Abstract
Since cognitive and behavioural characteristics of paediatric migraineurs have yet to be adequately defined, in this study we assessed the effect of migraine on the interictal functioning of children and adolescents by comparing the performance of two patient groups, 17 migraineurs with aura (MA) and 31 without aura (MoA) and by correlating the duration of the disorder, the frequency of attacks and interictal period with neuropsychological and behavioural findings. Both patient groups had cognitive performance within normal range except for a significant delay in the reaction time (RT) task. Both MA and MoA revealed a behavioural phenotype characterized by internalizing problems on Child Behaviour Check List (CBCL) scales. Slower RT to simple visual stimuli may be an early sign of a subclinical neuropsychological dysfunction, significantly correlated with the frequency of headache attacks and interictal period. The lack of a control group and other methodological limitations, such as patient selection bias and unadjusted P-value for multiple testing, make it difficult to give this finding a clearcut meaning. Further studies are needed on larger samples compared with a control group.
Introduction
Migraine is a common disorder in childhood and adolescence, although it is often underestimated because it is difficult to diagnose precisely. The classification criteria introduced by the International Headache Society in 1988 (1), and the subsequent revisions proposed for developmental age (2, 3), have led to a more accurate definition of the various forms of headache and have prompted a more careful differential diagnosis between migraine and tension headache.
It is important to emphasize that about 50% of children and adolescents with migraine continue to suffer from the disorder in their adult life (4). The connotation of migraine as a chronic, lifetime pathology makes assessing the neuropsychological and behavioural characteristics of paediatric migraineurs even more important, not only because of the diagnostic and therapeutic implications, but also for the prognostic aspects.
In the last 20 years, we have witnessed attempts to define the cognitive and behavioural phenotype of adult headache patients, migraineurs in particular, but the data available in the literature are still relatively scarce and partly contradictory, probably due to the use of different inclusion criteria and neuropsychological tests not always sensitive enough to detect mild cognitive dysfunction.
Some researchers have reported that migraineurs have cognitive deficits (5–11), while others did not support a link between migraine and cognitive impairment (12–15). The functions most often affected are memory (5, 8, 10, 11), information speed processing (5, 11), attention (8, 10, 11) and psychomotor ability (8, 9).
Interictal single proton emission computed tomography (SPECT) studies in adult migraineurs have recently recorded a greater incidence, by comparison with healthy controls, of focal and multiple brain perfusion anomalies affecting various cerebral regions (11, 16). The abnormal SPECT findings do not correlate with the clinical features of the headache (11) and their location in the brain is non-specifically related to functional deficits (11).
As for developmental age, only two studies have been published that assess the impact of migraine on neurocognitive performance. The first, conducted on a sample of 20 children with migraine (aged 7–11 years), failed to identify any deficiencies affecting intelligence, digit span or visual–motor integration, but found a significantly impaired performance for short- and long-delayed memory tasks, which might be responsible (according to the authors) for the high incidence of learning disabilities among juvenile migraineurs reported in the literature (17). Recently another study, comparing children aged 6–12 years with migraine and their healthy siblings, revealed no significant differences on a scale assessing sequential and simultaneous information processing (18).
More attention has been focused on investigating the psychological characteristics of juvenile headache (17, 19–21). Although the existence of a typical migraine sufferer personality has yet to be confirmed, these studies tend to agree that children and adolescents with migraine have certain personality traits in common, such as insecurity, perfectionism, inflexibility and competitiveness, together with a tendency to react over-anxiously to stressful outside events and a marked tendency for somatic complaints, which is the only feature that significantly differentiates paediatric migraine from other chronic pain disorders (19).
Due to the limited number of studies published on this topic in children and adolescents, the present study was designed to achieve the following objectives:
To assess the effect of migraine on the interictal cognitive and behavioural performance of children and adolescents;
To compare the performance of migraineurs with (MA) and without aura (MoA) in an attempt to pinpoint any specific function patterns for each group, and to identify tests sensitive to mild neurocognitive dysfunctions suitable for use in the diagnostic work-up;
To correlate the clinical variables (duration of disorder, frequency of attacks and interictal period) with neuropsychological and behavioural measures.
Methods
Participants
This is a single-institution, prospective study conducted from 2001 to 2003 at the Developmental Neurology Division of the C. Besta National Neurological Institute in Milan, Italy. In this period 92 children were evaluated for primary headache.
The following exclusion criteria were considered: (i) presence of any other systemic diseases or major psychiatric disorders; (ii) association of different types of headache; (iii) significant alterations in the neuroradiological and/or neurophysiological and/or blood chemistry test findings; (iv) anomalies emerging at neurological examination.
