How does the brain change in chronic migraine? Developing disease biomarkers

Abstract

Background

Validated chronic migraine biomarkers could improve diagnostic, prognostic, and predictive abilities for clinicians and researchers, as well as increase knowledge on migraine pathophysiology.

Objective

The objective of this narrative review is to summarise and interpret the published literature regarding the current state of development of chronic migraine biomarkers.

Findings

Data from functional and structural imaging, neurophysiological, and biochemical studies have been utilised towards the development of chronic migraine biomarkers. These biomarkers could contribute to chronic migraine classification/diagnosis, prognosticating patient outcomes, predicting response to treatment, and measuring treatment responses early after initiation. Results show promise for using measures of brain structure and function, evoked potentials, and sensory neuropeptide concentrations for the development of chronic migraine biomarkers, yet further optimisation and validation are still required.

Conclusions

Imaging, neurophysiological, and biochemical changes that occur with the progression from episodic to chronic migraine could be utilised for developing chronic migraine biomarkers that might assist with diagnosis, prognosticating individual patient outcomes, and predicting responses to migraine therapies. Ultimately, validated biomarkers could move us closer to being able to practice precision medicine in the field and thus improve patient care.

Keywords

Classification CGRP chronic migraine migraine neuroimaging neurophysiology predict outcomes

Introduction

Chronic migraine (CM) affects 1% to 2% of the global population (1) and 2.5% of persons with episodic migraine (EM) progress to CM each year (2). Compared to EM, CM is associated with more headache-related disability, higher rates of comorbidities, and lower health-related quality of life (3,4).

According to the US Food and Drug Administration, biomarkers are characteristics that can be objectively measured and evaluated as an indicator of normal biological, pathological, or pharmacological processes (5). Biomarkers for migraine should be helpful in diagnosing the disease and its subtypes, prognosticating its progression, and predicting the outcome of a therapeutic intervention (6).

Despite the prevalence and the burden of the disease, there are no validated biomarkers for CM. The diagnosis of CM is purely clinical and based on the criteria established by the International Classification of Headache Disorders (ICHD-3) (7). Currently, there can be partial subjectivity when making migraine diagnoses, and substantial difficulty when trying to prognosticate patient outcomes and predicting responses to migraine therapies. The diagnostic process is more complicated when migraine becomes chronic, especially because a) the characteristics of pain change and are less specific, b) the diagnostic criteria for CM are debated, and in continuous evolution, and c) the differential diagnosis widens (8,9).

Improvement in the diagnostic process would be facilitated by the identification and development of diagnostic biomarkers in the migraine field. Prognostic and predictive biomarkers would also be of great value.

In this review, we summarise and interpret the main studies in the literature (based on PubMed searches for neurophysiology, imaging, and biochemical study methods combined with chronic migraine, treatment, medication overuse, or classification), that contribute to the identification and development of CM biomarkers using imaging, neurophysiology, and neuromolecular techniques.

Brain imaging: Structural and functional chronic migraine biomarkers

Structural and functional brain imaging data can be used to develop models for prognosticating migraine patterns, predicting treatment response, and classifying migraine. For these imaging biomarkers to be most useful, they need to have high accuracy and to be applicable at the level of the individual patient rather than at the level of the patient group. In addition, the imaging and data analysis techniques have to be highly replicable, so that the imaging biomarker performs across different healthcare settings.

Efforts towards the development of imaging biomarkers have utilised the large datasets that result from brain imaging studies, such as MRI, within machine-learning multivariate pattern and deep learning analyses that identify the optimal set of imaging measurements for accurate prognostication, prediction, or classification.

CM is associated with brain structure and function that differ from EM and non-migraine healthy controls, and greater headache frequency is positively correlated with the extent of structural and functional aberrations (10 –12). For example, a functional MRI (fMRI) study demonstrated that interictal migraineurs have greater pain-induced brain activations of several pain processing regions compared to healthy controls (12). Similarly, another fMRI study demonstrated that, compared to EM, CM is associated with stronger activation of the hypothalamus in response to painful trigeminal stimulation (13). Resting state functional connectivity analysis of CM revealed altered connectivity of the default mode, salience, and central executive networks compared to healthy controls and a positive association between the frequency of moderate and severe headache days with the extent of abnormal connectivity in the salience and central executive networks (14,15). A resting-state functional connectivity study comparing CM to EM demonstrated that those with CM have stronger functional connectivity within the pain matrix (16).

Similar to the functional imaging studies, structural imaging studies demonstrate that higher headache frequency is associated with greater abnormalities in brain structure. Compared to healthy controls, migraineurs have less gray matter volume and thinner cortex of several pain and visual processing regions, and there are negative correlations between the structure of these regions with headache frequency (10,11).Furthermore, amongst those with CM, cortical thickness might correlate with specific clinical characteristics such as duration of time with CM, sleep quality, somatic symptoms and pain self-efficacy (17). CM is also associated with altered local and global structural network topology (18).

Prognostication of patient outcomes and predicting treatment response

A few publications suggest that the development of imaging-based predictive models for headache outcomes is possible. In one such study, right hippocampal volumes measured via MRI were positively associated with good patient outcomes 2 years later (defined as having at least a 50% reduction in headache frequency) (19). After adjusting for several variables that could impact patient outcomes (age, sex, anxiety/depression scores, baseline headache frequency), right hippocampal volume was associated with a 4.7 odds ratio for good outcomes. A longitudinal MRI study demonstrated that gray matter volume changes differed amongst those with EM who had worsening migraine over time versus those that had improved or showed stable migraine patterns (20). In an MRI study of individuals who had CM with medication overuse (MO), baseline orbitofrontal cortex volume predicted the response to various treatments (21). Further work is needed to determine if predictive models are accurate for the individual patient, are valid across similar patient populations, and if they can predict outcomes for individual modes of therapy.

Measuring early treatment response

Although not a study of CM, an investigation of pain-induced brain activations in EM demonstrated changes in activation patterns of the anterior cingulate cortex after 60 days of treatment with an external trigeminal neurostimulator (22). Migraineurs with more frequent migraine attacks had greater or more “hyper-reactive” pre- and post-treatment activations of the anterior cingulate. A MRI functional connectivity study of CM with MO demonstrated improvement in interregional intra-network connectivity of salience and executive networks following a series of treatments with sphenopalatine ganglion blocks (23). Brain imaging biomarkers for treatment response that could be used early after the initiation of treatment, prior to being able to determine response on clinical grounds alone, could help to increase the efficiency of screening new migraine therapies during their development. However, such a brain imaging biomarker is useful only if it measures or predicts treatment response earlier or more accurately than clinical indicators, such as headache frequency or severity.

Migraine classification using imaging data

Imaging data can be used for classification in at least two important ways. One method is useful for determining the specificity of imaging findings for individual headache diagnoses and for objectively testing the validity of subjective diagnostic criteria. With this method, imaging data are obtained from patients with known headache diagnoses, multivariable models of imaging data are developed, and the accuracy of those models for assigning an individual brain MRI (i.e. individual patient) to the correct diagnosis is determined. This approach has been taken in several published studies (24 –26). The classification accuracies for migraine, EM, and CM have ranged from about 65% to 95% depending on the study (26 –31). The most accurate models are those that classify individuals with more severe migraine burden (i.e. those with higher headache frequency/CM and more years with migraine). The higher classification accuracy for CM compared to EM further supports the observation that CM is associated with greater aberrations in brain structure and function compared to EM, thereby making it easier to classify. A useful imaging biomarker for CM classification would not only differentiate CM from EM and healthy controls, but also from other headache types. Although much more work is needed in this regard, studies have identified differences in brain structure or function between migraine and tension-type headache, cluster headache, and post-traumatic headache (32 –37).

Another method of using imaging data for classification purposes is to use “automated” data-driven clustering techniques to identify sub-groups of patients based upon imaging data alone. Once sub-groups are identified, similarities and differences in patient characteristics are examined. This approach could contribute to identification of new headache types and/or migraine sub-types. Automated MRI classification of 66 migraineurs and 54 healthy controls divided the group into two sub-groups based upon measures of brain structure (31). The migraineurs in one sub-group had more severe symptoms of allodynia during migraine attacks, more years with migraine, and greater migraine-related disability, suggesting that these clinical characteristics are associated with differences in brain structure.

Neurophysiology: Functional chronic migraine biomarkers

Several research groups have investigated possible neurophysiological biomarkers of CM. It is worth mentioning that most of the CM studies included patients with medication overuse headache (MOH). Among various possible candidates, electrophysiological markers of central sensitisation and defective central pain control systems are the most promising.

