Abstract
Increased negativity of contingent negative variation (CNV) in adult migraineurs is thought to reflect cortical hyperexcitability. CNV amplitude changes with age in healthy adults. Recently, evidence emerged that this might not be the case for migraineurs. Our study investigates age-dependency of CNV during childhood age. Seventy-six healthy controls and 61 children with migraine without aura (IHS code 1.1) between 6 and 18 years were examined using an acoustic S1-S2-CNV-paradigm with a 3-s inter-stimulus interval. The amplitude of the late component of CNV, as well as total CNV at the vertex (Cz according to the international 10-20 system), were significantly higher in migraineurs without aura than in controls. Healthy controls showed increasing amplitudes of CNV with age, whereas in migraine children without aura amplitudes did not change. Thus group differences were reduced during adolescence. Increased CNV negativity might reflect a biological vulnerability to migraine, rather than being a result of chronification. Migraineurs seem to lack age-dependent development of CNV also during early age, which supports the hypothesis of migraine as a maturation disorder.
Introduction
Increased negativity of contingent negative variation (CNV) amplitude in adult migraineurs without aura is well established and is thought to reflect cortical hyperexcitability. The late component (late CNV: lCNV) (1), as well as the early component (initial CNV: iCNV) (2, 3), has been found to be increased. Differences in iCNV amplitude found in adults might be due to a reduction of amplitude with increasing age in healthy controls (CO) but not migraineurs without aura (MO) (3). Also, evidence for an impaired maturation of sensory information processing in migraineurs is emerging (4). Therefore, understanding the age-dependent development of CNV components during childhood and adolescence in MO compared with the physiological development in CO is of great interest to migraine pathophysiology. Some pioneering research on the topic has already been carried out.
One study (5) showed total CNV (tCNV) to be augmented (8–14 years) without distinguishing between CNV components. A recent study (mean age 13.6 years) using a Go-NoGo paradigm attributed augmented CNV mainly to changes in iCNV (2). However, the influence of age wasn't investigated due to limited age ranges, and healthy siblings formed part of the control groups in both studies. In the only study with a larger number of subjects that investigated the influence of maturation on CNV, Kropp et al. (3), using a long inter-stimulus interval (ISI) of 6 s in a Go-NoGo paradigma (focusing on a clear separation of iCNV from lCNV), described differences in iCNV development between MO and CO regarding the age groups 15–19 years and 20–29 years (as mentioned above) but not between 8–14 years and 15–19 years. No differences were stated for lCNV and thus no detailed data concerning age groups were given. However, in CO an increase of CNV amplitude during childhood and adolescence has been found by various investigators (6–8).
The aim of the current study was to provide more detailed data regarding age-related effects on both CNV components in children (CO and MO) down to 6 years, using a 3-s ISI – long enough to distinguish between CNV components and not too long to be applied to children – in a simple CNV paradigm (without response selection). We intended to examine whether MO differ from CO in CNV amplitude during the headache-free interval and, if so, to which component these differences can be attributed. Special emphasis was put on the contribution of age-dependent development to possible differences.
Methods
Sixty-one children suffering from migraine without aura (MO, IHS classification code 1.1, mean age 10.3 ± 2.89 (standard deviation) years, 35 male, 26 female) and 76 healthy controls (CO, 11.1 ± 3.32 years, 43 male, 33 female) aged from 6 to 18 years were included following these criteria: MO disease duration > 1 years and at least three attacks within the last 3 months; CO were not permitted to have any first-degree relatives suffering from migraine or any neurological disorder. One subject was not included because of hearing impairment. Newspaper announcements were used to recruit 50.8% of MO subjects, and 49.2% of the subjects were patients presenting at the neuropaediatric clinic with headache as their primary complaint. Controls were also recruited by newspaper announcements, as well as by messages on the hospital intranet.
