeminal regions in CH [ 199 ]. Although these alterations can be bilateral, most are more pronounced on the symptomatic side and during the active phase [ 200 , 201 ]. Indeed, two studies documented lateralized habituation deficit of the nBR in the active phase of episodic CH [ 202 , 203 ], reflecting an impairment of sensory filtering that could share some similarity with migraine [ 204 ]. Contrasting results were reported by Holle et al., who found no differences in nBR habituation between healthy controls and patients with CH, irrespective of disease subtype or activity stage. However, it is important to note that recordings were performed while patients were receiving acute and prophylactic medications, which may partly explain the discrepancies with earlier studies [ 205 ]. By recording the nBR and the PREP and comparing the headache and non-headache sides, the same authors showed asymmetry of trigeminal nociceptive processing and central facilitation primarily at the brainstem level [ 200 ]. Interestingly, some of these alterations can persist during remission and vary in magnitude depending on the disease phase and chronicity. This evidence suggested a role for the supraspinal central mechanisms in facilitating the alterations of trigeminal nociceptive processing and favouring the transition to the chronic state [ 200 ]. Complementing this, Ellrich et al. [ 206 ] reported bilateral alterations in LEPs, suggesting broader central involvement beyond the side of attack. Indeed, though CH is strictly unilateral in presentation, there is some evidence of broader functional involvement in pain processing. Lozza et al. [ 207 ] used both cranial and extracephalic stimulation (index finger) to assess inhibition of the blink reflex and found that inhibitory deficits were present across both regions, suggesting that brainstem disinhibition is not confined to trigeminal circuits. Although Perrotta et al. [ 208 ] did not stimulate extracephalic sites directly, they identified trait- and frequency-dependent habituation deficits to trigeminal stimulation, implying dysfunction in more widespread central regulatory systems. Consistent with these findings, evidence from interventional studies further substantiates the interpretation of supraspinal dysfunction in pain modulation. Busch et al. [ 209 ] examined the effect of greater occipital nerve blockade on the nBR in chronic CH patients. They observed a reduction in R2 reflex area and increased latency on the injection side only, indicating localized modulation of trigeminal processing. However, this was not accompanied by significant clinical improvement, suggesting that while the reflex arc was physiologically affected, it May not play a direct role in mediating pain relief. These results were not replicated in a more recent study in a group of 33 patients with episodic and chronic CH, where the nBR was recorded before and 30 min, 1 week, and 4 weeks after GON block [ 210 ]. Similarly, Haane et al. [ 211 ] found that although effective in aborting attacks, high-flow oxygen therapy did not alter nBR parameters, highlighting some degree of independence between clinical efficacy and immediate brainstem reflex modulation. In contrast, Jürgens et al. [ 168 ] reported that DBS of the posterior hypothalamus led to bilateral increases in thermal detection thresholds, pointing to a more diffuse supraspinal modulation of pain-related sensory systems. While not a direct nociceptive reflex study, this intervention illustrates how central pain control structures interact with sensory perception in a state-dependent and bilateral manner. Collectively, this evidence supports asymmetric brainstem dysfunction along with a generalized disruption of central pain modulation at the supraspinal level in CH. Table 1 List of neurophysiological studies that investigated the pain processing in subjects with cluster headache. Abbreviations: auc = area under the curve; br = Blink reflex; ch = cluster headache; cCH = chronic cluster headache; dbsh = unilateral deep brain stimulation of the posterior hypothalamus; eCH = episodic cluster headache; ePH = episodic paroxysmal hemicrania; gon = greater occipital nerve; hc = healthy controls; i.v.