Quantitative EEG Power and Asymmetry Increase 36 h Before a Migraine Attack

Abstract

The aim was to estimate ictal, pre- and postictal brain function changes in migraine in a blinded paired quantitative EEG (QEEG) study. EEG recordings (n = 119) from 40 migraineurs were retrospectively classified as ictal, interictal, preictal or postictal. δ, θ, α and β power, and hemispheric asymmetry in frontocentral, temporal and occipitoparietal regions were calculated from artefact-free EEG. Power and power asymmetry were calculated for two time-windows, 36 and 72 h before/after the attack, and compared with the interictal values. Frontocentral δ power increased (P = 0.03), whereas frontocentral θ and α power tended to increase (P < 0.09) within 36 h before the next attack compared with the interictal period. Occipitoparietal (α and θ) and temporal (α) power were more asymmetric before the attack compared with the interictal baseline (P < 0.04). Ictal posterior a power increased slightly (P = 0.01). Postictal power and power asymmetry were not significantly different from interictal baseline. EEG activity seems to change shortly before the attack. This suggests that migraineurs are most susceptible to attack when anterior QEEG δ power and posterior α and θ asymmetry values are high. Changed activity patterns in cholinergic brainstem or basal forebrain nuclei and thalamo-cortical connections before the migraine attack are hypothesized.

Keywords

Migraine headache EEG QEEG migraine cycle

Introduction

Non-invasive neurophysiological methods are well suited to study the pathophysiology of migraine, and evidence in favour of a neural dysfunction has been found (1, 2). Decreased habituation (but also potentiation) to sensory stimuli as well as altered levels of cortical excitability have been reported (1–4). The results of contingent negative variation (CNV) studies seem to indicate that neural dysfunction fluctuates with the ‘migraine cycle’, resulting in pathological cortical information processing with ‘quasi-normalization’ just before and after the attack (5–8).

EEG detects activity in groups of cortical neurons, and neural rhythms can be objectively quantified with frequency analysis (QEEG) (9). This technique has been used previously in migraine, but with contradictory results (10–12). The inclusion of both preictal and postictal recordings in the analysis of interictal QEEG may explain some of the discrepancies (6). Since a pattern of either increased or decreased α power or increased asymmetry has emerged (13–19), α power and asymmetry were considered a major focus of the present study. Since increased δ and θ power have been reported in basilar migraine (20), and interictally in migraine without aura (MoA) and in migraine with visual aura (18, 21, 22), we considered slow wave power as a second major interest.

To our knowledge, only two studies have analysed the period before and/or after the attack specifically with QEEG (7, 23). However, only one of these compared both the pre- and postictal interval with an interictal interval, and both studies used relatively wide intervals (i.e. 72–96 h) in the definition of the pre(post)-ictal phase (7, 23). The day before the attack seems to be associated with most abnormalities in CNV studies (6, 8). However, a variety of neurophysiological methods have shown altered cerebral function in proximity to the attack: normalization of intensity-dependence of cortical auditory evoked potentials (24) and somatosensory evoked high-frequency oscillations (25), habituation of visual evoked potentials (VEPs) (24, 26), increased brainstem auditory evoked potential (BAEP) intensity dependence (26), changes in event-related P300 (27), sleep quality (28), dimensional EEG complexity (29), and reduced motor activity in the night before the attack (30) have been described. It should therefore be of interest to examine the QEEG in closer proximity to the attack.

This study aimed to compare regional QEEG power and asymmetry before, during and after the attack with an interictal (baseline) phase using a paired and blinded design. We specifically wanted to study changes that might occur rather close to the attack, hence 36-h intervals before/after the attack were used.

Materials and methods

Patients

Migraine patients were recruited by a newspaper advertisement. After telephone screening by nurses trained in headache research, 52 migraineurs where evaluated for inclusion by a neurologist. The diagnosis was made according to the International Headache Society classification of Headache Disorders, 2nd edn (31): MoA (1.1) and with aura (MA) (1.2). Eligible participants were men or women aged 18–65 years with two to six migraine attacks per month during the last 3 months. Exclusion criteria were coexisting frequent episodic or chronic tension-type headache, acute or chronic neurological disease, connective tissue disorder or other painful conditions, malignancy, previous craniotomy or cervical spine surgery, cardiopulmonary or cerebrovascular disease, pregnancy, medication for acute or chronic pain, neuroleptics, alcohol or drug abuse, ferromagnetic implants, antidepressive or antiepileptic drugs or migraine prophylactic drugs within 4 weeks before test. A headache questionnaire was completed and a semistructured interview was also performed by the nurse. Phonophobia and photophobia were classified on a three-point scale (none, moderate, severe). The usual headache attack rate (1, 1–3 days per month; 2, 4–7 days per month; 3, 8–14 days per month; 4, ≥ 15 days per month) and the usual headache intensity were scored (1, mild, can continue present activity; 2, moderate, can perform easy work; 3, strong, must lie down; 4, extreme, cannot lie down to rest) as estimated by the patient on four-point scales. Forty-one migraine patients (33 MoA, eight MA) entered the study. One patient was not willing to undergo the last EEG recording. Another patient had only one recording and could not contribute any EEG in the paired analysis. The reason for these dropouts was worsening of headache after the test. Thus, data from 119 EEGs were available for further classification. Patient data are reported in Table 1.

