Abstract
Evidence for deviant maturation of sensory processing in migraine has come recently from cross-sectional studies during childhood. Age-dependent development of response preparation and evaluation is characterized using a longitudinal design in school-aged migraine patients and controls in order to challenge the hypothesis of migraine as a maturation disorder. Forty-six children with migraine and 57 healthy controls aged 6-18 years were investigated and followed up 4 years later using a simple acoustic contingent negative variation (CNV) paradigm. Maturation in controls was characterized by increasing negativity of late and total CNV and stability of initial CNV (iCNV) and the motor postimperative negative variation (mPINV). Migraine patients showed a lack of development for late and total CNV and decreasing iCNV and mPINV negativity. This first longitudinal study confirms cross-sectional results of deviant CNV maturation in migraine. Altered maturation was not correlated with clinical improvement and may represent a vulnerability marker for migraine.
Introduction
Migraine is currently recognized as an information-processing disorder, and many results from electrophysiological studies support this pathophysiological concept (1). Lack of habituation is the principal and most reproducible interictal abnormality of sensory processing in migraine and is thought to be associated with the relatively consistent finding of more negative amplitudes of the contingent negative variation (CNV) in adult migraineurs (2). CNV is a slow cortical potential that corresponds to processes of response preparation and evaluation. CNV abnormalities may represent a possible biological elementary endophenotype for migraine: increased amplitudes of the early CNV component and—to a lesser extent—reduced habituation are greatly influenced by familial factors and seem therefore to reflect vulnerability to migraine rather than state markers of disease activity or manifestation (3).
Increased CNV negativity and disturbed habituation are consistently reported only in adult migraineurs. Before young adulthood, and especially in children, results are contradictory (4–9) because of maturational influences on cortical processing (5). Even healthy children present increased amplitudes and reduced habituation of the early CNV component compared with adults and thus resemble alterations in adult migraine patients (5).
However, some recent studies in children and adolescents have provided evidence for an altered maturation of information processing in migraine. Altered age-dependent development of CNV amplitude and habituation slope in migraine has recently been demonstrated by Kropp et al. (5) and by our group (6, 9). We also found deviant age-dependent development of the second negative component of the visual evoked potential (VEP) (10, 11); similar results have been reported for the amplitude/stimulus intensity function (ASF slope) of the auditory-evoked potential (AEP) (12). All recent studies investigating evoked (EPs) and event-related potentials (ERPs) (and maturation of the underlying processes) in children with migraine bear on cross-sectional designs.
Longitudinal studies are much more valid than cross-sectional ones for the assessment of maturation processes. They are much more sensitive to age-dependent changes because identical subjects are investigated at different time points (and not different groups of subjects suffering from migraine at different time points), and intraindividual development is addressed instead of group effects. Intraindividual comparisons have more statistical power and can substantially reduce variance. Additionally, intraindividual maturation of the different CNV components can be characterized for each single subject calculating maturational slopes. Thus mean maturation can be compared between migraine patients and controls, whereas cross-sectional designs can derive maturation only very indirectly by comparison of different groups of younger and older subjects. EP and ERP studies investigating maturation under normal and different illness conditions during childhood and adolescence are rare because they are complex and call for high numbers of subjects covering a large age range.
The CNV paradigm consists of a warning (S1) and a response stimulus (S2), and the elicited event-related slow brain potential is characterized by a characteristic augmenting negativity around the vertex between S1 and S2. Different components represent response preparation [initial CNV (iCNV), early component following S1; late CNV (LCNV), preceding S2; total CNV (TCNV)] and evaluation [motor postimperative negative variation (mPINV)] (13). These components have different topographies, are related to different behavioural processes and transmitter systems and are characterized by different maturational trajectories:
The iCNV represents both an unspecific orienting reaction (14) and early motor preparation as part of specific response anticipation (15–17) and is modulated by the noradrenergic system (15, 16). Pronounced iCNV negativity has been described in healthy children and is thought to reflect increased cortical excitability necessary for facilitation of information processing at an early developmental stage important for learning (primary school age) (4). Kropp et al. have demonstrated in a large cross-sectional study a marked decrease of iCNV negativity in healthy controls between 20 and 30 years of age. This iCNV decrease in early adulthood might indicate maturation of the frontal attention system (18) and, more general, of frontal cortex and its connections (19). However, subcortical structures are also thought to be involved in CNV maturation, such as basal ganglia, raphé nuclei, reticular formations and locus coeruleus. Source analysis of cross-sectional 64-channel high-resolution EEG data give evidence that higher negativity of iCNV at Cz in primary school children with migraine might be due to an additional ‘premature’ activation, e.g. in the locus coeruleus or the basal ganglia (9). Negativity of iCNV has been shown to correlate with the duration of migraine disease in adults and is therefore suggested as a marker of migraine chronicity (20, 21).