The clinical characteristics of the MO sample were (values±standard deviation): migraine intensity (numerical rating scale from 1 to 10) 7.3 ± 1.7, migraine duration 7.7 ± 11.6 h, duration of disease 4.5 ± 2.8 years, migraine frequency: once a month 47.5%, twice a month 8.2%, weekly 34.4%, more often 9.8%. In agreement with other studies (e.g. 9), about one-third of MO subjects suffered from clinically relevant (T > 64) behavioural and emotional problems (mainly internalizing disorders) (10), as dimensionally assessed by CBCL (child behaviour checklist, Achenbach) (11). Patients continued to take acute medication but did not use any pharmacological prophylactic treatment except for magnesium (one subject) and Petadolex (one subject). Five patients received acupuncture and four received psychological treatment (bioresonance, psychotherapy, headache group, eurhythmia).
EEG-recordings were taken in the headache-free interval with at least 72 h distance to the last and the next attack. Five children had to be excluded from analysis because of headache attacks after the recording (a headache diary was kept). We recorded 20 CNV trials using a warning stimulus S1, 1000 Hz, duration 50 ms, 90 dB, and an imperative stimulus S2, 2000 Hz, 50 ms, 90 dB. Subjects were instructed to press a button as rapidly as possible when S2 ocurred. The ISI was 3 s, the inter-trial interval varied randomly from 10 s to 15 s. Neuroscan Synamp amplifiers were used to record continuous DC EEG from Cz (Fz and Pz – electrode locations according to the international 10–20 system – served to distinguish CNV from artifacts considering known topography of CNV) at an AD-rate of 250 Hz. Electrode impedance was kept below 5 kΩ. Vertical and horizontal electro oculargram were also recorded. Linked mastoids served as reference channels and recordings 1 s before S1 as baseline.
Only trials with correct responses within 1 s were included in further analysis (response window, Neuroscan Stim). Reaction time was recorded using a trigger from the mouse. The EEG signal was segmented into epochs of 7.5 s (1 s before S1 until 3.5 s past S2), digitally filtered (30 Hz high cut-off), and corrected for DC-drifts (linear function, Brain Vision Analyser) and also for eye movements and blinks (algorithm described by Gratton and Coles, 12). Artifacts were rejected automatically if the signal amplitude exceeded 150 µV. This procedure was confirmed by visual inspection; remaining artifacts were removed. Six subjects of 132 had to be excluded because they did not respond properly or produced unremovable artifacts. The amplitude of iCNV was calculated as the mean amplitude 200 ms around the peak within a latency range from 550 to 750 ms after S1. The mean amplitude of the last 200 ms preceding S2 served to measure lCNV. tCNV was the mean amplitude between S1 and S2 (1–3).
Statistical analysis was carried out using Stata 7 (Stata Corporation College Station, TX, USA). In a first step the effect of diagnosis was examined by t-tests for iCNV, lCNV and tCNV. In a second step multivariate regression analysis procedures were calculated to evaluate the influence of diagnosis, age, sex and the interaction between diagnosis and age, in order to detect group differences in age-dependent development. The formation of age groups was avoided to show continuous development. The α-level was set to 0.05.
Results
Group differences in amplitude
MO showed significantly higher amplitudes of lCNV and tCNV than CO (P = 0.053 and 0.021, t-test, Table 1). These differences were mainly found due to differences at early age (see Fig. 1 and age dependency below). The early component (iCNV) showed more variability due to latency differences and a tendency to increase already at a younger age in controls; no significant differences were found between groups (P = 0.389).
Mean amplitudes and age slopes. Mean amplitude (±standard error of the mean) and age slope (coefficient and significancy level of the regression) of iCNV, lCNV and tCNV are presented for healthy controls (CO) and children with migraine without aura (MO). Significancy levels of the differences in amplitude (P t-test, third column) and age slope (diff., linear regression, last column) between the two groups are also given
Age development of lCNV. Scatter plots showing age-dependent development and variability of lCNV in CO (healthy controls, left) and MO (children with migraine without aura, right). CO show a significant increase of lCNV amplitude with age, whereas MO do not (for values of regression analysis see Table 1). The result for total CNV is similar.