Table 1

Clinical data for migraine patients entering the study [mean (s.d.) or count]

	Migraine (n = 41)	MoA (n = 33)	MA (n = 8)
Women/men	38/3	30/3	8/0
Age (years)	37.4 (13.0)	37.4 (12.6)	37.4 (15.5)
Head circumference (cm)	56.4 (1.3)	56.5 (1.3)	55.9 (1.3)
Headache history (years)	20.6 (11.8)	19.9 (12.0)	23.3 (11.4)
Headache attack rate (1–4)	2.0 (0.6)	2.0 (0.6)	2.1 (0.6)
Headache intensity (1–4)	2.4 (0.7)	2.4 (0.7)	2.3 (0.5)
Usual headache attack duration (h)	16.2 (20.3)	16.1 (19.6)	16.6 (24.4)
Headache days last 3 months	6.5 (4.8)	6.3 (4.2)	7.5 (6.9)
Phonophobia (0–2)	1.2 (0.7)	1.2 (0.8)	1.3 (0.5)
Photophobia (0–2)	1.4 (0.7)	1.5 (0.8)	1.1 (0.6)
Days from last menstruation in premenopausal women∗	14 (4.5) n = 29	14.3 (4.8) n = 23	12.5 (3.5) n = 6

∗

Mean for the three tests.

The usual headache attack rate and headache intensity were graded by the patient on four-point scales from mild to extremely severe. Phonophobia and photophobia were classified on a three point scale (none, moderate, severe).

MA (migraine with aura) vs. MoA (migraine without aura) differences: NS (Student's t-test or χ² test). One MoA patient was recorded only once and did not contribute data to the present study. Another MoA patient withdrew after two EEGs, hence 119 EEGs from 40 patients were eligible for inclusion in paired analysis. No significant differences.

Experimental set-up

Three EEG recordings with 3–10-day intervals were performed in each patient (days A, B and C). Caffeine-containing beverages were not allowed on the day of recording. Patients with ongoing headache were allowed to use their regular analgesics. The recordings were classified later using the patient's headache diary from the time period before, during and after the test period (minimum 2 weeks before and after this period). The diary included pain characteristics, accompanying symptoms and consequences for work and leisure, enabling a correct retrospective classification of the patient's headaches in relation to the time of the EEG recordings. Recordings (day A, B or C) were classified as preictal (Y: within 36 h before attack), postictal (X: within 36 h after attack), attack (U) and interictal (Z: > 36 h from an attack) (Fig. 1). Each patient with an attack-related recording, who also had an interictal recording, was included in three migraine subgroups: YZ₃₆, UZ₃₆, XZ₃₆ for paired analyses. Afterwards, a 72-h cut-off was applied to the same patients and the data were reanalysed in order to compare our 72-h results with the 36-h results as well as with data from the literature (Table 2). Patients recorded between two closely spaced attacks were classified within the U group (three of the 13 36-h pairs). For the 36-h cut-off, nine patients had all three recordings in the same phase, and they could accordingly not contribute any EEG to the paired analysis. Among the remaining 40 patients with three (n = 39) or two (n = 1) EEGs, 31 (40 − 9) patients contributed to the paired analysis. Three of the 31 patients had all three EEGs in different phases and they contributed a pair to two different subgroups in Table 2. The number of pairs in the three 36-h subgroups in Table 2 is accordingly 34 (31 + 3 = 12 + 13 + 9 = 34). For the 72-h period, four patients did not contribute and 36 (40 − 4) patients contributed to paired analysis. The subgroup counts for the secondary 72-h cut-off in Table 2 are 13, 13 and 10 (13 + 13 + 10 = 36).