The LCNV reflects expectancy and sensory anticipation as well as advanced motor preparation and is controlled by the dopaminergic system (15, 22, 23). LCNV can be clearly separated from iCNV if the interstimulus interval (ISI) is >2 s (24). The age-dependent development is different from that of iCNV: negativity increases with age during childhood especially over motor areas (6, 25, 26) and declines afterwards during ageing (27). We have recently demonstrated that LCNV topography during and after puberty is different from prepubertal state; younger children seem to prepare for the fast response in a qualitatively different way than adolescents do—that might reflect maturation of motor cortex itself, motor system circuitry or prefrontal cortex (23). Cross-sectional results have shown a lack of this age-dependent negativity increase in children with migraine (6).
The amplitude of TCNV undergoes similar age-dependent development as LCNV (which is not surprising, because LCNV accounts for a substantial part of TCNV).
The mPINV is a postmovement component peaking in adults about 500 ms after the button press following S2. mPINV indicates postmovement evaluation in the contralateral pre/primary motor cortex as well as supplementary motor area (28). Normal age-dependent development during school age is characterized by slightly decreasing negativity (23). Elevated PINV negativity might be specific for children with migraine, as it has been described in migraine children <14 years old (7, 9), but not in adults. Negativity drop with age seems to be much more pronounced in children with migraine than in healthy controls (9).
The aim of the present investigation was to assess the intraindividual age-dependent development of CNV amplitudes in a longitudinal design at two time-points with an interval of 4 years in school-aged children with migraine in comparison with healthy controls and thus to challenge the hypothesis of migraine as a maturation disorder.
Methods
Subjects
One hundred and nine children and adolescents participated successfully in a two-point longitudinal assessment with first examination in 1999–2001 (T1) and follow-up 2003–2004 (T2): 57 headache-free controls and 52 patients with primary headache. Participants were 6–18 years old at T1; follow-up period was 3.9 ± 0.4 years (mean ± SD). Of 150 right-handed participants with analysable data at T1 (9), 109 (72%) could be recalled for T2.
At both T1 and T2, structured headache interviews with a child and a parent were performed (10). Headache was categorized according to the revised criteria of the International Headache Society (IHS) (29), including information of both T1 and T2. Patients classified as migraineurs without (MoA) or with aura (MA) fulfilled the IHS criteria (codes 1.1 and 1.2, respectively) at least at T1 or T2. Six patients suffered from tension-type headache (TTH) only; these were excluded from analysis because of the small group number. Controls (CO) were not permitted to have first-degree relatives suffering from migraine or chronic pain. For headache diagnosis and age as well as gender distribution of the remaining subjects, see Table 1. Subjects reporting both migraine and TTH were included in the respective migraine groups (10). A child-adapted prospective daily headache diary (30) was slightly modified and used to assess headache frequency over an 8-week period beginning at both T1 and T2. For clinical characteristics of migraine, see Table 2. Note that migraine frequency was reduced at follow-up, whereas the mean attack duration had increased. Duration of headache history at T2 was (mean ± SD) 111 ± 38 months, 113.8 ± 34.7 months in MoA and 105.9 ± 44.7 months in MA patients. No participant showed clinical hearing impairment, neurological or psychiatric diseases (other than primary headache) or took any drugs affecting the central nervous system. All subjects were headache free for at least 72 h before and after the recording as CNV amplitudes show periodic changes before and after a headache attack (31). The study was conducted according to the Declaration of Helsinki (current version, 1996) on biomedical research involving human subjects (Tokyo amendment) and the protocol was approved by the University of Heidelberg Ethics Review Committee. All children and their parents were instructed about the study and written informed consent was obtained.
