Premonitory phase in primary headaches: shared and divergent mechanisms in migraine and cluster headache, the Junior Editorial Board Members’ vision

Abstract

Background and aim

Migraine and cluster headache are two highly disabling primary headache disorders, each characterized by distinct clinical phases. Both conditions feature a premonitory phase that precedes the onset of pain or the emergence of cluster bouts, and the symptoms occurring during this stage may provide critical insight into their underlying pathophysiology. This narrative review aims to synthesize current evidence on the premonitory phase of migraine and cluster headache, including its clinical manifestations, neurobiological mechanisms, and potential implications for early or pre-emptive therapeutic strategies.

Main results

Migraine patients frequently experience premonitory symptoms before the onset of headache, and multimodal neuroimaging indicates that migraine initiation involves coordinated functional alterations within hypothalamic–brainstem circuits. In cluster headache, premonitory phenomena can be categorized into pre-cluster symptoms, which emerge hours to days before a bout, and pre-attack symptoms, which develop minutes to an hour before an individual attack. Evidence from neuroimaging, genetic, and provocation studies implicates hypothalamic dysfunction, circadian dysregulation, and trigeminal-autonomic network activation as key mechanisms underlying the pre-attack and pre-cluster phases. Across both migraine and cluster headache, early recognition and targeted intervention during the premonitory phase may help prevent or attenuate attacks before the full cascade of nociceptive activation is initiated.

Conclusions

The premonitory phase represents a critical window for understanding the initiation and progression of both migraine and cluster headache. Current evidence suggests that these disorders share partially overlapping neurobiological substrates, particularly involving alterations in hypothalamic function. By elucidating the mechanisms underlying these early phases, future research may help reshape the current clinical paradigm toward truly pre-emptive intervention.

Introduction

Headache disorders represent one of the leading causes of disability worldwide, with migraine and cluster headache (CH) being two of the most prevalent and disabling primary headache syndromes [1, 2]. Although they differ markedly in clinical presentation, temporal pattern and epidemiology, both disorders share a complex, multiphasic course that extends beyond the headache attack itself [3–5]. Increasing evidence indicates that the early period before the pain phase, known as premonitory phase, plays a critical role in the initiation and modulation of headache attacks.

In migraine, the premonitory phase is characterized by diverse non-painful symptoms such as mood and cognitive changes, yawning, neck discomfort, and food cravings, which occur before the onset of pain in patients with migraine without aura (MO), or before the aura phase in patients with migraine with aura (MA) [6]. Functional magnetic resonance imaging (MRI) studies have identified the involvement of specific brain regions, including the hypothalamus, frontal, occipital, and brainstem areas, which may account for the symptoms reported during the premonitory phase of migraine [7, 8]. Additionally, a recent study showed that early treatment during this phase can effectively alleviate the pain phase, underscoring the therapeutic relevance of early intervention in migraine management [9].

Although the International Classification of Headache Disorders, 3rd edition (ICHD-3) does not formally define a premonitory phase in cluster headache (CH), parallel observations have described comparable premonitory phenomena, albeit with evaluation complicated by the distinct temporal pattern of the disorder [10]. In CH, recurrent attacks cluster into periods known as “cluster bouts,” which differ fundamentally from the episodic nature of migraine [3]. It is therefore critical to distinguish between symptoms that precede an entire cluster bout and those that precede individual CH attacks (Fig. 1).Pre-cluster symptoms refer to local or systemic manifestations that arise hours to one week before the onset of a cluster bout [11, 12]. These may include mild pain in the same distribution as subsequent CH attacks, constitutional symptoms, mood alterations, and sleep disturbances [11, 13]. Notably, some pre-cluster symptoms have been proposed as potential predictors of treatment response [11]. In contrast, pre-attack symptoms occur immediately prior to each CH attack, typically lasting from several minutes to about one hour [14].

Fig. 1.

Premonitory symptoms in cluster headache (CH) and migraine. Migraine and CH show a similarly high prevalence of symptoms preceding headache attacks (~ 80–85%), with overlapping features such as mood changes, concentration difficulties, neck stiffness, and yawning, indicating shared involvement of hypothalamic, brainstem, limbic, and dopaminergic networks. In migraine, premonitory symptoms typically develop gradually over hours to days. In CH, pre-cluster symptoms often emerge 1–7 days before the onset of an active cluster period. Pre-attack symptoms in CH are distinguished by a higher prevalence of head and neck “shadow” pain and lateralized cranial autonomic features, reflecting early activation of trigeminal–autonomic and nociceptive pathways. These symptoms arise rapidly, usually within minutes to less than an hour before pain onset, and are associated with focal, ipsilateral hypothalamic activation and pronounced circadian rhythmicity. Created with BioRender.com

Understanding these early phases is of increasing scientific and clinical interest. They not only offer insights into the neurobiological mechanisms underlying attack initiation but also present a therapeutic window for early or even preventive intervention. This review summarizes the current evidence on the premonitory phase of migraine and CH, emphasizing shared and distinct clinical features, underlying neurobiology, and the emerging implications for early treatment strategies.