Chronic migraine

In CM, a larger amplitude in cortical responses related to pain has been detected (38,39). The frequent recurrence of pain promotes neural plasticity, and remapping of the pain process over the anterior cingulate cortex has been described in CM (40).

While investigating simultaneous non-painful somatosensory evoked potentials (SSEP) habituation and thalamo-cortical loop activation in CM, researchers have observed a remarkable neurophysiological pattern similar to that of ictal EM. In fact, both groups of patients were characterised by higher initial amplitudes and by diminishing response (habituation) over sequential block averages, resulting in a “transient” central sensitisation. In ictal recordings of CM and EM patients, the decrease in interictal episodic thalamo-cortical activity (reflected by reduced early high-frequency oscillations [HFOs] embedded in the common broad-band SSEP) normalised, whereas the amplitude of the primary cortical component (late HFOs) consistently increased compared with those of healthy controls and interictal EM patients (41). Central sensitisation in CM is not due to a deficit of lateral inhibition in the parietal cortex (42) but a contribution of subcortical inhibitory circuits cannot be excluded (43). These neurophysiological abnormalities of CM are proportional to the duration of the headache chronification phase (44).

Contradictory results were obtained studying the brainstem using the blink reflex response (39,45). Higher initial mean block amplitude followed by a normal response decrement (i.e. habituation) was also observed in recordings of magnetic visual evoked potentials (P100m) in CM patients compared with EM patients recorded interictally (46). This is consistent with the SSEP studies mentioned above (39,43). Using a test of cortical inhibition known as transcranial magnetic suppression of perceptual accuracy, researchers observed that CM patients (with or without MO) exhibited the lowest suppression index compared with controls, with EM falling in between (47,48). Cosentino et al. (49) reported an inhibitory response in CM patients during trains of TMS at 5 Hz, instead of a progressive physiological facilitation. Prolonged inhibition was confirmed recording the transcranial magnetic stimulation (TMS)-induced cortical silent period (CSP) in a group of patients with CM, but most of them were taking prophylactic medication at the time of testing (50).

Migraine evolved in medication overuse headache

In patients with MOH, the initially higher SSEP amplitudes were further increased during stimulus repetition, resulting in a “persistent” sensitisation (51). Nonetheless, patients experiencing cutaneous allodynia exhibited greater SSEP sensitisation compared with those without allodynia (44). As for CM, the neurophysiological abnormalities of MOH are proportional to the duration of the headache chronification phase (51).

Nonetheless, despite the overall lack of habituation in MOH patients, SSEP amplitudes differed according to the drug of overuse, because amplitudes were smaller in triptan overusers than in patients overusing non-steroidal anti-inflammatory drugs (NSAIDs) or both medications combined (51). These abnormalities appear to be influenced by genetic factors (52). Moreover, MOH patients exhibited deficient habituation during contingent negative variation cognitive potentials (53) and laser evoked potentials (LEP) recordings (54).

Exploring inhibitory circuits, Currà et al. (55) measured the CSP in a group of MOH patients. Despite the overall similarity in CSP duration between MOH patients and healthy controls, subgroup analysis revealed that CSP duration was significantly shorter in triptan overusers than in the NSAID or triptan-plus-NSAID overuser groups. In NSAID and triptan-plus-NSAID MOH subgroups, CSP duration correlated positively with monthly tablet intake, while triptan overusers exhibited a negative correlation (55). In a study using trains of motor cortex TMS at 5 Hz, which enhanced motor evoked potential amplitudes in healthy controls and pure CM patients, depressed MEP amplitudes in MOH patients, helping to differentiate these two forms of chronic headache (56).

Perrotta et al. (57) explored spinal cord pain processing by studying the reflex threshold, area, and temporal summation threshold (TST) of the lower limb nociceptive withdrawal reflex in a group of MOH patients before and after drug withdrawal. A significantly lower reflex threshold, higher amplitude, and lower TST were detected in MOH patients before drug withdrawal compared with EM and healthy controls. All of these neurophysiological abnormalities tended to recover to normal values after withdrawal from acute medication overuse (57) in parallel with an elevation in endocannabinoid system activity (58).

In sum, in patients with CM, neurophysiological abnormalities showed a pattern similar to that of ictal episodic migraine; that is, initial sensitisation and delayed habituation. For this reason, CM was previously defined as a condition of “never-ending migraine attack” (59). Electrophysiological responses differ from those of ictal migraine only when the chronicity is due to drug abuse. In patients with MOH, SSEPs suggested a persistent cortical sensitisation; that is, initial sensitisation and delayed lack of habituation. Moreover, the recorded neurophysiological abnormalities appear to vary according to the overused drug.

Prognostication of patient outcomes and predicting treatment response

Although not explicitly performed in CM, some studies observed that fluctuations of several neurophysiological responses depend on the spontaneous clinical fluctuations in attack frequency occurring in the 6 months following the testing session. Between attacks of EM, the degree of intracortical somatosensory inhibition, the activation of the thalamocortical loops and the degree of somatosensory habituation were significantly reduced in patients experiencing a spontaneous worsening of the frequency of the attacks (60 –62). In noxious cortical evoked potentials using CO₂ laser, researchers observed reduced habituation of the main N2-P2 vertex response after both cephalic and extra-cephalic stimulation in MOH patients who did not improve after preventive treatment (54). However, MOH patients who had clinically improved, habituation was significantly more pronounced after treatment.

Measuring early treatment response

Researchers have used different neurophysiological tools to investigate biomarkers of response to treatment in CM and MOH.

In MOH patients with larger amplitude cortical responses related to pain, acute medication withdrawal and successful therapeutic intervention with onabotulinumtoxinA can both restore normal cortical activation (54,63,64).

In a study involving patients who remitted from CM to EM after preventive treatment with topiramate, the abnormally higher visual evoked magnetic response pattern shifted toward a form of EM interictally (i.e. decreased initial amplitude with subsequent deficient habituation) (65). Similarly, an anesthetic block of bilateral greater occipital nerves as treatment for CM+MOH resulted in the visual and auditory neurophysiological patterns reverting to those characterising interictal EM (i.e. deficient VEP habituation and flat intensity dependence of auditory evoked cortical potentials slope) and, in a subgroup of patients, in parallel with resolution of MO (66). The lowest degree of cortical inhibition detected using TMS in CM normalised after 1-month preventive treatment with topiramate (47).

Migraine classification using neurophysiological data

Neurophysiological data were used for reliably predicting the occurrence of EM attacks. The classification model using resting-state EEGs showed a mean accuracy of 76% for classifying interictal and preictal phases (67). Similarly, steady-state VEPs also detected pre-ictal state in EM with a 73% of accuracy (68). In a recent paper, the classification tasks were performed using a single-channel SSEP signal, and researchers attained a relatively high accuracy of above 88% in migraine ictal or interictal versus healthy controls discrimination, and above 80% in classification of migraine ictal versus migraine interictal states (69).

Although no studies have been done in CM and the number of studies is limited, the integration of new EP-EEG analytical methods to classify preictal, ictal, and interictal neurophysiological responses might be used further in predicting the evolution from episodic to CM, promoting the use of preventive treatment.

Neurochemistry: Biochemical chronic migraine biomarkers

Trigeminovascular system neuropeptides

Migraine is a complex disorder in which changes in the central modulating nociceptive inputs activate the trigeminovascular system. This activation is thought to release a number of vasoactive, potentially pain-producing neuropeptides. Considering that in CM there should be a permanent, or almost permanent, activation of the trigeminovascular system, it is possible that some of these neuropeptides could theoretically become biological and diagnostic markers of CM. Previous attempts to find a biomarker of the prodromal and aura phases have not been successful and sometimes, as in the case of glutamate levels as a potential marker of migraine aura, totally contradictory (71,72). Therefore, these will not be reviewed here.

Sensory neuropeptides

Calcitonin-gene-related peptide (CGRP), substance P, neurokinin A, amylin and cholecystokinin-8 are the vasodilator peptides found in the trigeminal cell bodies of the afferent arm of the system (72,73). To the best of our knowledge, the role of substance P and neurokinin A has not been studied in patients with CM. Substance P, an endothelium-dependent vasodilator, was suggested to be involved in nociceptive transmission. The 13- preprotachykinin that produces substance P may be differentially cleaved to yield a decapeptide with a similar profile of action and localisation in the trigeminal system, neurokinin A. Blockade of substance P, though effective in blocking plasma extravasation, has not demonstrated efficacy in migraine prevention. In addition, studies testing substance P in plasma as a potential migraine biomarker have provided contradictory results (74). Amylin is a 37 amino acid peptide structurally related to CGRP, with vasodilatory and pronociceptive actions (75,76). A very recent study has shown that interictal amylin levels are elevated in peripheral blood in a series of CM patients, suggesting that this peptide could play a role in migraine chronification (77).