Diagnoses (MoA, migraine without aura, MA, migraine with aura, CO, healthy controls), age at T1 and T2 (mean ± SD, minimum, maximum), gender distribution (M = male, F = female) of participants and number of subjects in the different age groups (6–11 years at T1; 12–18 years at T1)
Clinical characteristics of migraine (MoA, migraine without aura, MA, migraine with aura) at T1 and T2 (mean ± SD, minimum, maximum)∗
Differences of intensity, frequency, duration and analgesics intake were tested using repeated measurement ANOVA with the within-factor ‘T1T2’ and the between-factor ‘diagnosis’ (MoA vs. MA). Effects of the factor ‘diagnosis’ did not reach significance; level of significance for the factor ‘T1T2’ is given in the last column. Accompanying symptoms were tested for differences between T1 and T2 (level of significance given in the last column) using the Wilcoxon signed-rank test for paired data (separately for MoA and MA).
Experimental design, recordings and data preprocessing
At both T1 and T2, 60 CNV trials were recorded in three consecutive blocks of 20 trials each (short breaks were introduced to avoid exhaustion) using a warning stimulus S1 (frequency 1000 Hz, duration 50 ms, 90 dB) and an imperative stimulus S2 (frequency 2000 Hz, duration 50 ms, 90 dB). The interstimulus interval (ISI) between S1 and S2 was 3 s, intertrial intervals (ITI) varied randomly from 10 to 15 s (Gentask, Stim software package; Neuroscan Inc., Charlotte, NC, USA). Subjects were instructed to respond as fast as possible after S2 by pressing a mouse button with the dominant right hand. All participants were asked to fixate on a cross on a computer screen at 1 m distance during the recording session.
Synamp amplifiers (Neuroscan Inc.) were used at T1 and BrainAmp (Brain Products GmbH, Gilching, Germany) at T2 to record continuous DC EEG from 64 channels with a sampling rate of 250 Hz. Sixty-four sintered silver/silver chloride electrodes were fixed using equidistant electrode caps (Easycap; FMS, Munich, Germany) according to head size. Electrodes are named after the equivalent positions in the extended international 10–20 system. Small deviations are indicated by ‘. Vertical and horizontal electrooculogram (VEOG and HEOG) was also recorded from electrodes attached 1 cm above and below the left eye (VEOG) as well as next to the outer canthi (HEOG). Electrode impedances were kept below 5 kΩ. Reaction time and correctness of the response were measured using a trigger from the mouse.
Data were recorded DC to 70 Hz against a reference near Cz at T1 and to Iz at T2 and transformed offline to an average reference according to the recommendation of Nunez (32, 33). Recordings 1 s before S1 were taken as baseline. Only trials with correct responses within 1 s after S2 were included in analysis. The EEG signal was digitally filtered (30 Hz high cut-off), segmented into epochs of 7.5 s (from 1 s pre S1 to 3.5 s post S2), corrected automatically for DC drifts by subtraction of a linear function calculated by regression analysis over the whole segment (BrainVisionAnalyser; Brain Products GmbH, Gilching, Germany) and corrected for eye movements and blinks (34). Artefacts were rejected automatically if the signal amplitude exceeded 150 µV (because of higher background EEG amplitudes in children than in adults). This procedure was confirmed by visual inspection, and remaining smaller artefacts were removed. In the following we present the results of electrode Cz.
The iCNV was determined according to (35) as the mean amplitude 200 ms around the peak between 550 and 750 ms after S1. The mean amplitude of the last 200 ms preceding S2 served to measure the late component LCNV. TCNV was defined as the mean amplitude between S1 and S2 (35). mPINV was calculated as the mean voltage in the interval 500–1200 ms after S2 (28).