Migraine and the premonitory phase

The ICHD-3 defines the premonitory phase of migraine as a symptomatic period lasting up to 48 h that precedes the onset of headache pain MO or the aura phase in MA [15]. Interestingly, a systematic review and meta-analysis of observational studies found that the onset of premonitory symptoms ranged from 6.3 to 10.6 h, with one study reporting that 45% of symptoms lasted less than one hour and only 13% persisted for more than 12 h [6, 16–18]. In general, premonitory symptoms encompass neuropsychiatric (including cognitive and behavioural), sensory or autonomic, and more general or subjective symptoms [19]. Notably, these symptoms have been documented not only in adults but also in adolescents and in children as young as 18 months, as reported by parents [20].

The reported prevalence of premonitory symptoms in migraine is highly heterogeneous, ranging from 9% to 88%, depending on the study period and method of data collection (e.g., retrospective recall vs. prospective reporting). However, recent prospective studies have generally reported rates between 77% and 88% [6, 21]. A systematic review and meta-analysis of observational studies by Eigenbrodt and colleagues indicated that the pooled estimated prevalence of at least one premonitory symptom was 29% (95% CI 8–63) in population-based studies and 66% (95% CI 45–82) in clinic-based studies [6]. One of the largest studies included in this meta-analysis, a multicentre investigation by Wang et al., involved patients with migraine from tertiary headache and neurology clinics; 21.5% of participants reported at least one premonitory symptom [22]. The most frequently reported symptoms were neck stiffness, dizziness, yawning, and drowsiness. Moreover, premonitory symptoms were more common in patients with MA than in MO. Nevertheless, frequency estimates of premonitory symptoms vary widely across studies, likely reflecting differences in methodology, data collection approaches, and definitions, as well as potential bias and the limited number of large prospective diary-based investigations [6, 23, 24]. It is also important to note that self-reported premonitory features do not necessarily imply causal relationships and may be influenced by recall bias or reverse causality.

Distinguishing premonitory symptoms from migraine triggers remains challenging, as both are associated with an increased likelihood of a migraine attack. Premonitory symptoms, however, are considered part of the earliest phase of the migraine attack, whereas triggers are defined as endogenous or exogenous factors that precipitate an attack [25]. Consistent with this distinction, a recent cross-sectional study evaluating reported trigger factors alongside premonitory symptoms demonstrated a substantial overlap between specific triggers and corresponding premonitory features. Significant associations were observed between bright light and premonitory photophobia, loud noise and premonitory phonophobia, and sleep disturbances and premonitory tiredness, suggesting that some perceived triggers may represent early manifestations of an impending migraine attack [26]. Furthermore, in a three-month study using electronic diaries, a selected group of patients who reported non-headache features, were found to exhibit premonitory symptoms that served as relatively reliable predictors of migraine headache attacks [27]. Similarly, an analysis of data from the screening period of the PRODROME trial found that migraine patients who could identify premonitory symptoms most commonly reported sensitivity to light, fatigue, neck pain, sensitivity to sound, difficulty concentrating, and dizziness, with 81.5% of these premonitory events followed by headache within 1–6 h [28]. Overall, despite considerable variability in reported prevalence, the consistent occurrence of premonitory symptoms across studies reinforces the view that migraine pathophysiology is initiated prior to pain onset, providing a crucial clinical foundation for investigating the neural mechanisms underlying this early phase of the disorder.

Pathophysiology of the premonitory phase in migraine

Various functional neuroimaging techniques have been employed to elucidate the neural correlates of premonitory symptoms in migraine. One such modality is positron emission tomography (PET), which allows assessment of cerebral blood flow. In a study involving eight migraine patients, nitroglycerin (NTG)-triggered migraine-like attacks were associated with significant increases in blood flow within the hypothalamus, thalamus, cingulate cortex, and dorsolateral pons during both the early and late premonitory phases [29]. Premonitory symptoms such as nausea and photophobia were linked to increased perfusion in the occipital lobes in patients with photic hypersensitivity, while activation of a region in the medulla involving the nucleus tractus solitarius was observed in patients reporting nausea [30, 31]. Furthermore, in a double-blind, randomized, placebo-controlled, perfusion MRI study using arterial spin labelling (ASL), cerebral blood flow was analysed in 21 migraine patients who completed both NTG-triggered and placebo imaging sessions. After correction for NTG-related vascular effects, increased perfusion was observed in the hypothalamus, thalamus, basal ganglia, and limbic cortex during the premonitory phase [32].