Table 2

Clinical data and EEG background data: mean (s.d.) in paired migraine subgroups

	36 h			72 h
Preictal (n = 12)	Attack (n = 13)	Postictal (n = 9)	Preictal (n = 13)	Attack (n = 13)	Postictal (n = 10)
Clinical data
Women/men	11/1	12/1	7/2	11/2	12/1	9/1
MA/MoA	10/2	10/3	8/1	3/10	2/11	2/8
Age (years)	35.2 (13.0)	37.5 (12.5)	41.3 (12.8)	33.8 (12.4)	36.8 (12.6)	39.5 (10.5)
Headache history (years)	19.9 (12.2)	20.5 (9.9)	18.1 (13.1)	18.8 (8.8)	20.5 (9.3)	22.6 (11.3)
Headache frequency (1–4)	2.2 (0.6)	1.8 (0.6)	2.0 (0.5)	2.0 (0.6)	2.0 (0.6)	1.9 (0.6)
Headache intensity (1–4)	2.3 (0.8)	2.3 (0.6)	2.2 (1.0)	2.7 (0.5)	2.5 (0.5)	2.6 (0.5)
Usual headache attack duration (h)	15.6 (23.4)	14.9 (17.3)	18.4 (30.5)	18.1 (22.1)	17.5 (17.7)	13.6 (21.6)
Headache days last 3 months	7.2 (5.7)	7.2 (4.7)	4.2 (2.3)	4.7 (2.8)	6.6 (3.2)	4.5 (3.0)
Phonophobia (0–2)	1.3 (0.7)	0.8 (0.8)	1.2 (0.8)	1.2 (0.7)	0.8 (0.7)	1.2 (0.9)
Photophobia (0–2)	1.4 (0.8)	1.0 (0.8)	1.2 (0.7)	1.6 (0.7)	1.2 (0.8)	1.5 (0.7)
EEG background data∗
Epochs (no)
Attack-related†	16.8 (1.6)	15.8 (2.1)	16.4 (1.6)	16.1 (1.6)	16.1 (2.1)	16.8 (1.5)
Baseline‡	17.3 (2.8)	16.1 (1.7)	17.0 (3.2)	15.8 (1.4)	16.2 (1.9)	16.4 (1.2)
Stage 1 (%)
Attack-related†	8.9 (13.9)	11.2 (20.2)	0.7 (2.2)	10.9 (15.9)	5.3 (10.3)	0.7 (2.1)
Baseline‡	4.6 (9.7)	8.3 (14.2)	6.4 (8.7)	4.3 (9.4)	9.0 (15.1)	2.6 (6.4)
Eye blinks (per min)
Attack-related†	4.3 (4.4)	1.8 (2.5)	1.7 (2.6)	1.8 (3.5)	1.7 (2.5)	1.3 (1.9)
Baseline‡	3.3 (5.4)	3.0 (4.8)	2.4 (4.6)	0.5 (1.1)	1.5 (3.3)	0.3 (0.5)

∗

Number of EEG epochs accepted in the QEEG analysis, percent drowsiness (these epochs were also excluded) in the 5-min recording and the number of eye blinks per minute.

†

Preictal, attack or postictal recording.

‡

Paired baseline (interictal) recording.

Preictal (YZ) subgroup: migraine patients who had EEG both interictally (> 36 or 72 h after/before previous/next attack) and within 36 (72) h before next attack. Attack (UZ) group had EEG both in the attack phase and in the interictal period. Postictal (XZ) group had EEG both in the interictal phase and within 36 (72) h after termination of the previous attack. With 36- and 72-h cut-off, respectively, nine and four patients were recorded in the same phase. The number of pairs is accordingly less than the total number of patients in Table 1. No significant differences.

Figure 1

Classification of recordings. The same procedure was also used with 72-h intervals.

As attacks occurred ‘randomly’ with respect to predetermined EEG dates, interictal recordings were as probable to occur before as after a preictal recording. The same arguments hold for postictal and attack recordings, eliminating the effects of a possible order bias in the paired analysis. When two of the three EEGs were in the same phase, recording on day B was preferred over days A and C. Day C was preferred over day A. The selection algorithm minimizes a possible first-day effect.

The EEG technicians, the neurophysiologist and other staff involved in data reduction and analysis were blinded regarding the diagnostic status. Written consent was obtained from all patients, and they received US $150 (not mentioned in the newspaper advertisement) after completing the three sessions to cover expenses. Technicians and doctors who performed recordings were blinded to the diagnosis. The other examinations in the 2.5 h long neurophysiological battery were thermal pain thresholds, VEP, BAEP and pupillary reflex (to be reported in another paper). The Regional Ethics Committee approved of the study.