Data analysis
Clinical characteristics of migraine (intensity, frequency, duration and analgesics intake) were compared using repeated measurement
In a second step, intraindividual maturation was operationalized using slopes. For each subject, an individual maturation slope for each CNV component was computed using the linear equation CNV = intercept + slope × age (y = a + b × age). Amplitudes of CNV components at T1 and T2 and the respective ages in months were used to calculate maturation slope and intercept for every single subject. A linear equation of the form (y = a + b × age) or (y = a + b × 1/age) has been shown to be a good approximation of curvilinear developmental trajectories according to the general developmental model (y = a + age × b + 1/age × b + e; a = constant, b = multiple regression coefficient, e = error (36)) of CNV components during school age (23).
In order to test if (y = a + b × age) or (y = a + b × 1/age) would be more appropriate for the given age span in our study, correlation (r
2 and P-value) was calculated between individual slopes of CNV amplitudes derived from the formula (y = a + b × age) and age at T1. Additionally, possible differences between younger (6–11 years at T1) and older (12–18 years at T1) age group were tested using a one-way
To investigate differences between before and after puberty, analyses of variance for amplitude changes over time were rerun with ‘age group’ as an additional factor (age at T1 = 12 years. vs. age at T1 > 12 years) for both amplitudes at T1 and T2 and maturation slopes.
Correlation (r 2 and P-value) (i) between extent of amplitude changes and the amplitude at T1; and (ii) between linear maturation slopes and clinical improvement (frequency reduction at T2) as well as duration of migraine disease was calculated for all components.
The α-level was set at 0.05.
Results
Grand means of the CNV recordings are presented in Fig. 1. Note that migraine patients showed almost no change of iCNV and LCNV amplitude between T1 and T2, whereas in controls negativity increased with age. Negativity of mPINV amplitude decreased with age in migraine patients and was nearly stable in controls.
Grand means of the contingent negative variation (CNV) recordings at T1 (dotted line) and T2 (solid line) in migraine patients (left, n = 46) and controls (right, n = 57). Initial contingent negative variation (iCNV), late contingent negative variation (LCNV) and motor postimperative negative variation (mPINV) are indicated, total contingent negative variation (TCNV) is calculated as the mean amplitude between S1 and S2.
For amplitudes (μV) mean ± SEM and 95% confidence intervals (in parentheses) are given.
F- and P-values of the MANOVA are given for the factors ‘diagnosis’, ‘T1T2’ and their interaction.
Separate cross-sectional comparisons between migraineurs and controls at T1 and T2 showed significant differences for TCNV at T1 (
F = 4.395, P = 0.039
) and LCNV at T2 (
F = 5.068, P = 0.027
). No significant cross-sectional differences were found for the other components.
iCNV, Initial contingent negative variation; LCNV, late contingent negative variation; TCNV, total contingent negative variation; mPINV, motor postimperative negative variation.
Correlation between extent of amplitude changes and baseline amplitudes (T1) for all CNV components
R 2 and P-value are given for the total group, for the migraine group and the controls. For all contingent negative variation (CNV) components, extent of amplitude changes was clearly correlated to the amplitude at T1. For all CNV components, correlation was stronger in the control group (higher correlation coefficients).
iCNV, Initial contingent negative variation; LCNV, late contingent negative variation; TCNV, total contingent negative variation; mPINV, motor postimperative negative variation.
Cross-sectional comparisons showed that migraine children had significantly more negative TCNV amplitudes at T1 and less negative LCNV amplitudes at T2 than controls (see Table 3).
Individual linear maturation slopes differed significantly between migraine patients and controls for amplitudes of LCNV, TCNV and mPINV; and showed again a statistical trend for iCNV (Table 5). Note that CIs indicate that for iCNV and mPINV, >95% of the migraine patients had a positive slope (that means negative amplitude decreased between T1 and T2), whereas slopes of controls scattered about zero. For TCNV and LCNV we found a different pattern, with mostly positive slopes in migraine patients and >95% negative slopes in controls indicating an increasing negativity with age in controls.