The powerful vasodilator CGRP is a very abundant peptide in the trigeminal system and is able to induce migrainous pain when injected in migraine sufferers, which indicates a crucial role in the pathophysiology of migraine pain (78). CGRP seems to play a relevant role in the development of sensitisation of the trigeminal pain pathways, which leads to migraine pain chronification (79). CGRP is the only neurotransmitter reliably shown to be released acutely during the migraine attack (80 –82). In addition, peripheral plasma (83,84), saliva (85 –87) and cerebrospinal fluid (88 –91) CGRP levels have been shown to be increased in interictal, headache-free periods in CM patients as compared to controls without a history of migraine and EM patients (80 –83). Current results indicate that interictal concentrations of CGRP in serum from peripheral blood samples and saliva have a good sensitivity and specificity for CM detection (79,80). Increased CGRP levels have been shown to correlate with the response to onabotulinumtoxinA in CM (84,92) and this treatment significantly decreases CGRP levels as measured 1 month after pericranial injections (93). Therefore, interictal CGRP concentration from peripheral blood samples might be a specific and sensitive biomarker in CM in the context of a patient with a compatible clinical history. However, such a role must still be taken with caution. A recent study did not find differences in interictal serum CGRP levels in EM, CM and healthy controls (94). Even though the numbers of CM patients included in this study are lower compared to those which found increased CGRP levels (83,84) and CGRP levels in healthy controls are higher than those found in all previous studies, it is clear that there are a number of clinical and methodological doubts that should be clarified before CGRP levels could be considered a formal CM biomarker. For instance, the very short half-life of the heterogeneous ELISA tests currently in the market might influence results. It is unknown whether increased CGRP levels are the actual consequence of trigeminovascular system activation or if they are a chronic pain or chronic headache marker, or even a marker for a specific comorbidity of CM, though preliminary data indicate CGRP levels do not change in chronic tension-type headache (95), cervicogenic headache (96) or due to fibromyalgia (83,97). In addition, the use of peripheral CGRP as a CM biomarker is limited by the high interpersonal variability. Thirty percent of CM patients have a peripheral level of CGRP that is comparable to those in the healthy controls. It is possible that these patients suffer from other headaches mimicking CM. Alternatively, these data could suggest that CGRP is not the only neuropeptide involved in pain generation and maintenance, and that in some patients it might not be the most relevant.

Parasympathetic neuropeptides

Besides the trigeminal nerve, activation of the cranial efferent parasympathetic arm of the trigemino-vascular system also plays a role in migraine pathophysiology (72). The autonomic nervous system is involved in migraine as reflected by the “general” (nausea, vomiting, polyuria or diarrhea) and “cranial” (conjunctival injection, lacrimation, nasal congestion, etc) symptoms and signs appearing during attacks. A recent study has shown that around 80% of women with CM have at least one cranial autonomic parasympathetic symptom coinciding with pain (98). It has been suggested that parasympathetic outflow to cephalic vasculature may trigger activation and sensitisation of perivascular sensory afferents and thereby contribute to migraine pain chronification (99).

Goadsby and Edvinsson showed, more than 20 years ago, an increase in vasoactive intestinal peptide (VIP) during cluster headache attacks as a reflection of the intense parasympathetic activation (100). Few studies have tested VIP levels in migraine and such information is limited in CM. Interictal serum VIP levels have been shown to be increased in CM, though this increase was less pronounced and consistent as that seen for CGRP (92,101). VIP, but not CGRP, levels correlate with the presence and degree of cranial parasympathetic symptoms as measured by an ad-hoc quantitative scale (102). These data suggest that VIP levels could be an indication of the degree of activation of cranial parasympathetic system in CM patients.

Experimental data support a role for pituitary adenylate-cyclase-activating polypeptide (PACAP) in migraine pathophysiology. Unlike VIP, PACAP localises to both parasympathetic and sensory systems (103), which could explain why PACAP infusion produces a migrainous headache (104). However, data on PACAP levels in migraine patients are scarce and inconsistent. For instance, while PACAP levels were seen to be increasing in jugular samples during acute migraine attacks (105), decreased interictal PACAP levels (as compared to non-headache subjects or tension-type headache patients), which normalise during attacks, have been shown in EM patients (106,107). Contrary to CGRP and VIP; interictal serum levels of PACAP have been shown to be in the range of controls in a large series of CM (108).Therefore, in spite of the evidence suggesting a relevant role of PACAP in the pathophysiology of migraine pain and in its chronification, and in contrast to CGRP and VIP, serum levels of PACAP, as measured peripheral and interictally by ELISA techniques, do not seem to be a useful biomarker in CM, though again current data should be interpreted with caution.

Glial neuropeptides

Glial cells are abundant in the trigeminovascular system and mainly in the trigeminal ganglion. S100B is a calcium-binding protein, found in the cytoplasm of glial cells in the central nervous system, which is released in response to inflammatory stimuli. Previous works analysing S100B in migraineurs have offered contradictory results. The first study in CM has shown S100B levels in the range of healthy controls (109).

Discussion

There is supporting evidence that CM is associated with structural and functional changes in the brain that are measurable using imaging, neurophysiological, and neurochemical techniques. In addition to yielding insights into CM pathophysiology, the quantification, combination, and modeling of these changes can contribute to the development of diagnostic, prognostic, and predictive biomarkers (see Tables 1, 2, 3).

Table 1.

Brain imaging results that could contribute to developing chronic migraine biomarkers.

Publication	Measures	Patient Cohorts	Findings
Headache frequency is correlated with imaging findings
Kim 2008 (10)	Gray matter volume	• EM (n = 20)• HC (n = 33)	• EM with less gray matter in insula, motor/premotor, prefrontal, cingulate, posterior parietal, orbitofrontal• Negative correlations between gray matter volume and lifetime headache frequency
Valfre 2008(11)	Gray matter volume	• Migraine (n = 27) including EM (n = 16) and CM (n = 11)• HC (n = 27)	• Migraine with less gray matter than HC in superior temporal, inferior frontal, precentral• CM associated with less gray matter than EM in ACC, amygdala, parietal operculum, middle and inferior frontal, insula• Correlation between gray matter volume of ACC and migraine frequency
Schwedt 2014(12)	Pain-induced activation (heat)	• EM (n = 24)• HC (n = 27)	• EM greater activation in lentiform nucleus, fusiform gyrus, subthalamic nucleus, hippocampus, middle cingulate, premotor, somatosensory, dorsolateral prefrontal• EM less activation in precentral, superior temporal• Correlations between activation strength with headache frequency and years with migraine
Schulte 2017(13)	Pain-induced activation (ammonia)	• EM (n = 18)• CM (n = 17)• HC (n = 19)	• CM with stronger hypothalamic activation compared to HC• CM with stronger hypothalamic activation compared to EM, both groups with headaches at time of MRI
Androulakis 2017(14)	Resting functional connectivity	• CM (n = 29)• HC (n = 19)	• CM with less coherent default mode, salience, and central executive networks• Moderate to severe headache frequency associated with decreased salience network and central executive network connectivity
Lee 2019(16)	Resting functional connectivity	• EM (n = 44)• CM (n = 18)	• CM with greater degree centrality in the pain matrix
Prognostication and tracking of clinical outcomes
Liu 2016(19)	Gray matter volume	• Migraine (n = 122), including EM (n = 83) and CM (n = 39)• HC (n = 31)	• Right hippocampal volume positively associated with migraine improvement at two years
Lai 2016(21)	Gray matter volume	• CM, half with medication overuse (n = 66)• HC (n = 33)	• CM with medication overuse who subsequently responded to treatment had greater (i.e. more normal) orbitofrontal cortex volume at baseline
Messina 2018(20)	Gray matter volume	• EM (n = 73), including n = 24 with longitudinal imaging• HC (n = 27), including n = 25 with longitudinal imaging	• Increased attack frequency at follow-up (mean = 4 years) associated with volume changes in pain processing regions
Measuring treatment response
Russo 2017(22)	Pain-induced activation (heat)	• EM (n = 16)	• “Normalisation” of BOLD response in the anterior cingulate cortex following treatment with external trigeminal neurostimulation
Krebs 2018(23)	Resting functional connectivity	• CM (n = 10)• HC (n = 10)	• “Normalisation” of connectivity in salience and executive networks following treatment with sphenopalatine ganglion blocks
Classification
Schwedt 2015(26)	Cortical thickness, cortical surface area, volumes	• EM (n = 51)• CM (n = 15)• HC (n = 54)	Classification accuracy:• Migraine vs. HC = 68%• EM vs. HC = 67%• CM vs. HC = 86%• CM vs. EM = 84%
Zhang 2016(24)	Resting functional connectivity	• EM (without aura) (n = 21)• HC (n = 28)	Classification accuracy = 84%
Chong 2017(30)	Resting functional connectivity	• Migraine (n = 58), including EM (n = 51) and CM (n = 7)• HC (n = 50)	Classification accuracy = 86.1%Those with longer migraine disease duration were easier to classify
Schwedt 2017(32)	Cortical thickness, cortical surface area, volumes	• Migraine (n = 66)• HC (n = 54)	Automated sub-classification on brain structure identified two migraine subgroupsThe subgroups differed on years with migraine, severity of disability, and cutaneous allodynia severity
Chen 2018(34)	Gray matter volume	• Migraine (n = 56), including EM (n = 25) and CM (n = 24)• TTH (n = 49), including ETTH (n = 25) and CTTH (n = 24)• HC (n = 43)	Classification accuracy, migraine vs. TTH: Area under the curve = 0.806
Chen 2019(29)	Gray matter volume	• EM (n = 18)• CM (n = 16)• HC (n = 21)	Classification accuracy based on hypothalamic volume:• EM vs. HC = 0.77• CM vs. HC = 0.90
Dominguez 2019(27)	Iron deposition	• EM (n = 57)• CM (n = 55)• HC (n = 25)	Classification accuracy for CM:• Iron volume in red nucleus: sensitivity = 80%, specificity = 71%• Iron volume in periaqueductal gray: Sensitivity = 93%, specificity = 97%
Mu 2020(28)	Resting functional connectivity	• Low-frequency migraine (defined as fewer than 8 headache days per month) (n = 120)• High frequency migraine (defined as 8 or more headache days per month) (n = 59)	Classification accuracy: Area under the curve = 0.79