Most neuroimaging studies investigating the premonitory phase of migraine have utilized functional MRI (fMRI). In a placebo-controlled, multi-visit experimental study employing resting-state fMRI and seed-based connectivity analyses, each participant served as their own control and underwent imaging at identical time points, baseline, premonitory, headache, and recovery phases, following NTG or placebo infusion in a randomized, double-blind, crossover design. During the NTG-triggered premonitory phase, significant positive functional coupling was observed between the bilateral thalami and the right praecuneus and cuneus regions, along with a shift in connectivity direction, from positive to negative, between the pons and the limbic lobe in 25 migraine patients. These findings suggest an early functional reorganization of sensory and limbic networks during the premonitory phase, which may underlie the manifestation of premonitory symptoms [33]. Nevertheless, while NTG provocation is a widely used and informative experimental paradigm, the pharmacologically induced nature of these attacks may not fully capture the complexity of spontaneous migraine, particularly, the premonitory phase.

Trigemino-nociceptive stimulation using intranasal ammonia during the interictal period has been shown to elicit increased neuronal activity in the spinal trigeminal nucleus preceding migraine onset, as reflected by elevated blood oxygen level–dependent (BOLD) signal intensity during the preictal phase [34]. Longitudinal fMRI data further support dynamic brainstem and hypothalamic involvement across the migraine cycle. In one patient who underwent daily fMRI scanning for 30 consecutive days, intranasal ammonia provoked heightened hypothalamic activation during the preictal period, accompanied on the day before headache onset by altered functional connectivity between the hypothalamus and migraine-relevant regions, including the spinal trigeminal nucleus and dorsal rostral pons [7]. Extending these findings, a cohort of seven patients with episodic migraine demonstrated hypothalamic activation as early as two days prior to headache onset during daily trigemino-nociceptive, olfactory, and visual stimulation paradigms combined with serial fMRI scanning [35]. Consistent with these observations, a study that evaluated patients with migraine during the preictal (n = 8), postictal (n = 11), and interictal (n = 26) periods, as well as 78 healthy controls, observed an increase in infraslow oscillatory activity in the hypothalamus and brainstem regions, including the spinal trigeminal nucleus and dorsal pons, immediately prior to a migraine attack [36]. Additionally, enhanced functional connectivity strength and regional homogeneity were observed among several hypothalamic and brainstem regions, suggesting that oscillatory activity within these structures may contribute to migraine initiation, potentially through astrocytic mechanisms [36].

In a small longitudinal study involving three migraine patients and five healthy controls, resting-state fMRI was performed every weekday for four weeks to examine oscillatory brainstem function across the migraine cycle [34]. Variability in the dorsal pons and spinal trigeminal nucleus increased markedly within 24 h prior to migraine onset, reflecting a dynamic shift in brainstem function during the pre-attack period [37]. Similarly, in a longitudinal intra-individual fMRI study of 12 migraine patients across 82 sessions, including spontaneous headache attacks and follow-up recordings before the next attack, cyclical changes were observed in sensory, limbic, and salience networks. These connectivity changes increased over the interictal period, peaked immediately before headache onset, and returned to baseline during the headache phase [38].

Furthermore, an fMRI study examining orofacial nociceptive stimulation in 31 patients with migraine and 31 healthy controls demonstrated reduced pain sensitivity and increased activation of the spinal trigeminal nucleus within the 24 h preceding headache onset, with concomitant reductions in functional connectivity with the rostral ventral medulla [39], suggesting altered brainstem pain-modulating circuitry. During the interictal period, patients with migraine also demonstrated significantly increased mean diffusivity in the spinal trigeminal nucleus, dorsomedial and dorsolateral pons, periaqueductal grey matter (PAG), and cuneiform nucleus, which normalized to control levels within 24 h before a migraine attack, only to rise again over the subsequent three days, as shown by diffusion tensor imaging [40]. In addition, elevated fractional anisotropy in the medial lemniscus/ventral trigeminal thalamic tract was observed in patients with migraine compared to controls throughout the migraine cycle. These findings suggest dynamic microstructural changes in brainstem anatomy across the migraine cycle, with specific alterations occurring within the 24 h preceding headache onset [40].

The relationship between hypothalamic and brainstem activation during the premonitory phase has also been explored using pseudocontinuous arterial spin labelling (pCASL) and functional connectivity analyses [41]. Reduced blood flow in the lateral hypothalamus was observed within 24 h before headache onset, together with decreased functional connectivity between the hypothalamus and the PAG, dorsal pons, rostral ventromedial medulla, and cingulate cortex. These findings suggest that the hypothalamus contributes to migraine initiation and modulates pain sensitivity by altering its connectivity with brain regions involved in pain processing during the premonitory phase [41].