= intravenously; moa = migraine without aura; n = number; nBR = nociceptive Blink reflex; LEP = laser-evoked potential; PREP = pain-related evoked potentials; tsep = trigeminal somatosensory evoked potentials Publication year and Authors Sample size ( n ) Phase (in-bout/out-of-the-bout) Methods Main findings 1997, Lozza et al. [ 212 ] -eCH, n  = 10 -HC, n  = 10 In-bout BR after conditioning by supraorbital or index finger stimuli -CH: • No changes of R2 threshold, latency or area -After paired supraorbital stimuli: • More rapid R2 recovering in eCH patients on the symptomatic side -After index stimulations: • More rapid R2 recovery on both symptomatic and non-symptomatic sides in eCH compared to HC. Partial reversion of the R2 suppression induced by index finger stimuli in 2 eCH subjects after Naloxone i.v. administration 2003, Van Vliet et al. [ 206 ] -eCH, n  = 28 -HC, n  = 22 -In-bout, n  = 28 -Out-of-bout, n  = 22 TSEP and BR repeated measurement in two time point (in-and out-of-bout) -eCH in-bout: • Increased N2 TSEP latencies on the symptomatic side in comparison to the non-symptomatic side and with the same side out-of-bout • Increased N1, P1 and N2 latencies on the symptomatic side in comparison to HC -eCH out-of-bout: • Increased N1 latencies of both sides in comparison to HC • No differences of TSEP amplitudes and BR latencies in comparison to HC 2007, Ellrich et al. [ 211 ] -CH, n  = 25 (16 eCH, 9 cCH) -HC, n  = 10 -In-bout, n  = 7 -Out-of-bout, n  = 10 LEP -cCH: • Reduced P2 amplitude and delayed N1c in comparison to HC -eCH in-bout: • Reduced P2 latency on the headache side -eCH out-of-bout: • reduced N2P2 ratio on the headache side -CH: • in 19/26 examinations LEP deviated from normative data in HC without any specific pattern of altered parameters 2007, Busch et al. [ 214 ] -cCH, n  = 15 cCH nBR before and after unilateral GON block -cCH: • Decrease of the R2 response areas and increase of R2 latencies after the GON block only on the injection side 2008, Perrotta et al. [ 208 ] -eCH, n  = 27 -MoA, n  = 22 -HC, n  = 20 In-bout nBR elicited on the symptomatic side -eCH: • Deficit of habituation in the R2 and R3 components compared with HC and MoA -eCH: • More pronounced lack of habituation in comparison to MoA 2009, Jürgens et al. [ 217 ] -CH, n  = 26 (11 cCH with DBSh; 15 cCH without DBSh) -HC, n  = 29 cCH Perception and pain thresholds for hot and cold stimuli -cCH with DBSh: • Increased cold pain thresholds at the first trigeminal branch on the stimulated side in comparison to HC, and higher cold detection thresholds in comparison to cCH without DBSh -cCH with DBSh: • No changes after short-term interruption of stimulation 2012, Holle et al. [ 210 ] -CH, n  = 66 (46 eCH, 20 cCH) -HC, n  = 30 -In-bout, n  = 18 -Out-of-the-bout, n  = 28 nBR -CH (independently from the subtypes and the phase): • No differences in habituation in comparison to HC • No side-to-side differences of habituation between headache side and non-headache side 2012, Holle et al. [ 205 ] -CH, n  = 66 (46 eCH, 20 cCH) -HC, n  = 20 -In-bout, n  = 18 -Out-of-the-bout, n  = 28 nBR and PREP -CH (independently from the subtype): • Decrease in nBR latency ratio (headache side/non-headache side) in comparison to HC -eCH: • Increase in AUC ratio in patients with episodic CH in-bout only -cCH: • Decrease N2 latency ratio of PREP 2015, Coppola et al. [ 207 ] -eCH, n  = 18 -HC, n  = 18 In-bout nBR -eCH: • Decrease in reflex area on both sides in comparison with HC • Deficit of habituation only on the affected side • Positive correlation between the habituation slope and the number of days since the onset of the bout and the daily attack frequency 2016, Haane et al. [ 216 ] -CH, n  = 10 (3 eCH, 5 cCH) In-bout