Individual linear maturation slopes∗ (µV/year) of the amplitudes of CNV components in migraine patients and healthy controls
Note that positive slopes indicate decreasing negative contingent negative variation (CNV) amplitudes; negative slopes indicate an increase in negativity.
Dev., Direction of age-dependent development between T1 and T2; ↑, significantly decreasing negativity; ↓, increasing negativity; ≅, non-significant development; iCNV, initial contingent negative variation; LCNV, late contingent negative variation; TCNV, total contingent negative variation; mPINV, motor postimperative negative variation.
Significant development was assumed when the confidence intervals for the slope did not include zero.
Intercepts of the y-axis calculated from the linear equation were more negative in migraine patients for all components and differed between groups (Table 6).
Individual intercepts of y-axis (µV) calculated from the linear equation of the amplitudes of contingent negative variation components in migraine patients and healthy controls
iCNV, Initial contingent negative variation; LCNV, late contingent negative variation; TCNV, total contingent negative variation; mPINV, motor postimperative negative variation.
Maturation slopes of CNV amplitudes did not correlate significantly with age for the total group and were not different between the younger and older age group. These findings provide evidence that linear developmental trajectories are a sufficient model for the data. When we restricted analyses on development of LCNV and TCNV in healthy controls that showed increasing negativity with age, results were similar: no correlation of slope and age at T1 for TCNV (r 2 = 0.02, P = 0.26) and LCNV (r 2 = 0.02, P = 0.25), no effect of ‘age group’ on slopes for TCNV (F = 0.209, P = 0.649) and LCNV (F = 0.298, P = 0.588).
Changes of CNV parameters over time were not correlated to clinical improvement or changes of clinical characteristics: R 2 was <0.07 for correlation of iCNV, LCNV, TCNV and PINV maturation slopes with frequency reduction at T2 (P > 0.150); r 2 was <0.04 for correlation of iCNV, LCNV, TCNV and PINV maturation slopes with change of attack duration (P > 0.260). Maturation slopes were not correlated to duration of migraine disease (r 2 < 0.02 and P > 0.390 for all calculated slopes).
Additional analyses of variance including the factor ‘age group’ and its interactions with ‘diagnosis’ and ‘T1T2’ revealed significantly higher mPINV amplitudes in the younger age group (F = 6.235, P = 0.014). Effect of ‘age group’ was non-significant for the other components and the maturation slopes of all components. Double interactions of ‘age group’ and ‘diagnosis’ as well as triple interactions of ‘age group’, ‘diagnosis’ and ‘T1T2’ were non-significant for all amplitudes and maturation slopes.
Mean developmental lines according to the linear equation with the mean slope and intercept of either the migraine or the control group fit in with the observed values for different CNV amplitudes (Fig. 2), especially for LCNV and TCNV. For the latter, developmental differences between migraine and control were most pronounced and highly significant.
Mean developmental lines for age-dependent development of late contingent negative variation (LCNV) (top), total contingent negative variation (TCNV) (middle) and motor postimperative negative variation (mPINV) (bottom) in migraineurs (solid line) and controls (dotted line) in the age range 12–18 years. Lines represent the linear equation CNV amplitude = mean group slope × age + mean group intercept. For comparison with observed results and investigation of validity, mean amplitudes and SEM are given for age spans of 3 years in the range 6–23 years for migraineurs (grey symbols) and controls (white symbols) at T1 (squares) and T2 (circles). Number of subjects is given for all data points.
Discussion
For the first time, we have confirmed in a longitudinal design the hypothesis of migraine as a maturation disorder that was suggested by some cross-sectional studies investigating CNV (5, 6, 9) or visual or auditory-evoked potentials (10–12). Two consecutive investigations (with a follow-up of 4 years) have demonstrated altered age-dependent development of CNV components, suggesting deviant maturation of response preparation and evaluation in a large sample of children and adolescents with migraine compared with healthy controls.
The main finding is that migraineurs lack the age-dependent negativity increase of the late CNV component (and, respectively, the total CNV) during school age which is characteristic of healthy subjects. Previous cross-sectional data (6, 9) show identical findings.