After this selection, 48 subjects (19 males and 29 females, age 6–17 years) entered the study, 31 MoA (13 males and 18 females, age 6 years 8 months to 16 years 4 months) and 17 MA (six males and 11 females, age 9 years 4 months to 17 years 4 months). Headache was classified according to the diagnostic criteria of the ICHD-II (3).
None of the patients was taking any medication before or during the study.
The testing session took place on a symptom-free day at least 2 days after the latest attack in order to rule out the risk of the subject being in a postdrome period. The fact that the patients reported no discomfort or pain at the time of the assessment enabled us to rule out any prodrome state effects on their neuropsychological performance.
All the patients underwent magnetic resonance imaging (MRI) of the brain, a set of blood chemistry analyses and a comprehensive neurological examination. MRI was performed according to the following standard protocol: T2-weighted images in the axial or coronal views, and T1-weighted images before and after contrast medium in the coronal, sagittal and axial views. The slice thickness was 4 mm.
The frequency of the attacks, the duration of the patient's history of migraine and the interictal period were also recorded. Duration was recorded as the number of months since the onset of the disorder, which was at least 6 in all patients, while frequency was calculated as the average number of attacks per month in the last 6 months. The duration of the disorder was significantly longer in MA (MA: median = 72 months, range 8–120 months; MoA: median = 21 months, range 6–96 months; Z =−2.99, P < 0.01) because of the higher mean age in this group (MA 160 ± 29 months, MoA 132 ± 31 months, t =−2.99, P < 0.01), while the frequency of the attacks did not differ significantly between the two groups (MA: median = 4, range 0.5–15 attacks per month; MoA: median = 8, range 0.5–26 attacks per month; Z =−1.58, P = 0.113).
The interictal period was calculated as the mean number of days between two successive attacks in a month. No difference emerged between the groups (
Cognitive and behavioural assessment
Patients were assessed in a quiet room at the Besta National Neurological Institute, always by the same examiners with specific training on the neuropsychological assessment of children. The assessment was completed in a single session taking an average of approximately 1.5 h, but tests were always suspended whenever patients requested it, or when it was evident that their attention or interest was waning.
The following were investigated: intellectual abilities with the Raven Progressive Matrices (22, 23); short-term sequential memory: the Digit Span Test [an item of the Wechsler Intelligence Scale (24)] was used for verbal recall, and the Corsi Span Test (25) was used for visuospatial memory; divided and selective visual attention, evaluated with Trail-Making Test (TMT) (26) and Cancellation test (27), respectively; information speed processing, using computerized simple visual reaction time task (RT) (28). The children used their index finger to press the space bar as quickly and accurately as possible after a white square appeared in the centre of the screen. Thirty stimuli were given and the independent variable was the mean response time. The interstimulus interval varied randomly from 2.5 to 4 s to discourage anticipations, i.e. responses occurring before 100 ms had elapsed. A deadline was fixed at 3 s and the absence of response within 3 s was considered as an omission. Emotional and behavioural problems are assessed using the Child Behaviour Check List (CBCL) (29, 30), a 113-item checklist completed by the parents of children aged 4–18 years old, giving rise to nine behaviour problem scales and an internalizing–externalizing dichotomy. Each item is scored on a response scale of 0–2, where 0 = not true, 1 = somewhat or sometimes true and 2 = very true or often true. The responses are based on current observation or behaviour observed up to 2 months before completing the questionnaire.
In our Division, all children with neurological diseases undergo neuropsychological evaluation. The protocol used in the present study was specifically designed to study the behaviour and higher functions of children with migraine and is part of a broader protocol that evaluates neuropsychological functions in children with different type of neurological disease. The Digit Span, Corsi span, Cancellation test and CBCL are administered in their Italian versions (see References quoted above) and have standard norms for Italian children. For the TMT, the normative data have been collected by our group and are not yet published.
The RT task, part of the Fepsy battery (28), were based on Dutch norms but, given the independence of this test from language or other cultural factors, it is unlikely that the use of Dutch norms had any effect on the results, and this also is confirmed by our ongoing data collection.
Statistical analysis
Apart from the Raven Progressive Matrices (percentiles), all cognitive test results were expressed as Z scores. The Z score indicates the deviation from the mean population score (which is set at zero), stratified by gender and age, and enables comparisons between different tests within and between subjects. Since the scores obtained for most dependent variables are not normally distributed, the Mann–Whitney U-test was used to compare MA and MoA in cognitive and behavioural measures, and the Wilcoxon signed ranks test was used to compare repeated measures within the same sample.