nBR after high-flow-oxygen -CH: • No effects of oxygen, immediately and over time, on the nBR • No differences between the symptomatic and asymptomatic side 2018, Perrotta et al. [ 213 ] -eCH, n  = 28 -ePH in-bout, n  = 18 -HC, n  = 21 -In- bout, n  = 16 -Out-of-bout, n  = 12 nBR -eCH (both in-and out-of-the-bout) and ePH: • Lower mean percentage decrease of the R2 area across all blocks at 0.2 to 1 Hz stimulation frequency in comparison to HC • Frequency-dependent habituation deficit of trigeminal nociceptive responses at higher SFs 2025, Naber et al. [ 215 ] -CH, n  = 33 (eCH, n  = 17; cCH n  = 16) In-bout nBR 30 min before and 30 min, 1 week, and 4 weeks after GON block -CH: • No increased R2 latencies after GON infiltration were observed • No correlation between R2 latencies and clinical response were observed List of neurophysiological studies that investigated the pain processing in subjects with cluster headache. Abbreviations: auc = area under the curve; br = Blink reflex; ch = cluster headache; cCH = chronic cluster headache; dbsh = unilateral deep brain stimulation of the posterior hypothalamus; eCH = episodic cluster headache; ePH = episodic paroxysmal hemicrania; gon = greater occipital nerve; hc = healthy controls; i.v.= intravenously; moa = migraine without aura; n = number; nBR = nociceptive Blink reflex; LEP = laser-evoked potential; PREP = pain-related evoked potentials; tsep = trigeminal somatosensory evoked potentials -eCH, n  = 10 -HC, n  = 10 -CH: • No changes of R2 threshold, latency or area -After paired supraorbital stimuli: • More rapid R2 recovering in eCH patients on the symptomatic side -After index stimulations: • More rapid R2 recovery on both symptomatic and non-symptomatic sides in eCH compared to HC. Partial reversion of the R2 suppression induced by index finger stimuli in 2 eCH subjects after Naloxone i.v. administration -eCH, n  = 28 -HC, n  = 22 -In-bout, n  = 28 -Out-of-bout, n  = 22 -eCH in-bout: • Increased N2 TSEP latencies on the symptomatic side in comparison to the non-symptomatic side and with the same side out-of-bout • Increased N1, P1 and N2 latencies on the symptomatic side in comparison to HC -eCH out-of-bout: • Increased N1 latencies of both sides in comparison to HC • No differences of TSEP amplitudes and BR latencies in comparison to HC -CH, n  = 25 (16 eCH, 9 cCH) -HC, n  = 10 -In-bout, n  = 7 -Out-of-bout, n  = 10 -cCH: • Reduced P2 amplitude and delayed N1c in comparison to HC -eCH in-bout: • Reduced P2 latency on the headache side -eCH out-of-bout: • reduced N2P2 ratio on the headache side -CH: • in 19/26 examinations LEP deviated from normative data in HC without any specific pattern of altered parameters -cCH: • Decrease of the R2 response areas and increase of R2 latencies after the GON block only on the injection side -eCH, n  = 27 -MoA, n  = 22 -HC, n  = 20 -eCH: • Deficit of habituation in the R2 and R3 components compared with HC and MoA -eCH: • More pronounced lack of habituation in comparison to MoA -CH, n  = 26 (11 cCH with DBSh; 15 cCH without DBSh) -HC, n  = 29 -cCH with DBSh: • Increased cold pain thresholds at the first trigeminal branch on the stimulated side in comparison to HC, and higher cold detection thresholds in comparison to cCH without DBSh -cCH with DBSh: • No changes after short-term interruption of stimulation -CH, n  = 66 (46 eCH, 20 cCH) -HC, n  = 30 -In-bout, n  = 18 -Out-of-the-bout, n  = 28 -CH (independently from the subtypes and the phase): • No differences in habituation in comparison to HC • No side-to-side differences of habituation between headache side and non-headache side -CH, n  = 66 (46 eCH, 20 cCH) -HC, n  = 20 -In-bout, n  = 18 -Out-of-the-bout, n  = 28 -CH (independently from the subtype): • Decrease in nBR latency ratio (headache side/non-headache side) in comparison to HC -eCH: • Increase in AUC ratio in patients with episodic CH in-bout only -cCH: • Decrease N2 latency ratio of PREP -eCH, n  = 18 -HC, n  = 18 -eCH: • Decrease in reflex area on both sides in comparison with HC • Deficit of habituation only on the affected side • Positive correlation between the habituation slope and the number of days since the onset of the bout and the daily attack frequency -CH: • No effects of oxygen, immediately and over time, on the nBR • No differences between the symptomatic and asymptomatic side -eCH, n  = 28 -ePH in-bout, n  = 18 -HC, n  = 21 -In- bout, n  = 16 -Out-of-bout, n  = 12 -eCH (both in-and out-of-the-bout) and ePH: • Lower mean percentage decrease of the R2 area across all blocks at 0.2 to 1 Hz stimulation frequency in comparison to HC • Frequency-dependent habituation deficit of trigeminal nociceptive responses at higher SFs -CH: • No increased R2 latencies after GON infiltration were observed • No correlation between R2 latencies and clinical response were observed CAS are hallmark features of cluster headache, and objective electrophysiological studies have investigated both central and peripheral autonomic function. Using dynamic pupillometry, Micieli et al. [ 212 ] found altered pupillary dynamics during active periods, including abnormal constriction and dilation velocities, consistent with autonomic imbalance. In a subsequent study, the same authors [ 213 ] showed reduced amplitude and delayed latency of the trigemino-pupillary reflex, indicating compromised sympathetic responses. Another study that compared face and extremity sympathetic skin responses revealed face sympathetic hypoactivity on the symptomatic side, failing to show any differences in the extremity skin responses [ 214 ]. Further evidence in favour of autonomic system dysfunction comes from Tassorelli et al. [ 215 ], who used the cold pressor test to elicit simultaneous pupillary and cardiovascular responses. They showed that CH patients exhibited blunted autonomic reactivity, again pointing to dysfunction in central autonomic control mechanisms. Evers et al. [ 216 ] evaluated peripheral autonomic potentials and found reduced and delayed responses, reflecting altered function in peripheral autonomic reflex arcs. These instrumental findings reveal measurable dysregulation of the autonomic system in CH, affecting both central integration and peripheral execution of autonomic responses. These abnormalities correspond with the hallmark attack-related symptoms such as lacrimation, rhinorrhoea, and ptosis. Finally, the strong circadian and sleep-related patterns of cluster headache have prompted electrophysiological studies into sleep physiology. According to Gorgoni et al. [ 217 ], who reviewed available polysomnographic data, CH patients—particularly during active phases—exhibit altered sleep architecture. This includes fragmentation, reduced sleep efficiency, and abnormal Rapid-Eye-Movement (REM) regulation. These findings suggest that instability in sleep-wake transitions, especially those involving REM sleep, may play a role in attack initiation. Although early studies lacked detailed sleep staging, findings such as those by Polich et al. [ 218 ] and Silvestri et al. [ 219 ] noted atypical cortical activity and EEG abnormalities during nocturnal attacks, hinting at state-dependent vulnerabilities in sleep. Complementary findings from Jürgens et al. [ 168 ], where posterior hypothalamic stimulation altered both pain perception and thermal thresholds, reinforce the view that the hypothalamus is a critical node linking sleep regulation, circadian rhythms, and pain modulation. Taken together, electrophysiological studies in CH consistently reveal central and lateralized dysfunctions in pain processing, widespread abnormalities in autonomic control, and disrupted sleep physiology, all of which have been captured using reliable neurophysiological techniques. Collectively, these findings not only reflect the complexity of CH but also underscore its distinct neurophysiological fingerprint—one that bridges pain, autonomic function, and chronobiology.