Reaction time
In both groups reaction time (RT) decreased significantly with age (CO coef=−0.02, P < 0.001 and MO coef=−0.02, P < 0.001, with coef being the age slope [s/year]). There were no significant differences between CO and MO children (P = 0.988).
Age dependency
Linear regression showed significant age dependency (Table 1) of tCNV (age slope coef [µV/year]=− 0.45, P = 0.004), lCNV (coef=−0.90, P < 0.001) and iCNV (coef=−0.63, P = 0.040) for CO at Cz. MO children did not show this development: tCNV coef=0.04 (P = 0.83), lCNV coef=0.07 (P = 0.79) and iCNV coef=−0.13 (P = 0.75). The age slopes of CO and MO differed significantly between groups for tCNV (P = 0.049) and lCNV (P = 0.006) but not for iCNV (P = 0.305).
Discussion
Like adults, young MO children also show increased late and total CNV amplitudes with respect to CO. Thus differences are not likely to result from long
The differences we found during the headache-free interval were not the result of changes in the early component iCNV but of the late component lCNV. Our findings, however, do not contradict the hypothesis that iCNV is increased right before the attack; a periodicity of iCNV amplitude has recently been shown in migraineurs (14). Methodological differences could play an important role concerning which CNV component (if any) is found to be increased. A Go-NoGo-paradigm (requiring response selection) might relatively enhance iCNV (2, 3), a simple CNV-paradigm (motor preparation) lCNV amplitude (1). A long ISI of 6 s (3) could activate different processes than shorter ones (15). The exact age of the population under investigation is a third important factor. Finally, we have to state that the differences in lCNV in the present study were not produced by the subjects under psychological treatment or receiving acupuncture; these subjects fitted into the general pattern.
The main finding of our study is the missing age-dependent development of lCNV in MO, including an elevation at an early state of development. Altered maturation of preparation in the sensory-motor system could be an explanation. Müller et al. (16) have found a negative Bereitschaftspotential (BP, preparation to self-paced movements), which is closely related to lCNV (preparation to externally triggered movements), in migraine children, whereas young CO showed positive waveforms. The positive BP (or CNV) (17) in young CO has been attributed to structural immaturity of the cerebral cortex (17) – inhibition by axodendritic synapses instead of axosomatic ones – or to inhibition of spontaneous motor activity before the movement (18). MO might also fail to show the development in BP or lCNV because of a cortical hyperexcitability producing a ceiling effect: MO could differ from CO in these respects.
A widely accepted model for the generation of CNV (for a review see 19) states that prefrontal cortex and reticular formation (FR) control a thalamic gating system that regulates cortical activation. An important feedback loop projects back from the cortex to the thalamus via basal ganglia (BG). BG execute GABAergic inhibitory control that can itself be inhibited by dopaminergic neurones. Segawa et al. (20) described that symptoms involving the ascendent efferents of the BG to the thalamocortical pathways are not observed during the first decade but appear later in life, providing evidence that the development of this cortical activation regulation system is going on during childhood. MO might not show lCNV development due to a ceiling effect because of a hyperactivity in FR (locus coeruleus, noradrenergic) (21, 22) or a hyperdisinhibition in BG (dopaminergic) (23) or both (24). Further research is urgently needed to reveal the mechanism responsible for our results, as depending on their nature therapeutic consequences will be quite different.
In CO lCNV amplitude might increase until puberty and decline again afterwards (25). This would parallel the development of the dopaminergic system in the striatum revealed by pre-synaptic dopamine transporter density (26). Age is an important factor that future studies of CNV should not ignore, either in childhood or in adult headache.
Footnotes
Acknowledgements
This work was supported by the Pain Research Programme of the Medical Faculty, University of Heidelberg (F207040, E1). The authors would like to thank Kerstin Herwig for helping to acquire the data. We would like to thank the reviewers for their remarks.