BOLD: blood oxygen level detection; CM: chronic migraine; CTTH: chronic tension-type headache; EM: episodic migraine; ETTH: episodic tension-type headache; HC: healthy control; TTH: tension-type headache.

Table 2.

Neurophysiology results that could contribute to the development of chronic migraine biomarkers.

Publication	Goal	Measures	Patient cohorts	Findings
Headache frequency is correlated with neurophysiological finding
de Tommaso et al., Pain 2003(38)	To show changes in nociceptive brain responses related to dysfunction of pain elaboration at the cortical level	Heat pain thresholds and event-related potentials following CO₂-LEPs in hand and facial regions	HC = 15CM = 25MO = 15	Larger amplitude in cortical responses related to pain in CM, which was less suppressed by distractive tasks
Sohn et al., JHP 2016(39)	To investigate whether there are differences between patients with MO and patients with CM in trigeminal pain processing at the brainstem and cortical	Simultaneous recordings of nociceptive BR and PREP	HC = 40CM = 30MO = 38	Compared with HC, MO showed larger while CM smaller amplitudes values. Both MO and CM patients showed larger amplitude of PREP compare to HC
de Tommaso et al., Headache 2005(40)	To perform a topographic and dipolar analysis of nociceptive-evoked responses	Source location analysis of P2 component of LEP	HC = 15CM = 16MO= 10	Distinctive plastic remapping of the pain process over the anterior cingulate cortex in CM
Coppola et al., JHP 2013(41)	To analyse thalamocortical activity and sensitisation/habituation processes of the somatosensory system	SSEPs initial amplitude and delayed habituation	HC = 20CM = 19MO = 15MI = 19	Both MI and CM patients were characterised by higher initial amplitudes and by diminishing response (habituation) over sequential block averages, resulting in a “transient” central sensitisation
Coppola et al., Clin Neurophys 2020(42)	To study lateral inhibition and habituation/sensitisation in the somatosensory cortex of patients with CM and to identify correlations with clinical migraine features	Percentage of somatosensory lateral inhibition (LI)	HC = 17CM = 16	Despite in CM the percentage of LI was negatively correlated with the monthly headache-day number, it did not differ from that of HC
Hsiao et al., Cephalalgia 2017(43)	To explore if somatosensory gating is altered in migraine and linked to migraine chronification	Paired pulse somatosensory stimuli and gating ratios.	HC = 36CM = 22MO = 21	The gating ration increased from HC to CM, with MO falling in between. Somatosensory gating was associated with headache frequency
Kalita et al., Clin J Pain 2014(44)	To study sensitisation/habituation in the somatosensory system and correlation with the presence of allodynia	SSEPs initial amplitude and delayed habituation	HC = 29CM = 23MOH = 31MO = 42	These neurophysiological abnormalities of CM are proportional to the duration of the headache chronification phase. Patients experiencing cutaneous allodynia exhibited greater SSEP sensitisation compared with those without allodynia
De Marinis et al., Clin Neurophys 2007(45)	To study habituation of the trigeminal system in CM and its relationship with various clinical features of migraine	R2 component of BR	HC = 35CM = 35	Lack of R2 BR habituation in CM. Negative correlation between the degree of habituation and the mean frequency of the attacks
Chen et al., Pain 2011(46)	To verify whether the level of cortical excitability differs between CM and MO	VEF measurements by MEG	HC = 32CM = 25MO = 29MI = 9	Higher initial mean block amplitude followed by habituation was observed in recordings of VEF (P100m) both in patients with CM and with MI compared with MO patients
Aurora et al., Headache 2005(47)	To verify whether cortical excitability differs between CM and MO patients	Magnetic suppression of perceptual accuracy using TMS	HC = 5CM+MOH = 5MO = 5	CM patients exhibited the lowest suppression index compared with controls, with MO falling in between
Aurora et al., Headache 2007(48)	To verify whether cortical excitability differs between CM and MO patients	Magnetic suppression of perceptual accuracy using TMS	HC = N.S. MO = N.S. CM+MOH = 25	CM patients exhibited the lowest suppression index compared with controls, with EM falling in between
Cosentino et al., Pain 2014(49)	To study changes in cortical excitability and homeostatic plasticity related to migraine chronification	Brief trains of high-frequency repetitive TMS	HC = 20CM = 14MO = 66MA = 48	Inhibitory response in CM patients during trains of TMS at 5 Hz, instead of a progressive physiological facilitation
Ozturk et al., J Neurol 2002(50)	To evaluate cortical excitability in CM and to assess similarities between CM and MO patients	Thresholds, latencies, and amplitudes of MEP and duration of CSP	HC = 20CM = 20MO = 20	Longer duration of CSP in patients with CM
Coppola et al., BMC Neurol 2010(51)	To investigate sensitisation/habituation of the somatosensory system, whether they vary according to the drug overused, and whether they differ between patients with MI and MO and those with MOH	SSEPs grand-average, early amplitude response and delayed habituation	HC = 42MOH = 29MO = 41MI = 23	In patients with MOH, the initially higher SSEP amplitudes were further increased during stimulus repetition, resulting in a “persistent” sensitisation. These neurophysiological abnormalities further depend on the acute medication overused. Sensitisation depends on the duration of the chronic headache phase
Di Lorenzo et al., Cephalalgia 2012(52)	To searching for differences in cortical excitability in patients with MOH who underwent ACE I/D polymorphism analysis	SSEPs initial amplitude and delayed habituation	MOH = 43	In D/D carriers, habituation slope correlated positively with the duration of the overuse phase. In D/D carriers the degree of initial sensitisation depended on the type of medication overused
Siniatchkin et al., Cephalalgia 1998(53)	To investigate for possible neurophysiological features of clinical chronification of headache	Amplitude and habituation of contingent negative variation recording early and late components, and of PINV	HC = 15MOH = 15	MOH patients exhibited significant reduced negativity of the late component of PINV. Deficient habituation during contingent negative variation cognitive potentials recordings were observed in both MO and CM patients
Ferraro et al., Headache 2012(54)	To investigate habituation to experimental pain in a group of MOH patients	CO₂ LEPs amplitude and habituation	HC = 14MOH = 14	MOH patients exhibited deficient habituation during laser evoked potentials recordings
Currà et al., Cephalalgia 2011(55)	To investigate whether MOH is associated with changes in cortical inhibition compared with HCs and MO	CSP evoked by TMS over the primary motor cortex and recorded from contracted perioral muscles	HC = 13MO = 12MOH = 40	Despite the overall similarity in CSP duration between MOH patients and healthy controls, subgroup analysis revealed that CSP duration was significantly shorter in triptan overusers than in the NSAID or triptan-plus-NSAID overuser groups. In subgroups, CSP further depended on the number of medications overused
Perrotta et al., Cephalalgia 2010(57)	To verify whether MOH patients have abnormal processing of pain stimuli at the spinal level and defective functioning of the diffuse noxious inhibitory controls	Threshold, the area and the TST of the nociceptive withdrawal reflex before, during and after activation of the diffuse noxious inhibitory controls by means of the cold pressor test	HC = 23MOH = 31MO = 28	Lower reflex threshold, higher amplitude, and lower TST were detected in MOH patients before drug withdrawal compared with MO and healthy controls. All these neurophysiological abnormalities tended to recover to normal values after withdrawal from acute medication overuse