Supporting these clinical findings, several animal studies have provided mechanistic insights into how hypothalamic and midbrain circuits may contribute to early migraine symptoms. Stress, a recognized migraine trigger, has been linked to dynorphin–kappa opioid receptor (KOR) signalling. In one rodent model, administration of a KOR agonist led to increased urination and water consumption without inducing yawning or allodynia, indicating that hypothalamic KOR activation can promote thirst and polyuria, symptoms frequently reported during the premonitory phase, and may further predispose to headache onset [42]. In addition, studies have shown that the ventral tegmental area (VTA), a midbrain region involved in reward and homeostatic regulation, modulates trigeminovascular sensory processing, a key pathway in migraine pathophysiology. Pharmacological modulation of serotonin, Pituitary Adenylate Cyclase-Activating Polypeptide (PACAP), and dopamine signalling within the VTA altered neuronal responses in the trigeminal cervical complex and lowered circulating glucose levels, suggesting that homeostatic and reward-related networks interact with sensory pathways during the premonitory phase, possibly underlying symptoms such as food craving [43].

Collectively, converging evidence from multimodal imaging studies suggests that migraine initiation involves coordinated functional alterations within hypothalamic–brainstem circuits. These systems integrate stress, homeostatic, and sensory signals, supporting their pivotal role in generating early premonitory symptoms and setting the stage for the subsequent headache phase.

Cluster headache and the premonitory phase

Pre-cluster symptoms are those occurring before the onset of upcoming cluster bouts [11, 12], and are distinct from the premonitory phase that immediately precedes individual attacks [12, 44]. Current epidemiological studies have reported the presence of pre-cluster symptoms in CH patients, although prevalence varies across populations. Both Danish and Taiwanese cohorts found that up to 85–90% of CH patients experience such symptoms preceding cluster bouts [11, 12], whereas a Chinese study reported a prevalence of 20.8% [45]. Despite this variation, studies consistently note that pre-cluster symptoms typically emerge 1–7 days before the onset of an active cluster period.

The most commonly reported pre-cluster symptoms across both Western and East Asian studies include head and neck pain, such as dull facial or head discomfort and neck pain [11, 12]. However, constitutional symptoms such as concentration difficulties and phonophobia appear to be more frequently observed among East Asian CH patients [11, 12]. CH bouts exhibit a seasonal or circannual rhythmicity (usually peaking in the spring or autumn; reviewed in [46]), possibly associated with seasonal variations in temperature [47] and daylight duration [48]. Therefore, dysfunction of the master circadian clock located within the hypothalamic suprachiasmatic nucleus has been hypothesized to contribute to the pathophysiology of the pre-cluster phase [11, 49].

Pre-attack symptoms are now widely recognized as a distinct, non-painful phase that precedes the onset of pain in cluster headache attacks. These typically emerge within minutes, and occasionally up to one hour, before the onset of pain. The presence of pre-attack symptoms in CH was first described by Blau and Engel in 1998, who reported that 61% of patients experienced warning signs prior to an attack [50]. Since then, multiple studies have confirmed and expanded upon this finding, underscoring the consistency and clinical relevance of this early phase of CH.

Across cohorts, a consistent pattern of pre-attack symptoms has been identified. The most frequently reported is a dull, aching sensation in the region typically affected by the pain, often referred to by patients as a “shadow” pain. This vague discomfort mirrors the laterality and location of the full-blown attack and is frequently accompanied by ipsilateral neck stiffness. In addition to localized pain, many patients report unilateral cranial autonomic symptoms such as lacrimation, conjunctival injection, and nasal congestion or rhinorrhoea. These may be accompanied by cognitive and homeostatic disturbances, including fatigue, mood changes, concentration difficulties, excessive yawning, and sleep disruption [14, 45, 51–53]. Temporal analyses indicate that “shadow” pain and autonomic symptoms typically appear approximately 10 min before pain onset, whereas cognitive and homeostatic symptoms tend to occur earlier—often 20 to 60 min before the attack [52, 53]. In a Danish observational study involving 80 patients with CH, 83.3% reported experiencing pre-attack symptoms. Interestingly, no significant associations were found between the presence of premonitory symptoms and either sex or CH subtype (episodic vs. chronic), apart from a significant association between the occurrence of “shadow” pain and both male sex and the episodic form of the disorder. These results demonstrate the consistency of the premonitory profile across diverse patient populations [52]. Notably, a recent study by González-Martínez et al. involving 164 patients with both episodic and chronic CH found that up to 85% of participants reported identifiable premonitory symptoms [51]. This study also observed a higher prevalence of pre-attack symptoms among patients using high-flow oxygen for acute treatment, suggesting that individuals who recognize early warning signs may initiate therapy more promptly, thereby improving treatment efficacy. Other factors associated with an increased likelihood of reporting pre-attack symptoms include a higher number of reported attack triggers and a personal history of smoking [45].