EEG recording

A 5-min eyes closed EEG-videometry was performed, always at the same time of day, followed by photostimulation and motor tasks for 25 min. Data from the first 5 min of relaxed supine EEG are reported in the present paper. Digital EEG (Nervus 3.0 with M40 amplifier) was recorded with a common reference. Twenty electrodes were placed according to the 10/20 international system. Electrocardiography and eye movement detection channels were also applied. Electrode impedance was kept below 5 kΩ. To avoid drowsiness, the patients were asked to open and close their eyes every minute, and the EEG technician also alerted them if drowsiness occurred in between. A neurophysiologist interpreted the EEGs visually and quantified the amount of drowsiness as percent stage 1 during the first 5 min, while eye blinks (defined as a > 50 µV typical deflection in the Fp1-chin vertical eye movement or in the T1–T2 horizontal eye movement channel) were counted for the first minute. There were no differences in drowsiness or eye blinks between the pre-, post- and interictal intervals (Table 2).

QEEG analysis

The complete EEG file was imported into ASA (Advanced Source Analysis, version 3.1, ANT Software, Cambridge, UK) for further quantitative analysis. Segments with eye movements, other artefacts and periods of drowsiness were manually excluded. Average reference montage, band pass filters of 0.5–70 Hz and 50-Hz notch filter were used. Fast Fourier Transform was performed on 4-s epochs yielding a power spectrum with 0.25-Hz resolution. The time/amplitude series had a sampling frequency of 256 Hz. Band power values (µV²) were calculated by summing power across all bins in the 0.5–3.5 Hz (δ), 3.75–7.5 Hz (θ), 7.75–12.5 Hz (α) and 12.75–30 Hz (β) frequency spectrum. Occipitoparietal (O1, O2, P3, P4), temporal (T3, T4, T4, T6) and frontocentral (F3, F4, C3, C4) regional average values were computed. Asymmetry was calculated as the absolute difference between the left- and the right-sided regional power.

Statistics

Analysis was performed with SPSS (version 13.0; SPSS Inc., Chicago, IL, USA) and SYSTAT (version 11; Systat Software, San Jose, CA, USA). The preictal, ictal and postictal QEEGs were compared with a paired interictal record. Because there were rather few patients within each paired group and because the asymmetry data were skewed and difficult to transform, it was decided to use the non-parametric Wilcoxon signed ranks test within each subgroup. Effect size in paired analysis was calculated as 100% × mean difference/s.d._dif (s.d. of differences). The number of tests was reduced when predetermined regional variables were computed. Since δ, θ, α and β activity represent largely different neurophysiological processes, and since preictal EEG has seldom been studied in the past, our focus was upon regional bandpower and asymmetry. Hence, we performed a study with multiple predefined hypotheses. P-values were not adjusted for multiple comparisons because Bonferroni-corrected statistics test the often non-relevant hypothesis that all variables are simultaneously significant and because uncorrected P-values yield the best balance between type I and type II errors in a study with multiple relevant predetermined hypotheses (32, 33). Bonferroni correction yields large type II error rates (32, 33). Two-sided P-values < 0.05 are reported as significant and P-values < 0.10 as trends. A paired analysis with 12 (or nine) pairs has approximately 80% (or 65%) power to detect an effect = 90% of group s.d.

Results

In the 36 h before the attack, frontocentral δ power increased (74% of s.d._dif, Z = 2.12, P = 0.03) and frontocentral θ and α tended to increase (55% and 61% of s.d._dif, Z > 1.73, P < 0.09) compared with interictal baseline (Figs 2 and 3). (s.d._dif, has not been tabulated.) Temporal β power tended to increase (66% of s.d._dif, Z = 1.80, P = 0.07).

Figure 2

Frontocentral δ power is significantly higher (P = 0.03) and θ and α power tend to be higher (P ≤ 0.08) before attack compared with interictal baseline.

Figure 3

Individual line diagram. Nine of 12 patients had higher frontocentral (FC) δ power values within 36 h before attack compared with interictal baseline.

Occipitoparietal (α and θ) and temporal (α) power were more asymmetric before the attack compared with interictal baseline (56, 71 and 64% of s.d._dif, respectively, Z > 2.04, P < 0.04; Figs 4 and 5).

Figure 4

Occipitoparietal α and θ asymmetry is significantly higher (P ≤ 0.04) before attack compared with interictal baseline. Asymmetry is calculated as the absolute power difference between the left and right hemisphere: abs (left side power – right side power).

Figure 5

Individual line diagram. Nine of 12 patients had higher occipitoparietal (OP) α power asymmetry within 36 h before attack compared with interictal baseline. Asymmetry is calculated as the absolute power difference between the left and right hemisphere: abs (left side power – right side power).