The non-parametric χ2 test was used to compare the observed frequencies of patients’ pathological scores with those predicted for the normal population. For the neuropsychological test, a Z score |≥2|, i.e. lying outside 95% of the normal distribution, is considered statistically abnormal. As for the CBCL, the observed frequencies of pathological scores were calculated for each scale with respect to cut-offs defined according to the Achenbach method (28). In short, the standards are based on the 98th percentile (Z score = 2) for the syndrome scales, and on the 90th percentile (Z score = 1.3) for the internalization and externalization scales.
Finally, the correlation between clinical data (duration of disorder, frequency of attacks and interictal period) and cognitive and behavioural variables, and between CBCL and neuropsychological scores was computed. Spearman's bivariate correlation coefficients (ρ) are reported. All the statistical analyses were two-tailed and P-values of ≤0.05 were considered significant.
Results
Table 1 shows the median and range performance obtained by MA and MoA in all neuropsychological tests. None of the subjects revealed performance deficits on the Raven Progressive Matrices (five had a performance from the 10th to the 25th percentile, nine from the 25th to the 50th percentile, 13 from the 50th to the 75th percentile, 10 from the 75th to the 90th percentile, 11 > 90th percentile) and there were no significant differences between MA and MoA in the distribution of the scores (χ2 = 2.742; d.f. = 4; P = 0.602).
Median Z scores and range performance obtained in the neuropsychological tests
Since the Mann–Whitney U-test recorded no significant difference in cognitive performance and behavioural problem subscales between MA and MoA, the subsequent statistical analyses were performed collapsing all the subjects into a single group.
Analysing neuropsychological performance for individual subjects revealed a percentage of pathological scores significantly higher than expected only for RT (dominant hand χ2 = 110.08, P < 0.001), as shown in Fig. 1. To be more precise, nine (29%) of the migraineurs without aura and three (18%) with aura had a pathologically slow RT, i.e. 2 SD above the mean for the normative group.
Percentage of pathological scores obtained in neuropsychological test in migraineurs (▪) compared with the normal paediatric population (□).
As regards the results of the CBCL, Wilcoxon's signed ranks test found internalizing (but not externalizing) scores significantly higher (i.e. pathological) among the migraineurs (Z =−4.95, P < 0.001). The observed frequencies show a distinctly higher proportion of pathological scores on the internalizing scale in our migraineurs than might have been expected (χ2 = 9.06; d.f. = 2; P < 0.05, see Fig. 2). In fact, approximately 20% of the patients obtained pathological scores—a considerably higher proportion than the 2% of pathological subjects estimated in the general population. By contrast, the percentage of subjects with pathological scores on externalizing scale (10%) did not differ significantly from the expected values (χ2 = 3.07; d.f. = 2; P = 0.215).
Percentage of pathological scores obtained in Child Behaviour Check List scales in migraineurs (▪) compared with the normal paediatric population (□).
Statistical significance was also reached in the scale-by-scale assessment on withdrawal (χ2 = 38.781; P < 0.001), somatic complaints (χ2 = 432.634; P < 0.001), anxiety/depression (χ2 = 69.134; P < 0.001), social problems (χ2 = 11.18; P < 0.01), thought problems (χ2 = 27.57; P < 0.001) and attention problems (χ2 = 10.07; P < 0.01), whereas the score for delinquent and aggressive behaviour did not indicate an incidence of difficult cases significantly higher than expected (respectively, χ2 = 0.00, P = 0.992; χ2 = 1.20, P = 0.548), confirming globally adequate behavioural patterns in externalization.
Most neuropsychological scores were not correlated with the clinical variables, but the simple RT to visual stimuli correlated significantly both with the frequency of the attacks (ρ= 0.394, P < 0.01) and with the interictal period (ρ= 0.394, P < 0.01).
None of the CBCL scores correlated with the cognitive measures or the clinical variables.
Discussion
This study achieved the following results: (i) both MA and MoA had normal neuropsychological findings, but significantly reduced information processing speed and a behavioural phenotype characterized by pathological internalizing features (withdrawal, somatic complaints and anxiety/depression); (ii) there were no significant differences on neuropsychological and behavioural measures between MA and MoA; (iii) only the simple RT to visual stimuli correlated significantly with the frequency of the migraine attacks and the interictal interval.
Concerning the first point, MA and MoA performed on Raven Progressive Matrices within the normal range, consequently ruling out any general cognitive deficits. These results are consistent with data both in adults (10) and in children (17). No significant problems were found in the auditory and visuo-spatial sequential memory, confirming previous findings in children (17) and adults (10) with migraine. No data are available on visual selective attention in children, but a good preservation of this cognitive function, and divided attention in particular, was confirmed in one adult study (11).
In our subjects, a significant dysfunction was seen only in the information processing rate. The simple RT to visual stimuli were slow in a significantly larger number of our cases than in the normal population. The resulting data are consistent with reports on adults (11, 31), demonstrating the persistence of the dysfunction in adulthood.