Similar to premonitory symptoms in migraine, CH patients often report exposure to specific triggers within minutes to hours before an attack. Among these, alcohol consumption is the most frequently reported [54]. More than 50% of CH patients identify alcohol intake as a precipitant [55, 56], with red wine appearing more strongly associated with attack provocation than other alcoholic beverages [57]. While up to 90% of CH patients report a history of tobacco exposure, either first- or second-hand, the role of tobacco as a direct trigger of cluster attacks remains poorly defined [56, 58–60]. Additional epidemiologic and clinical data are needed to further elucidate its role. Other environmental factors, including sleep disturbance, strong odours, and changes in weather, atmospheric pressure, or altitude, may also act as triggers for CH [61, 62].

Pathophysiology of the premonitory phases in CH

Although studies on the pathophysiology of CH are scarce, the symptomatology of premonitory features offers clues to their biological basis. Genetic studies have reported mutations in several core circadian clock genes (e.g., CLOCK, PER3, CRY) [63, 64]. Moreover, a recent study in individuals with CH found reduced seasonal variation, both during and outside of bouts, in peripheral CLOCK gene expression compared with healthy controls [65]. These findings support the hypothesis that dysfunctional hypothalamic circadian rhythms contribute to the premonitory phase of CH. In contrast, current biochemical studies have not demonstrated consistent abnormalities in hypothalamic-regulated hormones such as cortisol, thyroid-stimulating hormone, melatonin, or follicle-stimulating hormone [66].

Functional neuroimaging studies have consistently suggested the involvement of the hypothalamus in CH, showing that altered neural function persists even during out-of-bout phases [67]. In a longitudinal resting-state fMRI study, Yang et al. reported reduced hypothalamic functional connectivity with regions including the medial frontal gyrus and occipital cuneus during out-of-bout phases compared with healthy controls, with even greater reductions during active bouts [68]. Similarly, Chou et al. demonstrated persistent dysfunction across several large-scale networks, including the default mode, frontal, salience, and somatosensory systems, during out-of-bout states [69].

In addition to the hypothalamus, alterations have also been observed in other brain regions, particularly the anterior cingulate cortex, insula, and prefrontal cortex [67]. Such findings support the notion of persistent vulnerability, potentially predisposing individuals to recurrent CH episodes. A diffusion tensor imaging study also revealed chronic microstructural white matter changes in tracts like the corpus callosum and cingulum, reinforcing the view that even episodic CH involves long-term remodelling beyond symptomatic phases [70].

Structural neuroimaging has also provided insight into potential CH generators. Voxel-based morphometric studies comparing in-bout and out-of-bout phases [71, 72], as well as comparisons between CH patients and healthy controls [73], have generally found no significant alterations in hypothalamic volume. However, distinct regional changes have been reported in other brain regions, including the temporal lobe, hippocampus, insula, and cerebellum. Beyond the traditional pain networks, Giorgio et al. noted decreased frontal grey matter and increased cerebellar and occipital volume, suggesting widespread changes beyond traditional pain-processing networks [74].

Evidence from provocation studies has identified multiple neuropeptides and signalling molecules capable of triggering CH attacks. Among these, nitric oxide is perhaps the most extensively studied. As early as 1953, small clinical trials demonstrated that nitroglycerin, a nitric oxide donor, could reliably induce CH attacks in human participants [75]. Subsequent placebo-controlled trials confirmed this effect, with attack rates ranging from 60% to 81% following nitroglycerin infusion. In a more recent crossover trial involving 24 participants, nitroglycerin infusion triggered cluster attacks in 76% of the study population compared with only 4% during placebo infusion [76]. Of interest, the majority of subjects reported non-headache symptoms in the lead-up to the onset of the pain from the attack.

Calcitonin gene–related peptide (CGRP), a potent vasodilatory neuropeptide, has also been implicated in CH pathophysiology. In a randomized crossover trial of 32 CH patients, CGRP infusion induced cluster attacks in 89% of participants with active episodic CH and in 50% of those with chronic CH, but not in any participant in remission. However, no premonitory symptoms were reported in this study [77]. These results underscore the complex and multifactorial nature of CH pathophysiology and the importance of disease phase in modulating susceptibility to attacks.