No QEEG differences were found after the attack compared with interictal baseline. Posterior α power was increased during attack (74% of s.d._dif, Z = 2.62, P = 0.01), whereas frontocentral β power also was slightly increased (24% of s.d._dif, Z = 2.27, P = 0.02). Temporal β power asymmetry tended to be higher during the attack (51% of s.d._dif, Z = 1.64, P = 0.10).

When the pre- and postictal phases were redefined to start 72 h before and end 72 h after the attack, respectively, regional paired power differences and asymmetries were not found, apart from increased posterior α power during attack (64% of s.d._dif, Z = 2.83, P = 0.005).

Discussion

Our main findings in the present longitudinal blinded study were the increase in cortical frontocentral δ activity and posterior α asymmetry within the 36-h interval before the next migraine attack. Our results are in general concordance with several CNV studies, which have also found profoundly different cortical activity about 1 day before the attack (6, 8), and partly concordant with two previous EEG studies (7, 23).

Nyrke et al. (23) have compared QEEG values before and after attack with interictal baseline, but they did not report absolute power values due to ‘scanty findings’. Our results suggest that the lack of QEEG power differences in the study of Nyrke et al. (23) was caused by their use of a 72-h preictal interval. However, they found asymmetry in α power from 3 days before to 3–6 days after the attack. Asymmetry was predominantly observed in the preictal phase, in accordance with our 36-h preictal findings. Facchetti et al. (17) also found α asymmetry among 31 MA patients, three of whom were recorded within 24 h after the attack, but the time to the next attack was not reported.

Siniatchkin et al. (7) found that δ and θ power increased (in the right hemisphere), θ increased (in the left hemisphere), whereas α asymmetry increased with right side α predominance (7). However, they included patients from 1–4 days before the attack, and we could not find significant paired differences in our 72-h analysis. If most of their patients had the attack 1–2 days before the recording, it might explain the similarity to our 36-h results, but they did not report the proximity distribution. Another methodological difference is that they compared the preictal recordings with the postictal ones, not with an interictal baseline interval as we did. To our knowledge, our present study is the first to analyse and report QEEG findings this close to the migraine attack.

Interhemispheric asymmetry of the α rhythm, diffuse and focal slowing, and increase of fast activity have been reported with some consistency in the literature on QEEG in migraine (10, 11, 34). However, most authors have studied QEEG activity in migraine without considering the proximity to the previous and next attack (15, 17, 35) or the next attack (19, 21). Since we found increased asymmetry and increased δ power before the attack, it could possibly signify that a majority of EEGs included in earlier studies have been followed rather closely by a migraine attack. This seems plausible, since the circumstances around an EEG recording session may provoke an attack (stress, flicker stimulation, etc).

The increase–decrease pattern in our results resembles published CNV data, where the early components in amplitudes and dishabituation increase in the last 2 days before attack and decrease 1–2 days after attack before they normalize (6, 8). Based on the pattern similarities, it is tempting to speculate if the findings in EEG and CNV result from the same pathophysiological process.

CNV is an event-related preparatory potential, whereas EEG in this study was measured in relaxed wakefulness. The connection between EEG and CNV is not well explained. One possibility relates to arousal mechanisms mediated by cholinergic, noradrenergic and serotonergic activity in brainstem nuclei. It is known, for example, that reduced activity, lesions in the cholinergic brainstem and basal forebrain projection system, as well as anticholinergic scopolamine administration (36) may cause slow δ EEG activity (37).

Reduced activity in the cholinergic brainstem and basal forebrain projection system may also cause cortico-thalamic disconnection and increase θ and β in wakefulness (38). This is in concert with our present finding (i.e. increase in frontocentral θ and a trend towards increased temporal β power before the attack), suggesting a possible ‘thalamo-cortical dysrhythmia’ in migraine within 36 h before the attack (39). During thalamocortical dysrhythmia, protracted hyperpolarization of thalamic cells, either because of deafferentation or excess inhibition, increases low-frequency neuronal oscillations in the θ range. Thalamocortical activation, as measured by somatosensory high-frequency oscillations, does indeed seem to be reduced interictally in migraine (25). At the cortical level, low-frequency activation of intracortical inhibitory neurons can reduce lateral inhibitory drive and result in high frequency, phase-locked coherent activation of neighbouring cortical modules (39).

Adrenergic modulation from the nearby mesencephalic reticular formation has also been hypothesized to underlie the increased CNV amplitude in migraine (40). However, noradrenaline is generally assumed to exert an activating influence on EEG, increasing fast activity and diminishing slow potentials. We can accordingly not conclude that EEG and CNV changes share a common mechanism in the preictal period. It is possible that EEG reflects a basal (‘preactivation’) CNS state, which indeed may be reduced in migraine according to some studies (2, 3), whereas CNV reflects another dimension, i.e. excitability in preparatory systems (40, 41).