Perrotta et al., Headache 2012(58)	To investigate for a possible relationship between the functional activity of the endocannabinoid system and the facilitation of pain processing in MOH	Temporal summation threshold of the nociceptive withdrawal reflex, before and after drug withdrawal	HC = 14MOH = 27	Compared with HCs, MOH patients showed reduced temporal summation threshold and increased related pain sensation in MOH. This neurophysiological pattern improved after drug withdrawal and in parallel with elevation in endocannabinoid system activity
Prognostication and prediction of outcomes
Restuccia et al., Clin Neurophys 2012(60)	To verify whether the recording of thalamocortical activity could represent a marker to evaluate migraine progression in the next 6 months	High-frequency oscillations embedded in the SSEPs, reflecting the activity of the thalamocortical drives and of primary cortical activation	HC = 21MO+MA = 28	The activation of the thalamocortical loops was significantly reduced in patients experiencing a spontaneous worsening of the frequency of the attacks
Restuccia et al., Cephalalgia 2013(61)	To verify whether the recording of N20 SSEP recovery cycle could represent a marker to evaluate migraine progression in the next 6 months	N20 SSEP recovery cycle after paired pulse stimulation, which is likely to reflect the functional refractoriness of primary somatosensory cortex neurons	HC = 18MO+MA = 21	The degree of intracortical somatosensory inhibition was significantly reduced in patients experiencing a spontaneous worsening of the frequency of the attacks
Restuccia et al., Cephalalgia 2014(62)	To verify whether the recording of N20 SSEP habituation could represent a marker to evaluate migraine progression in the next 6 months	N20 SSEP amplitude habituation	HC = 18MO+MA = 25	The degree of somatosensory habituation was significantly reduced in patients experiencing a spontaneous worsening of the frequency of the attacks
Ferraro et al., Headache 2012(54)	To verify whether LEP habituation could predict clinical improvement after preventive treatment	CO₂ laser-evoked potentials (LEPs) amplitude and habituation	HC = 14MOH = 14	Reduced habituation of the main N2-P2 vertex response after both cephalic and extra-cephalic stimulation in MOH patients who did not improve after preventive treatment. Whereas MOH patients who had clinically improved, habituation was significantly more pronounced after treatment
Measuring treatment response
Ferraro et al., Headache 2012(54)	To verify whether acute medication withdrawal can modify pain-processing	CO₂ laser-evoked potentials (LEPs) amplitude and habituation	HC = 14MOH = 14	Acute medication withdrawal can restore normal cortical activation
Ayzenberg et al., Cephalalgia 2006(63)	To verify whether acute medication withdrawal can restore normal pain processing at the brainstem and cortical level	Simultaneous recording of nociceptive BR and PREP following electrical stimulation of the forehead	HC = 15MOH = 29MO = 16	Acute medication withdrawal can restore normal cortical activation. No effects at the brainstem level
de Tommaso et al., Toxin 2016(64)	To evaluate pain processing modifications after single onabotulintoxinA administration	CO₂ LEPs before and 7 days after onabotulintoxinA administration	CM = 20	Successful therapeutic intervention with onabotulinumtoxinA can restore normal cortical activation
Chen et al., Cephalalgia 2012(65)	To investigate the changes of visual cortical excitability in CM patients who remitted to MO from CM after treatment	VEF measurements by MEG before and 3 months after topiramate (50–100 mg per day)	CM = 25	The abnormally higher visual evoked magnetic response pattern of CM patients shifted toward a form of MO interictally after preventive treatment with topiramate
Viganò et al., JHP 2018(66)	To verify whether neurophysiological measures can predict treatment response	Habituation of VEP amplitude and IDAP before and 1 week after the GON-B	HC = 19CM+MOH = 17	GON block resulted in the visual and auditory neurophysiological patterns reverting to those characterising interictal MO. This was more evident for IDAP, so it was considered as potentially useful predictor of GON-B efficacy
Classification
Cao et al., Cephalalgia 2018(67)	To develop of an EEG-based classification model to recognise the difference between interictal and preictal phases of migraine	Complexity of EEG at the prefrontal and occipital areas in 4 migraine phases (interictal, preictal, ictal, and postictal)	HC = 40MO = 40	The classification model using resting-state EEGs showed a mean accuracy of 76% for classifying interictal and preictal phases
Ko et al., IEEE 2013(68)	To develop a classification system to determine different migraine states	Steady-state VEP habituation in four migraine phases (interictal, preictal, ictal, and postictal)	MO = 29	Steady-state VEPs also detected preictal state in MO with a 73% of accuracy
Zhu et al., Cephalalgia 2019(69)	To verify whether SSEP can be considered as a practical diagnostic tool and a sensitive predictor for migraine classification	SSEPs amplitudes, habituation, and thalamocortical/cortical activation in two migraine phases (interictal, preictal, ictal, and postictal)	HC = 15MO = 14MA = 15MI = 13	SSEP signal attained a relatively high accuracy of above 88% in MI or MO versus HCs discrimination, and above 80% in classification of MI versus MO states

ACE: angiotensin converting enzyme; BR: blink reflex; CM: chronic migraine; CSP: cortical silent period; GON-B: greater occipital nerve block; HC: healthy control; IDAP: intensity dependence of auditory evoked potentials; LEP: laser evoked potential; LI: lateral inhibition; MA: migraine with aura; MEG: magnetoelectroencephalography; MEP: motor evoked potential; MI: migraine during the attack; MO: episodic migraine without aura between attacks; MOH: medication overuse headache; PINV: post-imperative negative variation; PREP; pain-related evoked potential; SSEP: somatosensory evoked potential; TMS: transcranial magnetic stimulation; TST: temporal summation threshold; VEF: visual evoked magnetic field; VEP: visual evoked potential.

Table 3.

Interictal trigeminovascular neuropeptides in different biological fluids that could contribute to the development of chronic migraine biomarkers.

Publication	Molecule, fluid	Patients cohorts	Main findings
Cernuda-Morollón et al, Neurology 2013 (83)	CGRP, blood	HC = 31CM = 103EM = 43	CGRP levels significantly increased in CM as compared to HC, and EM patients
Domínguez et al, Headache 2018 (84)	CGRP, blood	HC = 24CM = 62	CGRP levels significantly increased in CM as compared to controls and predicted efficacy of onabotulinumtoxin-typeA
Lee et al, J Headache Pain 2018 (94)	CGRP, blood	HC = 27CM = 34EM = 96	CGRP levels did not significantly differ among the three groups
Irimia et al, in press, Cephalalgia (77)	CGRP and amylin, blood	HC = 68CM = 191EM = 58	Both CGRP and amylin levels significantly increased in CM vs. HC. Amylin, but not CGRP, significantly increased in CM vs. EM
Fekrazad et al. Neurol Sci 2018 (86)	CGRP, blood and gingival crevicular fluid	HC = 15 CM = 24	CGRP significantly increased in CM patients in both fluids
Jang et al, Oral Dis 2011 (87)	CGRP, blood and saliva	HC = 36CM = 33	CGRP significantly increased in CM as compared to controls in both fluids
Sarchielli et al, J Pain 2007 (91)	CGRP, cerebrospinal fluid	HC = 20CM = 20	CGRP significantly increased in CM patients
Sarchielli et al, J Headache Pain 2002 (89)	CGRP, cerebrospinal fluid	HC = 20CM = 25	CGRP significantly increased in CM patients
Peres et al. Cephalalgia 2004 (90)	CGRP, cerebrospinal fluid	HC = 20CM = 30	CGRP significantly increased in CM patients
Cernuda-Morollón et al. 2015 (101)	VIP, blood	HC = 33CM = 119EM = 51	VIP levels in CM significantly higher than HC and numerically higher than EM
Cernuda-Morollón et al, Headache 2016 (108)	PACAP, blood	HC = 32CM = 86EM = 35	PACAP levels did not differ among groups
Riesco et al, Headache 2020 (109)	S100B, blood	HC = 29CM = 43EM = 19	S100B levels did not differ among groups

CGRP: calcitonin gene-related peptide; HC: healthy controls; EM: episodic migraine; CM: chronic migraine; VIP: vasoactive intestinal peptide; PACAP: pituitary adenylate cyclase-activating peptide; S100B: S100 calcium-binding protein.