Finally, the roles of PACAP and vasoactive intestinal peptide (VIP) have been examined in a crossover trial involving 14 participants with active episodic CH. Of these, 43% experienced a cluster-like attack following PACAP infusion, compared with 36% after VIP administration. In contrast, only 0–20% of participants in remission reported such attacks [78]. Further clinical studies are warranted to clarify the role of PACAP and VIP signalling in the acute mechanisms underlying CH. Nonetheless, while PACAP infusion can induce cluster-like attacks in susceptible individuals, mirroring its role in migraine, PACAP-provoked CH attacks have not been reported to be preceded by pre-attack symptoms, highlighting a mechanistic divergence in how this neuropeptide engages hypothalamic pathways across the two disorders [79, 80].

In summary, pre-cluster and pre-attack symptoms in cluster headache represent distinct, clinically relevant phases that provide valuable insight into disease pathophysiology and potential windows for early intervention. While the timing, prevalence, and specific manifestations of these premonitory features vary across populations, consistent patterns, such as “shadow” pain, cranial autonomic signs, and cognitive or homeostatic disturbances, underscore the reproducibility and predictive value of these early symptoms. Converging evidence from neuroimaging, genetic, and neuropeptide studies suggest hypothalamic dysfunction, and trigeminal–autonomic network activation as central mechanisms underlying these premonitory phases. Moreover, the differential response to provocation and treatment across disease phases highlights the importance of temporal context in CH pathophysiology. Continued research into the neurobiological basis, clinical characterization, and therapeutic targeting of pre-cluster and pre-attack symptoms is essential to improve early detection, optimize preventive strategies, and ultimately reduce the burden of cluster headache.

Comparative insights into the premonitory phase of migraine and cluster headache

The premonitory symptoms in migraine have been extensively studied. As previously mentioned, the reported prevalence is highly heterogeneous, ranging from 9% to 88%, depending on the methodology [6, 21]. In CH, the prevalence of symptoms preceding a CH bout have been observed in 20.8–90% of patients [11, 12, 45], while for pre-attack symptoms has been reported to be 82.3% [45], closely aligning with the prevalence of migraine premonitory symptoms. Recently, González-Martínez et al. directly compared the prevalence of symptoms that precedes the headache attacks between migraine and cluster headache, which reported a similar overall prevalence of pre-attack symptoms of CH and premonitory symptoms of migraine (85% vs. 84%; regression model adjusted p = 0.523) [51]. Both conditions sharing common features such as mood changes, concentration difficulties, neck stiffness, and yawning (see Fig. 1) [51]. These overlapping symptoms suggest the involvement of shared brain structures, including the hypothalamus, brainstem, limbic system, and dopaminergic pathways [81].

Nevertheless, there are important qualitative differences distinguish both conditions. The premonitory symptoms of migraine can be generally divided into three categories: fatigue and cognitive symptoms, symptoms related to homeostatic changes, and sensory sensitivities [82, 83]. Fatigue and cognitive symptoms include concentration difficulties, tiredness, memory impairment, mood changes, and irritability. Homeostatic alterations comprise yawning, sleep disturbance, changes in urinary frequency, thirst, and food craving. Sensory sensitivities encompass neck pain or stiffness, photophobia, phonophobia, osmophobia, and nausea [82, 83]. In the case of the symptomatology of premonitory symptoms in CH, the presence of ‘shadow’ pain and lateralized cranial autonomic symptoms is more common during the pre-attack and pre-cluster phases of CH, with a prevalence of approximately 70–80%, but is less frequent during the premonitory phase of migraine, where it occurs in about 40% of cases [11, 14, 28]. The higher prevalence of pain symptoms over the head and neck region, together with cranial autonomic symptoms, likely reflects early activation of the trigeminal–autonomic reflex and nociceptive pathways [11, 52]. Additionally, a key distinguishing aspect lies in the temporal profile, in CH, pre-attack symptoms typically emerge within minutes, often less than an hour before the onset of pain, whereas in premonitory symptoms of migraine, they tend to develop more gradually, frequently evolving over several hours to days [52, 53].