Several authors have suggested that migraine can be characterized by increased cortical arousal (40, 41) caused by dysfunction of serotonergic and catecholaminergic structures in the brainstem, especially in nucleus raphe and locus coerulus (40, 42). It was hypothesized, based on current theories (43) and published EEG data in migraine, that cyclic raphe and nucleus coreulus dysfunction causes ischaemia, EEG slowing and associated neurological dysfunction in basilar migraine and migraine with complicated aura (10). However, arousal seems to depend more on the widely distributed, but more specific ascending pontine and mesencephalic cholinergic system than on the noradrenergic system (37, 44–46). It may accordingly be more likely that EEG slowing in basilar migraine is related to transiently reduced neuronal activity in these ascending cholinergic pathways.

Basilar migraine with reduced consciousness, a type of aura that is associated with signs of bilateral brainstem dysfunction, is often associated with bilateral frontocentral δ activity (20, 47), i.e. abnormal bilateral paroxysmal δ activity (48), and this may represent one extreme end of the spectrum of brainstem dysfunction in migraine. The slight QEEG slowing that emerged in our study may represent the opposite extreme of this putative preictal brainstem dysfunction. It can be hypothesized that a slight depression or instability in the cholinergic activity of the basal forebrain and pontine tegmentum may trigger some low-amplitude slow cortical (< 1 Hz) oscillations and possibly also a few cortical and thalamic 1–4-Hz oscillations (49), even in the awake preictal migraine patient.

Slow wave power increases with drowsiness and during slow-wave sleep due to activation of pacemaker currents in thalamic reticular relé cells. Sleepiness is a well-known migraine prodrome (50), hence the preictal recordings could theoretically be influenced by this. However, we activated patients, excluded epochs with drowsiness, and found no evidence of sleep-related bias in the paired groups. We accordingly do not believe that drowsiness per se can explain the preictal QEEG change in the present study.

Ictal EEG changes such as occipital δ and reduced posterior α have been reported by some (16, 51, 52), but not all researchers (53, 54). None of our patients had basilar-type aura with or without reduced consciousness, a clinical picture which can be associated with profound EEG slowing (55, 56). In the present paired and blinded study, moderately increased occipitoparietal α activity was found. This is in contrast to Schoenen et al. (16), who reported unilateral α and θ reduction in common migraine attacks, possibly because the timing of EEG related to attack onset differed, as our patients were studied rather late in the attack phase. Moreover, we were not able to perform paired analysis within MA and MoA subgroups because too few MA patients were recruited.

In conclusion, frontocentral δ power increased, whereas frontocentral θ and α power tended to increase within 36 h before the next migraine attack compared with the interictal period. Occipitoparietal α and θ power and temporal α power were more asymmetric before the attack compared to the interictal baseline. Our results support the theory of a fluctuating cortical neuronal dysfunction in migraine that may reach a critical level within a few days before the attack. It is possible from available evidence in the literature, including positron emission tomography data (57, 58), that the cortical dysfunction follows or parallels subcortical cholinergic dysregulation in the mesencephalic reticular substance and basal forebrain. Thalamocortical dysfunction may also contribute to the observed preictal change in EEG activity. According to our results, is seems that migraineurs are most susceptible to attack when δ power and α and θ asymmetry levels are high. It should be emphasized that the magnitude of the effects are small and certainly not detectable at the individual level.

Footnotes

Acknowledgements

The authors are most grateful for the invaluable assistance of Marit Stjern, Grethe Helde, Knut Hagen, Lars Jacob Stovner and G⊘ril Bruvik Gravdal.

References

Schoenen

Neurophysiological features of the migrainous brain. Neurol Sci 2006; 27:S77–81.

Ambrosini

de Noordhout

Sandor

Schoenen

Electrophysiological studies in migraine: a comprehensive review of their interest and limitations. Cephalalgia 2003; 1:13–31.

Ambrosini

Schoenen

The electrophysiology of migraine. Curr Opin Neurol 2003; 114:327–31.

Welch

KM.

Contemporary concepts of migraine pathogenesis. Neurology 2003; 61:S2–8.

Kropp

Gerber

WD.

Contingent negative variation during migraine attack and interval: evidence for normalization of slow cortical potentials during the attack. Cephalalgia 1995; 15:123–8.

Kropp

Gerber

WD.

Prediction of migraine attacks using a slow cortical potential, the contingent negative variation. Neurosci Lett 1998; 257:73–6.

Siniatchkin

Gerber

Kropp

Vein

How the brain anticipates an attack: a study of neurophysiological periodicity in migraine. Funct Neurol 1999; 14:69–77.