Imaging data have shown promise for contributing to CM biomarkers. However, they need to be optimised and then validated in independent datasets. This should include testing model accuracy on completely independent cohorts and data collected from different scanners at different institutions.

Neurophysiological data have also demonstrated the presence of functional changes in the migraine brain. Normal habituation, central sensitisation, and higher amplitude of non-noxious (SSEP and VEP) and noxious (LEP and PREP) evoked responses may play important roles in the chronic nature of migraine with or without MO.

At a neurochemical level, there is evidence of changes that occur in the brain when migraine becomes chronic. Technical issues with measuring these neuropeptides must be resolved. If resolved, the future might see a practical neurochemical biomarker for migraine.

To optimise the accuracy of CM biomarkers, future models should include multiple data types such as structural and functional brain imaging, results of neurophysiologic studies, genetic profiles, patient demographics, migraine symptoms, neuropeptide levels, medical comorbidities, and treatment responses. However, a clinically useful biomarker needs to not only be highly accurate, but it also must consist of data that are relatively easy and practical to obtain.

In the future, CM biomarkers might include epigenetic and deep clinical phenotyping data alone or in combination with imaging, neurophysiological and biochemical data. Whether these biomarkers will eventually become useful for further defining migraine and for practicing precision medicine will depend upon their ultimate accuracy, their generalisability, and the practicality of collecting the data that contribute to the biomarker. Identification and development of biomarkers with utility to diagnose CM, to prognosticate and evaluate treatment response would greatly advance the field.

Conclusions

Existing studies show feasibility for developing diagnostic, prognostic, and predictive CM biomarkers based on imaging, neurophysiological, and neurochemical data.

Article highlights

Validated chronic migraine biomarkers could improve diagnostic, prognostic, and predictive abilities, as well as increase knowledge on pathophysiology.

Structural and functional changes that occur in the brain with the progression from episodic to chronic migraine might be useful for developing biomarkers.

Chronic migraine biomarkers would help us move closer to practising precision medicine and improve patient care.

Footnotes

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: PPR has received honoraria for participation in clinical trials and contribution to advisory boards or presentations from Allergan, Almirall, Amgen, Biohaven, Chiesi, Electrocore, Eli Lilly, Medscape, Novartis, and Teva. Financial support for research projects was provided by Allergan and Novartis. Headache research is supported by la Caixa Foundation, MINECO, AGAUR, Fundacion La Marato TV3, Instituto Investigacion Carlos III, Migraine Research Foundation, Mutual Medica, and PERIS. She serves as an Associate Editor for Cephalalgia, Frontiers in Neurology (Headache Section), Editor for Revista de Neurologia, and on the Advisory Board of The Journal of Headache and Pain.

GC has received contribution to advisory boards or presentations from Novartis, Teva, and Eli Lilly. He serves as an Associate Editor for Cephalalgia, BMC Neurology – Pain section, and on the Advisory Board of The Journal of Headache and Pain.

JP has served on the Advisory Boards of Allergan, Lilly and Novartis-Amgen. Financial support for research projects was provided for Instituto de Investigación Carlos III. He serves as an Associate Editor for Headache.

TJS receives research support from the American Migraine Foundation, Henry Jackson Foundation, National Institutes of Health, Patient-Centered Outcomes Research Institute, US Department of Defense, and Amgen. Within the past 12 months, he has received personal compensation for serving as a consultant or advisory board member for Alder, Allergan, Amgen, Biohaven, Cipla, Click Therapeutics, Eli Lilly, Equinox, Weber & Weber, and XoC Pharmaceuticals. He holds stock options in Aural Analytics and Nocira. He serves as an Associate Editor for Headache and Cephalalgia.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Author contributions

All authors contributed equally to the draft and final version of the paper.

ORCID iD

Gianluca Coppola

References

Natoli

Manack

Dean

, et al. Global prevalence of chronic migraine: A systematic review. Cephalalgia 2010; 30: 599–609.

Bigal

Serrano

Buse

, et al . Acute migraine medications and evolution from episodic to chronic migraine: a longitudinal population-based study. Headache 2008; 48: 157–168.

Vos

Abajobir

Abbafati

, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017; 390: 1211–1259.

Stovner

Nichols

Steiner

, et al. Global, regional, and national burden of migraine and tension-type headache, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol 2018; 17: 954–976.

FDA-NIH Biomarker Working Group. BEST (Biomarkers, EndpointS, and other tools) resource. Silver Spring, MD: Food and Drug Administration (US); Bethesda (MD): National Institutes of Health (US), www.ncbi.nlm.nih.gov/books/NBK326791/ (2016, accessed 22 September 2017).

Durham

Papapetropoulos

Biomarkers associated with migraine and their potential role in migraine management. Headache 2013; 53: 1262–1277.

Headache Classification Committee of the International Headache Society. The International Classification of Headache Disorders, 3rd edition (ICHD-3). Cephalalgia 2018; 38: 1–211.

Chalmer

Hansen

Dodick

, et al. Proposed new diagnostic criteria for chronic migraine. J Headache Pain 2018; 19: 1–8.

Diener

Dodick

Goadsby

, et al. Chronic migraine – classification, characteristics and treatment. Nat Rev Neurol 2012; 8: 162–171.

10.

Kim

Suh

Seol

, et al. Regional grey matter changes in patients with migraine: A voxel-based morphometry study. Cephalalgia 2008; 28: 598–604.

11.

Valfrè

Rainero

Bergui

, et al. Voxel-based morphometry reveals gray matter abnormalities in migraine. Headache 2008; 48: 109–117.

12.

Schwedt

Chong

Chiang

, et al. Enhanced pain-induced activity of pain-processing regions in a case-control study of episodic migraine. Cephalalgia 2014; 34: 947–958.

13.

Schulte

Allers

May

Hypothalamus as a mediator of chronic migraine. Neurology 2017; 88: 2011–2016.

14.

Androulakis

Krebs

Peterlin

, et al. Modulation of intrinsic resting-state fMRI networks in women with chronic migraine. Neurology 2017; 89: 163–169.

15.

Androulakis

Krebs

Jenkins

, et al. Central executive and default mode network intranet work functional connectivity patterns in chronic migraine. J Neurol Disord 2018; 6: 393.

16.

Lee

Park

Cho

, et al. Increased connectivity of pain matrix in chronic migraine: A resting-state functional MRI study. J Headache Pain 2019; 20: 29.

17.

Woldeamanuel

DeSouza

Sanjanwala

, et al. Clinical features contributing to cortical thickness changes in chronic migraine – a pilot study. Headache 2019; 59: 180–191.

18.

DeSouza

Woldeamanuel

Sanjanwala

, et al. Altered structural brain network topology in chronic migraine. Brain Struct Funct 2020; 225: 161–172.

19.

Liu H-Y, Chou K-H, Lee P-L, et al. Hippocampus and amygdala volume in relation to migraine frequency and prognosis. Cephalalgia 2016; 37: 1329–1336.

20.

Messina

Rocca

Colombo

, et al. Gray matter volume modifications in migraine: A cross-sectional and longitudinal study. Neurology 2018; 91: e280–e292.

21.

Lai

T-H

Chou

K-H

Fuh

J-L

, et al. Gray matter changes related to medication overuse in patients with chronic migraine. Cephalalgia 2016; 36: 1324–1333.

22.