For the early component iCNV, a statistical trend was found suggesting deviant development in migraineurs: negativity decreased with age in migraine patients during school age, whereas controls showed nearly stable amplitudes (see Tables 3 and 5). Kropp et al. (5) have demonstrated in a cross-sectional sample a reduction of iCNV negativity from 15 to 19 years onwards to young adulthood in healthy controls compared with relatively stable amplitudes in migraineurs. Using a longitudinal design with a higher power to characterize maturation trajectories, we have been able to show that a negativity drop of iCNV amplitude occurs also in migraine patients, but earlier than in controls, i.e. before age 23, most probably between 10 and 20 years of age. The earlier drop might be less pronounced, but there is no real ‘lack’ of this developmental step as suggested by the cross-sectional data of (5).
The negativity drop of the mPINV with age is much more pronounced in migraine than in healthy control children—these results from our cross-sectional data (9) are now fully supported by the longitudinal design.
Furthermore, our longitudinal results can integrate partly contradictory results from cross-sectional group comparisons during childhood and adolescence. Differences between migraine patients and healthy controls are not stable and seem to depend on age. During primary school age, migraine patients differ from healthy controls in cross-sectional designs and tend to present higher amplitudes of all CNV components (especially for iCNV (4, 9) and mPINV (7, 9), less stable for LCNV (5, 6, 12, 14) and TCNV (7, 8)). The observation of elevated negativity of CNV amplitudes in younger children is supported by the results of y-axis intercepts, calculated from the linear maturation equations (see Table 6). The y-axis intercept is the extrapolated value for the respective CNV amplitudes at an assumed age of 0 years—under the presumption of linear development. Despite the fact that development is surely not linear in the complete age range from 0 to 18 years (but can be sufficiently explained by linear models from 6 years onwards, see below), the more negative y-axis intercepts point towards a higher negative level of amplitudes in young children with migraine.
From about 30 years of age on, migraineurs have again more negative amplitudes, especially for iCNV (for a review see (2)), less stable for LCNV (5, 35) and TCNV (37, 38). In the presented study, cross-sectional comparisons between migraine patients and controls have revealed weak statistical differences, i.e. for migraineurs more negative TCNV amplitudes at T1 and less negative LCNV amplitudes at T2 than controls (see Table 3). Note that comparisons at T1 and T2 were done in an explorative way without Bonferroni correction, which points towards weakness of the statistical differences. Application of Bonferroni correction would result in lack of significance for the reported differences.
As our results show, between about 12 and 30 years of age, group differences between migraineurs and healthy controls are not reliably recorded in cross-sectional designs, because developmental trajectories of migraineurs and controls converge during this age range. The different development in children with migraine during school age makes the group differences of early school age disappear.
We covered an age range of 6–23 years with analysable longitudinal data for a total of 103 subjects (46 migraineurs, 57 controls). The high number of subjects supports the validity of results. Maturation slopes confirm the results from longitudinally calculated amplitude changes. Longitudinal results (amplitude changes and maturation slopes) fit in with our cross-sectional data at both T1 and T2, and cross-sectional data at T1 and T2 do not show systematic differences within small age ranges (see Fig. 2). That makes it improbable that development between T1 and T2 represents merely a regression to the mean.
However, the extent of amplitude changes was clearly correlated to the baseline amplitude (T1) for all CNV components. That means the more negative amplitudes at T1, the greater the reduction during follow-up time. The respective correlations were stronger for controls than for migraine patients (see Table 4)—this also argues against a mere regression to the mean.
Thus, the main impact of this longitudinal study is confirmation of the hypothesis that the development of special cerebral functions in migraineurs differs from healthy controls.
Background and impact of these findings such as structural correlates, functional consequences and implications for migraine pathophysiology are presently not clear and remain to be further elucidated. Nevertheless, we will present some explanatory remarks which remain partly speculative:
Definition and operationalization of maturation
Maturation means the gradual change from a simple to a more complex level of organization. Evoked and event-related potentials monitor specific cerebral functions with a close relationship to neuronal structures. Maturation of cerebral functions is reflected by change of latency and amplitude of main components with age. Neural activity and behaviour are interrelated with genetic activity and environmental factors by bidirectional influences (39). Maturation of cerebral structures and functions is thus interdependent with many other internal (e.g. other somatic factors, such as hormones, puberty state) factors and linked to the environment, e.g. by learning processes.