Siniatchkin

Kropp

Gerber

Stephani

Migraine in childhood—are periodically occurring migraine attacks related to dynamic changes of cortical information processing?

Neurosci Lett 2000; 279:1–4.

Nuwer

MR.

Quantitative EEG: II. Frequency analysis and topographic mapping in clinical settings. J Clin Neurophysiol 1988; 5:45–85.

10.

Sand

EEG in migraine: a review of the literature. Funct Neurol 1991; 6:7–22.

11.

Sand

Electroencephalography in migraine: a review with focus on quantitative electroencephalography and the migraine vs. epilepsy relationship. Cephalalgia 2003; 1:5–11.

12.

Gronseth

Greenberg

MK.

The utility of the electroencephalogram in the evaluation of patients presenting with headache: a review of the literature. Neurology 1995; 45:1263–7.

13.

Jonkman

Lelieveld

MH.

EEG computer analysis in patients with migraine. Electroencephalogr Clin Neurophysiol 1981; 52:652–5.

14.

Polich

Ehlers

Dalessio

DJ.

Pattern-shift visual evoked responses and EEG in migraine. Headache 1986; 26:451–6.

15.

Drake

Jr Du Bois

Huber

Pakalnis

Denio

LS.

EEG spectral analysis and time domain descriptors in headache. Headache 1988; 28:201–3.

16.

Schoenen

Jamart

Delwaide

PJ.

Cartographie electroencephalographique dans les migraines en periodes critique et intercritique. Rev Electroencephalogr Neurophysiol Clin 1987; 17:289–99.

17.

Facchetti

Marsile

Faggi

Donati

Kokodoko

Poloni

Cerebral mapping in subjects suffering from migraine with aura. Cephalalgia 1990; 10:279–84.

18.

Lia

Carenini

Degioz

Bottachi

Computerized EEG analysis in migraine patients. Ital J Neurol Sci 1995; 16:249–54.

19.

Neufeld

Treves

Korczyn

AD.

EEG and topographic frequency analysis in common and classic migraine. Headache 1991; 31:232–6.

20.

Walser

Isler

Frontal intermittent rhythmic delta activity, impairment of consciousness and migraine. Headache 1982; 22:74–80.

21.

Genco

de Tommaso

Prudenzano

Savarese

Puca

FM.

EEG features in juvenile migraine: topographic analysis of spontaneous and visual evoked brain electrical activity: a comparison with adult migraine. Cephalalgia 1994; 14:41–6.

22.

Thomaides

Tagaris

Karageorgiou

EEG and topographic frequency analysis in migraine attack before and after sumatriptan infusion. Headache 1996; 36:111–14.

23.

Nyrke

Kangasniemi

Lang

Alpha rhythm in classical migraine (migraine with aura): abnormalities in the headache-free interval. Cephalalgia 1990; 30:177–81.

24.

Judit

Sandor

Schoenen

Habituation of visual and intensity dependence of auditory evoked cortical potentials tends to normalize just before and during the migraine attack. Cephalalgia 2000; 20:714–19.

25.

Coppola

Vandenheede

Di Clemente

Ambrosini

Fumal

De Pasqua

Schoenen

Somatosensory evoked high-frequency oscillations reflecting thalamo-cortical activity are decreased in migraine patients between attacks. Brain 2005; 128:98–103.

26.

Sand

Vingen

JV.

Visual, long-latency auditory and brainstem auditory evoked potentials in migraine: relation to pattern size, stimulus intensity, sound and light discomfort thresholds and preictal state. Cephalalgia 2000; 20:804–20.

27.

Evers

Quibeldey

Grotemeyer

Suhr

Husstedt

IW.

Dynamic changes of cognitive habituation and serotonin metabolism during the migraine interval. Cephalalgia 1999; 19:485–91.

28.

Goder

Fritzer

Kapsokalyvas

Kropp

Niederberger

Strenge

Polysomnographic findings in nights preceding a migraine attack. Cephalalgia 2001; 21:31–7.

29.

Fritzer

Strenge

Goder

Gerber

Aldenhoff

Changes in cortical dynamics in the preictal stage of a migraine attack. J Clin Neurophysiol 2004; 21:99–104.

30.

Bruni

Russo

Violani

Guidetti

Sleep and migraine: an actigraphic study. Cephalalgia 2004; 24:134–9.

31.

31 Headache Classification Committee of the International Headache Society. Classification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain. Cephalalgia 2004; 24 (Suppl. 1):1–160.

32.

Perneger

TV.

What's wrong with Bonferroni adjustments. BMJ 1998; 316:1236–8.