Russo

Tessitore

Esposito

, et al. Functional changes of the perigenual part of the anterior cingulate cortex after external trigeminal neurostimulation in migraine patients. Front Neurol 2017; 8: 1–9.

23.

Krebs

Rorden

Androulakis

XM.

Resting state functional connectivity after sphenopalatine ganglion blocks in chronic migraine with medication overuse headache: A pilot longitudinal fMRI study. Headache 2018; 58: 732–743.

24.

Zhang

, et al. Discriminative analysis of migraine without aura: Using functional and structural MRI with a multi-feature classification approach. PLoS One 2016; 11: e0163875.

25.

Gaw

Schwedt

Chong

, et al. A clinical decision support system using multi-modality imaging data for disease diagnosis. IISE Transactions on Healthcare Systems Engineering 2018; 8: 36–46. DOI: 10.1080/24725579.2017.1403520.

26.

Schwedt

Chong

, et al. Accurate classification of chronic migraine via brain magnetic resonance imaging. Headache 2015; 55: 762–777.

27.

Domínguez

López

Ramos-Cabrer

, et al. Iron deposition in periaqueductal gray matter as a potential biomarker for chronic migraine. Neurology 2019; 92: E1076–E1085.

28.

Chen

Quan

, et al. Neuroimaging features of whole-brain functional connectivity predict attack frequency of migraine. Hum Brain Mapp 2020; 41: 984–993.

29.

Chen

Liu

, et al. Volume of hypothalamus as a diagnostic biomarker of chronic migraine. Front Neurol 2019; 10: 1–11.

30.

Chong

Gaw

, et al. Migraine classification using magnetic resonance imaging resting-state functional connectivity data. Cephalalgia 2017; 37: 828–844.

31.

Schwedt

, et al. Migraine subclassification via a data-driven automated approach using multimodality factor mixture modeling of brain structure measurements. Headache 2017; 57: 1051–1064.

32.

Schwedt

Chong

Peplinski

, et al. Persistent post-traumatic headache vs. migraine: An MRI study demonstrating differences in brain structure. J Headache Pain 2017; 18: 1–8.

33.

Giorgio

Lupi

Zhang

, et al. Changes in grey matter volume and functional connectivity in cluster headache versus migraine. Brain Imaging Behav 2019; 14: 496–504.

34.

Chen

Chou

Lee

, et al. Comparison of gray matter volume between migraine and “strict-criteria” tension-type headache. J Headache Pain 2018; 19: 4.

35.

Chong

Aguilar

Schwedt

TJ.

Altered hypothalamic region covariance in migraine and cluster headache: A structural MRI study. Headache 2020; 60: 553–563.

36.

Chong

Peplinski

Berisha

, et al. Differences in fibertract profiles between patients with migraine and those with persistent post-traumatic headache. Cephalalgia 2019; 39: 1121–1133.

37.

Dumkrieger

Chong

Ross

, et al. Static and dynamic functional connectivity differences between migraine and persistent post-traumatic headache: A resting-state magnetic resonance imaging study. Cephalalgia 2019; 39: 1366–1381.

38.

de Tommaso

Valeriani

Guido

, et al. Abnormal brain processing of cutaneous pain in patients with chronic migraine. Pain 2003; 101: 25–32.

39.

Sohn

Kim

Choi

HC.

Differences in central facilitation between episodic and chronic migraineurs in nociceptive-specific trigeminal pathways. J Headache Pain 2016; 17: 35.

40.

de Tommaso

Losito

Difruscolo

, et al. Changes in cortical processing of pain in chronic migraine. Headache 2005; 45: 1208–1218.

41.

Coppola

Iacovelli

Bracaglia

, et al. Electrophysiological correlates of episodic migraine chronification: Evidence for thalamic involvement. J Headache Pain 2013; 14: 76.

42.

Coppola

Cortese

Bracaglia

, et al. The function of the lateral inhibitory mechanisms in the somatosensory cortex is normal in patients with chronic migraine. Clin Neurophysiol 2020; 131: 880–886.

43.

Hsiao

F-J

Wang

S-J

Lin

Y-Y

, et al. Somatosensory gating is altered and associated with migraine chronification: A magnetoencephalographic study. Cephalalgia 2017; 38: 744–753.

44.

Kalita

Bhoi

Misra

UK.

Is lack of habituation of evoked potential a biological marker of migraine?

Clin J Pain 2014; 30: 724–729.

45.

De Marinis

Pujia

Colaizzo

, et al. The blink reflex in “chronic migraine.” Clin Neurophysiol 2007; 118: 457–463.

46.

Chen

Wang

Fuh

, et al. Persistent ictal-like visual cortical excitability in chronic migraine. Pain 2011; 152: 254–258.

47.

Aurora

Barrodale

Chronicle

, et al. Cortical inhibition is reduced in chronic and episodic migraine and demonstrates a spectrum of illness. Headache 2005; 45: 546–552.

48.

Aurora

Barrodale

Tipton

, et al. Brainstem dysfunction in chronic migraine as evidenced by neurophysiological and positron emission tomography studies. Headache 2007; 47: 996–1003.

49.

Cosentino

Fierro

Vigneri

, et al. Cyclical changes of cortical excitability and metaplasticity in migraine: Evidence from a repetitive transcranial magnetic stimulation study. Pain 2014; 155: 1070–1078.

50.

Ozturk

Cakmur

Donmez

, et al. Comparison of cortical excitability in chronic migraine (transformed migraine) and migraine without aura: A transcranial magnetic stimulation study. J Neurol 2002; 249: 1268–1271.

51.

Coppola

Currà

Di Lorenzo

, et al. Abnormal cortical responses to somatosensory stimulation in medication-overuse headache. BMC Neurol 2010; 10: 126.

52.

Di Lorenzo

Coppola

Currà

, et al. Cortical response to somatosensory stimulation in medication overuse headache patients is influenced by angiotensin converting enzyme (ACE) I/D genetic polymorphism. Cephalalgia 2012; 32: 1189–1197.

53.

Siniatchkin

Gerber

Kropp

, et al. Contingent negative variation in patients with chronic daily headache. Cephalalgia 1998; 18: 565–569.

54.

Ferraro

Vollono

Miliucci

, et al. Habituation to pain in “medication overuse headache”: A CO₂ laser-evoked potential study. Headache 2012; 52: 792–807.

55.

Curra

Coppola

Gorini

, et al. Drug-induced changes in cortical inhibition in medication overuse headache. Cephalalgia 2011; 31: 1282–1290.

56.

Cortese

Pierelli

Pauri

, et al. Short-term cortical synaptic depression/potentiation mechanisms in chronic migraine patients with or without medication overuse. Cephalalgia 2019; 39: 237–244.

57.

Perrotta

Serrao

Sandrini

, et al. Sensitisation of spinal cord pain processing in medication overuse headache involves supraspinal pain control. Cephalalgia 2010; 30: 272–284.

58.

Perrotta

Arce-Leal

Tassorelli

, et al. Acute reduction of anandamide-hydrolase (FAAH) activity is coupled with a reduction of nociceptive pathways facilitation in medication-overuse headache subjects after withdrawal treatment. Headache 2012; 52: 1350–1361.

59.

Schoenen

Is chronic migraine a never-ending migraine attack?

Pain 2011; 152: 239–240.

60.

Restuccia

Vollono

Del Piero

, et al. Somatosensory high frequency oscillations reflect clinical fluctuations in migraine. Clin Neurophysiol 2012; 123: 2050–2056.

61.

Restuccia

Vollono

del Piero

, et al. Different levels of cortical excitability reflect clinical fluctuations in migraine. Cephalalgia 2013; 33: 1035–1047.

62.

Restuccia

Vollono

Virdis

, et al. Patterns of habituation and clinical fluctuations in migraine. Cephalalgia 2014; 34: 201–210.

63.

Ayzenberg

Obermann

Nyhuis

, et al. Central sensitization of the trigeminal and somatic nociceptive systems in medication overuse headache mainly involves cerebral supraspinal structures. Cephalalgia 2006; 26: 1106–1114.

64.

de Tommaso

Delussi

Ricci

, et al. Effects of onabotulintoxinA on habituation of laser evoked responses in chronic migraine. Toxin (Basel) 2016; 8: 163.

65.

Chen

Wang

Fuh

, et al. Visual cortex excitability and plasticity associated with remission from chronic to episodic migraine. Cephalalgia 2012; 32: 537–543.

66.