Current evidence indicates that both migraine and CH involve early activation of deep-brain structures, but with distinct patterns and clinical consequences. In migraine, the premonitory phase is consistently associated with early engagement of the hypothalamus, limbic system, and brainstem [35, 84]. Fatigue and cognitive symptoms may reflect functional reorganization of frontal–parietal regions [33], while photophobia and nausea have been linked to activation of extrastriate visual areas and the rostral dorsal medullary area with the PAG, respectively [30, 31]. This phase is tightly linked to neuropeptides including CGRP, PACAP, and orexins, all of which participate in trigeminal nociceptive signalling [85–89]. Experimentally, administration of both CGRP and PACAP can provoke migraine attacks and premonitory symptoms, though PACAP appears more potent, likely due to its ability to cross the blood–brain barrier and directly modulate hypothalamic and thalamic networks [87, 90]. In contrast, CH is characterized by a more focal and lateralized pattern of activation, with functional imaging demonstrating ipsilateral posterior hypothalamic activation during attacks [91, 92]. Given the role of the hypothalamus in circadian regulation, its dysfunction is thought to underlie the striking chronobiological rhythmicity seen in CH [11]. Pre-attack and pre-cluster symptoms in CH typically include cranial autonomic features and pain, mediated by trigeminal–autonomic reflex circuits and the trigeminal nociceptive system, respectively [93]. Similar to migraine, CGRP seems to play an important role, as CGRP infusion can trigger CH attacks, galcanezumab reduces attack frequency, and commonly used preventives such as corticosteroids and verapamil influence CGRP levels [77, 94–97]. Additional evidence suggests that oxygen therapy may modulate early hypothalamic or trigeminal–autonomic activation, given its association with premonitory symptoms [51]. Nevertheless, as noted earlier, although PACAP infusion can trigger cluster-like attacks in susceptible individuals, similar to its effect in migraine, these PACAP-induced CH attacks have not been observed to be preceded by pre-attack symptoms, underscoring a mechanistic difference in how this neuropeptide interacts with hypothalamic pathways in the two disorders [79, 80].

Collectively, these findings indicate that migraine and CH share partially convergent pathways, yet remain distinguished by disorder-specific symptom profiles, temporal dynamics, and patterns of subcortical network engagement. The overlap in pre-attack features and signalling pathways suggests common upstream mechanisms within hypothalamic, limbic, and brainstem circuits, whereas the prominent cranial autonomic involvement, lateralized hypothalamic activation, and rapid temporal evolution of CH premonitory symptoms point to divergent modes of trigeminal–autonomic system recruitment [93]. A more precise delineation of these shared and distinct processes will be essential for improving early recognition, refining mechanistic models, and ultimately guiding the development of targeted interventions aimed at interrupting attacks before the onset of disabling headache attacks.

Therapeutic implications

For both migraine and CH, early identification and targeted treatment during premonitory phase may help prevent or attenuate attacks before the full cascade of nociceptive activation is initiated.

In migraine, clinical studies have been conducted to better characterize the premonitory phase, not only to raise awareness among patients, but also to develop interventions and provide pre-emptive therapies that decrease the likelihood of migraine onset or progression [98]. As reviewed by Becker [99] and Karsan & Goadsby [100], a few studies have shown the potential of drugs with anti-dopaminergic effects, such as flunarizine, prochlorperazine, haloperidol, droperidol, and domperidone, given during the premonitory phase. Furthermore, Luciani and colleagues reported a small open-label study indicating that naratriptan 2.5 mg may reduce headache occurrence when taken by patients who anticipated an impending attack [101]. Results from the PRODROME trial, a phase 3, randomized, double-blind, placebo-controlled trial in 518 participants, demonstrated that ubrogepant was effective when taken during the premonitory phase, when headache was still absent. A total of 190 (46%) of 418 qualifying events treated with ubrogepant did not progress to moderate or severe headache within 24 h, compared to 29% of events treated with placebo. Similar benefits were observed at 48 h [9]. In addition, ubrogepant significantly resolved common premonitory symptoms such as photophobia, phonophobia, neck pain, fatigue, dizziness, and cognitive difficulties, with effects appearing as early as 1 to 4 h post-dose and lasting up to 48 h [102].

Several studies have examined preventive treatments aimed at reducing the frequency of premonitory symptoms. Both topiramate and flunarizine significantly improved non-headache features, including premonitory symptoms, across the premonitory, headache, and resolution phases after eight weeks of treatment in patients with episodic migraine, with no significant difference between the two drugs [103]. Furthermore, an open-label observational cohort study using daily headache diaries showed that three months of galcanezumab treatment led to a reduction of the incidence of premonitory symptoms that were followed by headache by 48% in responders compared with 25% in non-responders, and by 50% in super-responders compared with 12% in super non-responders [104].

Currently, there is a paucity of studies investigating the role of the premonitory phase in CH treatment, and no studies have directly evaluated interventions targeting pre-cluster or pre-attack symptoms themselves. This limitation may largely be attributed to the short duration of the premonitory phase in CH, which makes it challenging to deliver treatment in time and to assess therapeutic response. Cohort studies show that many individuals with episodic CH can identify symptoms that occur days before a new bout, supporting the idea of a pre-bout window for early treatment [11, 12, 44]. For pre-cluster symptoms, a recent Taiwanese study involving 168 patients reported that patients presenting with sleep disturbances or any cranial autonomic symptom showed a better response to verapamil as a preventive treatment for CH [11]. These findings provide important insights into the potential for individualized treatment strategies in CH. Long acting triptans have also been used in practice as short-term preventive options in patients with predictable attack patterns, such as regular nocturnal attacks [105, 106]. However, this approach is based on timing rather than premonitory symptoms and is supported only by clinical experience [105, 106].