In this study, we have tried to rule out the influence of skull growth as one somatic factor as far as possible by transforming of CNV recordings to average reference. This technique is recommended when recording systems with 32 and more electrodes are used and allows valid investigation of age-dependent development under Cz, whereas mastoid-referenced recordings are more susceptible to artefacts and changes concerning the mastoid regions. Especially the influence of increasing skull thickness and bone density during childhood (40) can be minimized using an average reference technique, because changes are cancelling themselves in a mathematical way.
In the following the term ‘maturation’ is used for the age-dependent development of CNV, including the interdependence with other somatic as well as environmental factors and learning processes.
Possible correlates of deviant maturation in migraine
In migraine patients, the pattern of age-dependent development of CNV components differs from that in healthy controls. For LCNV and TCNV, no significant maturation can be recorded in migraine during school age, whereas healthy controls present increasing negativity. For iCNV and PINV, a statistical trend indicates a qualitatively different pattern: we found a decrease of negativity with age in migraineurs in contrast to nearly stable amplitudes over time in controls.
LCNV and TCNV
Possible explanations for the lack of maturation in migraine during school age are as follows: (i) maturation in migraineurs occurs before school age, i.e. earlier than in controls (premature maturation); (ii) maturation in migraineurs occurs after school age, i.e. later than in controls (delayed maturation); and (iii) maturation is completely lacking in migraineurs.
The integration of our results with those of previous studies allows the following plausible explanation: LCNV and TCNV negativity increased earlier in migraine patients than in controls, i.e. before the early school age (statistical trends for higher negative LCNV and TCNV amplitudes in younger children (6, 7)). However, this ‘premature negativity’ occurs only at an intermediate level and then seems to stagnate. Thus during school age, development in healthy controls overtakes development in migraineurs—with the result of slightly higher negative amplitudes in controls (see cross-sectional comparisons in Table 3).
Possible correlates of the deviant maturation can only be characterized if the generators of CNV are known. These have been described previously: supplementary motor area (SMA) and anterior cingulate become activated early during CNV and, together with the prefrontal cortex, recruit during late CNV those motor and sensory areas needed for a fast response after the imperative stimulus (15). Additionally, subcortical structures such as locus coeruleus or the basal ganglia are engaged. The involvement of the contralateral sensimotor cortex in LCNV generation seems to be a maturation step during puberty (23). Source analyses of the longitudinal data are necessary to characterize the cerebral substrates of the premature and subsequently stagnating development in migraine. Analysis of our cross-sectional data made ‘premature’ activation, e.g. in the locus coeruleus (leading to diffuse cortical activation summing up to a maximum over the vertex) or the basal ganglia (interacting with SMA) more likely to explain the rather stereotyped CNV elevation around the vertex in young children with migraine than involvement of the cortical systems responsible for orienting, motor preparation or sensory attention (9). The locus coeruleus may be involved in pubertal developmental spurts, as it has been shown under the control of sex hormones in rats (41).
iCNV
For iCNV, the negativity drop with age might be less pronounced in migraine patients than in controls and seems to occur at younger age. The explanation of an earlier maturation of the frontal attention system in migraine that is proposed to be associated with the iCNV decrease in early adulthood (18) is rather speculative at present and has to be challenged by functional investigations of the frontal attention system during school age in migraine. Another possibility is ‘premature’ subcortical activation in migraine that might be attenuated during puberty and probably returns after the age of 20 years in case of persistent (‘chronic’) migraine. Elevated iCNV amplitudes have been described as markers for chronicity (20, 21). Keeping this in mind, increased iCNV negativity in primary school children seems to reflect qualitatively other mechanisms than increased negativity in adult migraineurs.