Viganò

Torrieri

Toscano

, et al. Neurophysiological correlates of clinical improvement after greater occipital nerve (GON) block in chronic migraine: Relevance for chronic migraine pathophysiology. J Headache Pain 2018; 19: 73.

67.

Cao

Lai

Lin

, et al. Exploring resting-state EEG complexity before migraine attacks. Cephalalgia 2018; 38: 1296–1306.

68.

Lai

Huang

, et al. Steady-state visual evoked potential based classification system for detecting migraine seizures. In: Proceedings of International IEEE/EMBS Conference on Neural Engineering, NER 2013. pp.1299–1302.

69.

Zhu

Coppola

Shoaran

Migraine classification using somatosensory evoked potentials. Cephalalgia 2019; 39: 1143–1155.

70.

Martínez

Castillo

Rodríguez

, et al. Neuroexcitatory amino acid levels in plasma and cerebrospinal fluid during migraine attacks. Cephalalgia 1993; 13: 89–93.

71.

Ferrari

Odink

Bos

, et al. Neuroexcitatory plasma amino acids are elevated in migraine. Neurology 1990; 40: 1582–1586.

72.

Goadsby

PJ.

Autonomic nervous system control of the cerebral circulation. Handb Clin Neurol 2013; 117: 193–201.

73.

Riesco

Cernuda-Morollón

Pascual

Neuropeptides as a marker for chronic headache. Curr Pain Headache Rep 2017; 21: 18.

74.

Tajti

Szok

Majláth

, et al. Migraine and neuropeptides. Neuropeptides 2015; 52: 19–30.

75.

Edvinsson

Goadsby

Uddman

Amylin; location, effects on cerebral arteries and on local cerebral blood flow in the cat. Scientific World J 2001; 1: 168–180.

76.

Gebre-Medhin

Mulder

Zhang

, et al. Reduced nociceptive behavior in islet amyloid polypeptide (amylin) knockout mice. Brain Res Mol Brain Res 1998; 63: 180–183.

77.

Irimia

Martínez-Valbuena

Mingúez-Olaondo

, et al. Interictal amylin levels in chronic migraine patients. A case-control study. Cephalalgia 2020 (in press).

78.

Lassen

Haderslev

Jacobsen

, et al. CGRP may play a causative role in migraine. Cephalalgia 2002; 22: 54–61.

79.

Iyengar

Ossipov

Johnson

KW.

The role of calcitonin gene-related peptide in peripheral and central pain mechanisms including migraine. Pain 2017; 158: 543–559.

80.

Goadsby

Edvinsson

The trigeminovascular system and migraine: Studies characterizing cerebrovascular and neuropeptide changes seen in humans and cats. Ann Neurol 1993; 33: 48–56.

81.

Goadsby

Edvinsson

Ekman

Vasoactive peptide release in the extracerebral circulation of humans during migraine headache. Ann Neurol 1990; 28: 183–187.

82.

Rodríguez-Osorio

Sobrino

Brea

, et al. Endothelial progenitor cells: A new key for endothelial dysfunction in migraine. Neurology 2012; 79: 474–479.

83.

Cernuda-Morollon

Larrosa

Ramon

, et al. Interictal increase of CGRP levels in peripheral blood as a biomarker for chronic migraine. Neurology 2013; 81: 1191–1196.

84.

Dominguez

Vieites-Prado

Perez-Mato

, et al. CGRP and PTX3 as predictors of efficacy of onabotulinumtoxin type A in chronic migraine: An observational study. Headache 2018; 58: 78–87.

85.

Bellamy

Cady

Durham

PL.

Salivary levels of CGRP and VIP in rhinosinusitis and migraine patients. Headache 2006; 46: 24–33.

86.

Fekrazad

Sardarian

Azma

, et al. Interictal levels of calcitonin gene related peptide in gingival crevicular fluid of chronic migraine patients. Neurol Sci 2018; 39: 1217–1223.

87.

Jang

Park

Kho

, et al. Plasma and saliva levels of nerve growth factor and neuropeptides in chronic migraine patients. Oral Dis 2011; 17: 187–193.

88.

van Dongen

Zielman

Noga

, et al. Migraine biomarkers in cerebrospinal fluid: A systematic review and meta-analysis. Cephalalgia 2017; 37: 49–63.

89.

Sarchielli

Alberti

Gallai

, et al. Brain-derived neurotrophic factor in cerebrospinal fluid of patients with chronic daily headache. Relationship with nerve growth factor and glutamate levels. J Headache Pain 2002; 3: 129–135.

90.

Peres

MFP

Zukerman

Senne-Soares

, et al. Cerebrospinal fluid glutamate levels in chronic migraine. Cephalalgia 2004; 24: 735–739.

91.

Sarchielli

Mancini

Floridi

, et al. Increased levels of neurotrophins are not specific for chronic migraine: Evidence from primary fibromyalgia syndrome. J Pain 2007; 8: 737–745.

92.

Cernuda-Morollon

Martinez-Camblor

Ramon

, et al. CGRP and VIP levels as predictors of efficacy of onabotulinumtoxin type A in chronic migraine. Headache 2014; 54: 987–995.

93.

Cernuda-Morollon

Ramon

Martinez-Camblor

, et al. OnabotulinumtoxinA decreases interictal CGRP plasma levels in patients with chronic migraine. Pain 2015; 156: 820–824.

94.

Lee

Cho

, et al. Feasibility of serum CGRP measurement as a biomarker of chronic migraine: A critical reappraisal. J Headache Pain 2018; 19: 53.

95.

Ashina

Bendtsen

Jensen

, et al. Plasma levels of calcitonin gene-related peptide in chronic tension-type headache. Neurology 2000; 55: 1335–1340.

96.

Frese

Schilgen

Edvinsson

, et al. Calcitonin gene-related peptide in cervicogenic headache. Cephalalgia 2005; 25: 700–703.

97.

Ramon

Cernuda-Morollon

Pascual

Calcitonin gene-related peptide in peripheral blood as a biomarker for migraine. Curr Opin Neurol 2017; 30: 281–286.

98.

Riesco

Perez-Alvarez

Verano

, et al. Prevalence of cranial autonomic parasympathetic symptoms in chronic migraine: Usefulness of a new scale. Cephalalgia 2016; 36: 346–350.

99.

Yarnitsky

Goor-Aryeh

Bajwa

, et al. Wolff Award: Possible parasympathetic contributions to peripheral and central sensitization during migraine. Headache United States 2003; 43: 704–714.

100.

Goadsby

Edvinsson

Human in vivo evidence for trigeminovascular activation in cluster headache. Neuropeptide changes and effects of acute attacks therapies. Brain 1994; 117: 427–434.

101.

Cernuda-Morollón

Martínez-Camblor

Alvarez

, et al. Increased VIP levels in peripheral blood outside migraine attacks as a potential biomarker of cranial parasympathetic activation in chronic migraine. Cephalalgia 2015; 35: 310–316.

102.

Riesco

Cernuda-Morollón

Martínez-Camblor

, et al. Relationship between serum levels of VIP, but not of CGRP, and cranial autonomic parasympathetic symptoms: A study in chronic migraine patients. Cephalalgia 2017; 37: 823–827.

103.

Edvinsson

Role of VIP/PACAP in primary headaches. Cephalalgia 2013; 33: 1070–1072.

104.

Vollesen

Guo

Ashina

PACAP38 dose-response pilot study in migraine patients. Cephalalgia 2017; 37: 391–395.

105.

Zagami

Edvinsson

Goadsby

PJ.

Pituitary adenylate cyclase activating polypeptide and migraine. Ann Clin Transl Neurol 2014; 1: 1036–1040.

106.

Tuka

Helyes

Markovics

, et al. Alterations in PACAP-38-like immunoreactivity in the plasma during ictal and interictal periods of migraine patients. Cephalalgia 2013; 33: 1085–1095.

107.

Han

Dong

Hou

, et al. Interictal plasma pituitary adenylate cyclase-activating polypeotide levels are decreased in migraineurs but remain unchanged in patients with tension-type headache. Clin Chim Acta 2015; 450: 151–154.

108.

Cernuda-Morollón

Riesco

Martínez-Camblor

, et al. No change in interictal PACAP levels in peripheral blood in women with chronic migraine. Headache 2016; 56: 1448–1454.

109.

Riesco

Cernuda-Morollón

Martínez-Camblor

, et al. Peripheral, interictal serum S100B levels are not increased in chronic migraine patients. Headache 2020. Epub ahead of print 17 August 2020. DOI: 10.1111/head.13919.