33.

Feise

RJ.

Do multiple outcome measures require p-value adjustment?

BMC Med Res Methodol 2002; 2:8.

34.

Schoenen

Clinical neurophysiology studies in headache: a review of data and pathophysiological hints. Funct Neurol 1992; 40:191–204.

35.

Hughes

Robbins

LD.

Brain mapping in migraine. Clin Electroencephalogr 1990; 21:14–24.

36.

Buzsaki

Bickford

Ponomareff

Thal

Mandel

Gage

FH.

Nucleus basalis and thalamic control of neocortical activity in the freely moving rat. J Neurosci 1988; 8:4007–26.

37.

Steriade

Gloor

Linas

Lopes da Silva

Mesulam

MM.

Report of IFCN Committee on Basic Mechanisms. Basic mechanisms of cerebral rhythmic activities. Electroencephalogr Clin Neurophysiol 1990; 76:481–508.

38.

Llinas

Steriade

Bursting of thalamic neurons and states of vigilance. J Neurophysiol 2006; 95:3297–308.

39.

Llinas

Ribary

Jeanmonod

Kronberg

Mitra

PP.

Thalamocortical dysrhythmia: a neurological and neuropsychiatric syndrome characterized by magnetoencephalography. Proc Natl Acad Sci USA 1999; 96:15222–7.

40.

Kropp

Gerber

WD.

Is increased amplitude of contingent negative variation in migraine due to cortical hyperactivity or to reduced habituation?

Cephalalgia 1993; 13:37–41.

41.

Böcker

Timsit-Berthier

Schoenen

Brunia

CH.

Contingent negative variation in migraine. Headache 1990; 30:604–9.

42.

Schoenen

Maertens de Noordhout

Timsit-Berthier

Timsit

Contingent negative variation and efficacy of beta-blocking agents in migraine. Cephalalgia 1986; 6:229–33.

43.

Lance

Lambert

Goadsby

Zagami

AS.

5-Hydroxytryptamine and its putative aetiological involvement in migraine. Cephalalgia 1989; 9 (Suppl. 9):7–13.

44.

Aston-Jones

Cohen

JD.

Adaptive gain and the role of the locus coeruleus–norepinephrine system in optimal performance. J Comp Neurol 2005; 493:99–110.

45.

McCormick

DA.

Neurotransmitter actions in the thalamus and cerebral cortex and their role in neuromodulation of thalamocortical activity. Prog Neurobiol 1992; 39:337–88.

46.

Lance

Adams

Lambert

GA.

Bulbo-cortical pathways and their possible relevance to migraine and epilepsy. Funct Neurol 1986; 1:357–61.

47.

Pietrini

Terzano

D’Andrea

Parrino

Cananzi

Ferro-Milone

Acute confusional migraine: clinical and electroencephalographic aspects. Cephalalgia 1987; 7:29–37.

48.

Schaul

Pathogenesis and significance of abnormal nonepileptiform rhythms in the EEG. J Clin Neurophysiol 1990; 7:229–48.

49.

Amzica

Steriade

Electrophysiological correlates of sleep delta waves. Electroencephalogr Clin Neurophysiol 1998; 107:69–83.

50.

Blau

JN.

Resolution of migraine attacks: sleep and the recovery phase. J Neurol Neurosurg Psychiatry 1982; 45:223–6.

51.

Engel

Ferris

Romano

Focal electroencephalographic changes during the scotomas of migraine. Am J Med Sci 1945; 209–650.

52.

Logar

Grabmair

Lechner

Das EEG bei Migrane. EEG EMG Z Elektroenzephalogr Elektromyogr Verwandte Geb 1986; 17:153–6.

53.

Dow

Whitty

CWM.

Electro-encephalographic changes in migraine. Lancet 1947; 2:52–4.

54.

Lauritzen

Trojaborg

Olesen

EEG during attacks of common and classical migraine. Cephalalgia 1981; 1:36.

55.

Parrino

Pietrini

Spaggiari

Terzano

MG.

Acute confusional migraine attacks resolved by sleep: lack of significant abnormalities in post-ictal polysomnograms. Cephalalgia 1986; 6:95–100.

56.

Gaji

Basilar artery migraine: EEG and evoked potential patterns during acute stage. Headache 1986; 26:220–3.

57.

Weiller

May

Limmroth

Jüptner

Kaube

Schayck

Brain stem activation in spontaneous human migraine attacks. Nat Med 1995; 1:658–60.

58.

Bahra

Matharu

Buchel

Frackowiak

Goadsby

PJ.

Brainstem activation specific to migraine headache. Lancet 2001; 357:1016–17.