Implications for migraine
Altered age-dependent development was not correlated to either clinical improvement or duration of the disease; this indicates a probability that altered maturation reflects unspecific migraine predisposition rather than the activity of the disease or chronification processes. Additionally, no differences were found between MA and MoA patients—this argues in favour of a quite unspecific marker of migraine disposition. Migraine has an important genetic component which seems to be higher in MA than in MoA; but the disease is genetically complex and both multiple genes and environmental factors play a role (42). Whether MoA and MA are distinct entities is still a matter of debate: Clinical characteristics have been proposed to provide evidence of different entities (43), but up to one-third of MA patients suffer from MoA attacks also (44). In our sample, most MA patients suffered from both types of attacks. Recently, even convergence of pathophysiological concepts has occurred: it is widely accepted that a phenomenon called cortical spreading depression (CSD) underlies the migraine aura (45). This may be true also in MoA: recent data point to the possibility of a clinically silent CSD (for a review see (46)).
In genetically complex diseases, gene action and disease manifestation may be linked by vulnerability markers or biological elementary endophenotypes (3). These are partial factors closer to gene action, e.g. abnormalities in information processing and neurophysiological features that may contribute to the determination of a migraine threshold which is modulated by internal (e.g. hormonal) and environmental factors (47). The habituation deficit in adult migraineurs has recently been suggested to represent a trait marker for the genetic predisposition to migraine (48). Similarly, altered maturation of response preparation and evaluation might reflect a vulnerability marker for migraine. Further investigations in healthy relatives of migraine patients can show if the altered maturation can be used as an endophenotype, i.e. a biological trait marker that is genetically determined. Altered maturation could be useful as a trait marker especially during childhood and adolescence. During this age span, the habituation deficit characteristic of adult migraine can not be detected because of the influence of sensory maturation on cortical information processing (3, 5). The maturation slopes may be particularly suitable for further studies investigating migraine predisposition and the association with both genetic factors and behaviour (e.g. executive functions). Slopes can be calculated intraindividually based on at least two investigations at intervals of a few years and are relatively independent of age and the length of follow-up period. Thus reference values may be easily generated. We have been able to show that slopes allow a rather good separation of diagnostic groups: the 95% CIs for TCNV and LCNV amplitude do not overlap, which means 95% of controls show clearly negative slopes, whereas most migraine patients show no or positive slopes (see Table 5).
During school age, linear trajectories seem to sufficiently model our data—keeping in mind that this is a simplification and approximation of a definite age span. However, linear slopes may be especially useful in the age range 10–18 years. Maturation seems to be not substantially age-dependent in the investigated age range (no significant interaction of ‘age group’ and ‘T1T2’, no correlation between age and maturation slopes). Therefore the assessment of maturation in the examined age range without building of age subgroups is reasonable. Younger children in our study did not show steeper slopes than older ones (either in the total group or in the control group presenting ‘physiological’ maturation), which would be the case if development occurred following a curvilinear trajectory according to the equation (y = a + 1/age × b).
In summary, our data show that during puberty children with migraine differ from controls in their age-dependent development of response preparation and evaluation. Maturation in controls is characterized by an increase in LCNV and TCNV amplitude and stability of iCNV and mPINV amplitude during school age. Migraine patients show a different pattern, with a lack of development for LCNV and TCNV, a ‘premature’ negativity drop for iCNV amplitude and a more pronounced negativity decrease of mPINV. Cross-sectionally, migraine patients show more negative CNV amplitudes during primary school age and from about 30 years of age on, but not between the ages of 12 and 30 years. During this period of puberty and adolescence, converging developmental trajectories of migraineurs and controls result in similar extents of amplitude. Altered maturation reflects migraine predisposition rather than the activity of the disease and can be validly investigated longitudinally even in single subjects. Single or cross-sectional investigation of CNV in migraine between the ages of 12 and 30 years, however, is not promising to reveal deviant results. Maturation slopes may be used as trait markers for migraine; further research including unaffected relatives should show if slopes are useful as an endophenotype of migraine.
Footnotes
Acknowledgements
This work was supported by Deutsche Forschungsgemeinschaft (OE 265/1-1, 2). The authors thank Kerstin Herwig for helping to acquire the data, for preprocessing of the EEG recordings, including removing of artefacts, and for very helpful discussion.