Despite increasing interest in pre-emptive treatment strategies targeting the premonitory phase, several limitations currently restrict their clinical applicability. To begin with, the overlap between premonitory symptoms and perceived migraine triggers introduces uncertainty regarding when pre-emptive therapy should be initiated [26]. Moreover, frequent early use of acute medications may increase the risk of medication overuse headache, particularly in patients with high attack frequency [107]. Furthermore, from a mechanistic perspective, currently available therapies may not adequately target the central hypothalamic and brainstem pathways thought to be involved in the premonitory phase, potentially limiting their effectiveness in preventing headache onset.

Overall, the existing evidence highlights the clinical importance of recognizing and therapeutically targeting the earliest phases of both migraine and CH, despite substantial differences in the duration and practical treatability of their premonitory periods. Isn migraine, growing data suggest that intervention during the premonitory phase can meaningfully alter attack progression, reduce symptom burden, and improve patient outcomes, underscoring the value of patient education and the development of mechanism-based, rapid-acting therapies. In contrast, although the premonitory phase in CH remains far less studied and is often too brief to allow timely therapeutic intervention, emerging findings, particularly regarding identifiable pre-bout features and differential responses to preventive therapies, suggest that early-phase treatment may nonetheless hold promise for select patients. Nonetheless, the clinical applicability of pre-emptive therapy is limited by factors such as the difficulty of distinguishing premonitory symptoms from perceived triggers, the risk of medication overuse headache with frequent early intervention, and the possibility that currently available therapies may not adequately target central hypothalamic and brainstem pathways critical to attack initiation. Continued efforts to refine the characterization of premonitory symptoms, develop reliable biomarkers of attack initiation, and evaluate targeted early interventions will be crucial for advancing personalized, phase-specific treatment strategies in both disorders.

Conclusion

The premonitory phase represents a critical window for understanding the initiation and evolution of both migraine and cluster headache. Across both disorders, early non-painful symptoms provide insight into the neural circuits that drive attacks and offer potential opportunities for pre-emptive intervention. Despite differences in duration, prevalence, and clinical manifestations, migraine and CH share partially overlapping neurobiological substrates; however, disorder-specific distinctions are evident. From a clinical perspective, the recognition and characterization of premonitory symptoms hold significant therapeutic implications. Moving forward, advancing our understanding of premonitory phases in both disorders will require large-scale, prospective, and diary-based studies, combined with multimodal neuroimaging, and mechanistic investigation of neuropeptide signalling. Such efforts will be critical for refining models of attack initiation, enabling the development of personalized, phase-specific treatments, and ultimately reducing the burden of these highly disabling headache disorders. By bridging the gaps in knowledge regarding shared and divergent premonitory mechanisms, future research promises to transform early detection and intervention, shifting the paradigm from reactive management toward true pre-emptive therapy.

Abbreviations

ASL

Arterial Spin Labelling

BOLD

Blood Oxygen Level-Dependent

CGRP

Calcitonin Gene-Related Peptide

CH

Cluster Headache

fMRI

Functional Magnetic Resonance Imaging

ICHD-3

International Classification of Headache Disorders, 3rd Edition

KOR

Kappa Opioid Receptor

MA

Migraine With Aura

MO

Migraine Without Aura

MRI

Magnetic Resonance Imaging

NTG

Nitroglycerin

PACAP

Pituitary Adenylate Cyclase-Activating Polypeptide

PAG

Periaqueductal Grey Matter

pCASL

Pseudocontinuous Arterial Spin Labelling

PET

Positron Emission Tomography

VIP

Vasoactive Intestinal Peptide

VTA

Ventral Tegmental Area

Author contributions

Conceptualization: J-WW, ER-B. Literature review and data synthesis: all authors. Writing – original draft: all authors. Writing – review & editing: all authors. All authors read and approved the final manuscript.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval

Not applicable.

Consent for publication

Not applicable.

Competing interests

All authors are Editorial Board Members of The Journal of Headache and Pain. Additionally, J-WW, reports speaker honoraria from Biogen-Idec, Pfizer, AbbVie, Eli Lilly, Organon, and HAVA Bio-Pharma. LA-H, is Member of the Methodology and Statistics Editorial Board of Neurology. C-CC is consultant for Pfizer, AbbVie, Amneal, Satsuma, and eNeura; and reports research support from the American Heart Association, Pfizer, and Lundbeck. RM, reports personal fees from AbbVie, Biomedia, Lundbeck, Organon, Pfizer, and Teva, and grants from the Italian Ministry of Health. WW-G, is Editorial Board Member of SN Comprehensive Clinical Medicine. ER-B is Associate Editor of Frontiers in Molecular Neuroscience.