Simple visual RT involves different functions. Slower RT could be interpreted as a lower rate of transmission of information over white matter pathways connecting separate brain areas involved in most mental chronometrical tasks, even quite simple ones. There may consequently be a slow-down in information processing by the posterior cortical areas responsible for detecting the visual stimulus and by the premotor areas that program and implement the motor response (32). According to this interpretation, normal execution time for a complex task, such as TMT, is an unexpected finding, even if it is consistent with the results obtained by Calandre et al. in 60 patients with migraine, aged from 15 to 68 years (11).
A reconsideration of the attentional function underlying the RT may offer a plausible explanation for this unusual but recurrent dissociation in migraineurs. The simple RT are almost pure measures of psychomotor rate and depend mainly on the speed of signal detection and response initiation, whereas the TMT depends largely on divided attention processes, while a very marginal role is left for detecting the signal and initiating the reaction. According to Posner's model, attention includes three main functions: orienting to sensory stimuli, executive functions and maintaining the alert state (33). The explanation of RT lengthening could be the consequence of a selective disorder of the alerting network, which is involved in maintaining readiness to react according to the functional approach, rather than the general slowing of visuo-motor information processing from stimulus to response on the structural approach (33, 34). The structural approach interpretation is unconvincing and cannot account for patterns associating an impaired simple reaction time with a spared performance in more complex attentive tasks such as TMT.
Simple RT tests could be sensitive to attentional deficits because they stress the ability to trigger the same preprogrammed motor response as fast as possible throughout the task. The higher variability of RT (most notably in test with a skewed distribution) associated with a relatively well-preserved ability to produce a few fast responses (as shown by a normal minimum RT), and the disproportionate lengthening of median simple RT relative to other complex cognitive tasks, could be interpreted as evidence of a disorder of the supra-level system required to ensure the adequate activation of processes throughout the test (34). Similar considerations have recently been proposed by Buodo et al. (35), who measured event-related potentials and RTs during an acoustic oddball paradigm in 18 children with MoA. The authors suggested that difficulties in adjusting alertness level (possibly due to the repetitive stimulation features) and motivational engagement to task demands might explain the higher proportion of omission errors in migraineurs than in controls, and the increasing auditory RT over blocks of trials (35).
In our study the fact that longer simple RT was the only parameter indicative of any cognitive deficit may mean that difficulty in maintaining the alert state in visuo-motor processing speed tasks is the early, subtle sign of a subclinical neuropsychological dysfunction. The high rate of internalizing symptoms that could have slower information processing but the absence of correlation between CBCL scales and RT tasks confirms that unsteady alertness depends only on the migraine itself.
As for the behavioural assessment, the CBCL findings confirm that children and adolescents with migraine have traits such as introversion, a marked sense of insecurity, strong anxious reactions to stressful events and depressive symptoms, together with a significant tendency to develop somatic complaints. There are no apparent differences between MA and MoA.
In our sample, this psychosomatic predisposition was confirmed by clinical findings recording a high incidence in early childhood of various disorders with a mainly psychosomatic component, e.g. abdominal pains, motion sickness, paroxysmal vertigo and sleep disorders. It is so common to find such disorders in the remote clinical history of migraineurs that, for more than 20 years, they have been considered a risk factor for headache (4).
Difficulties also emerged on other CBCL scales. Judging from a qualitative item-by-item analysis, the relational difficulties seem to be attributable to a greater tendency to introversion, the thought problems to a prevalent ideation relating to pain and somatic complaints, and the attention problems to a poor concentration, selectively compromised with respect to spared selective visual attention and divided attention, as shown by cognitive results.
On the other hand, the delinquent and aggressive behaviour scales identified no more pathological cases than might have been expected, confirming globally adequate behavioural patterns on externalization.
The CBCL results are consistent with those reported in the literature in both adults (11, 36, 37) and children (17, 20, 21). Most studies conducted on the paediatric population did not use standardized checklist for parents, but self-reporting questionnaires and projective tests. These findings and our results indicate consistency between different sources of information and confirm the validity of the CBCL as a useful screening instrument for emotional and behavioural problems.
Concerning the second aim, the comparisons revealed no significant neurocognitive and behavioural differences between MA and MoA. This negative result confirms a uniform cognitive and behavioural phenotype among migraineurs in developmental age, irrespective of presence of aura—a finding consistent with the data already available in adults (10, 11, 37).
Finally, turning to our third aim, to correlate clinical variables (duration of disorder, frequency of attacks and interictal period) with neuropsychological and behavioural scores, only the RT correlated significantly with the frequency of headache attacks, as already shown by a previous report on adults (11). As for the duration of the disorder, most publications have reported no significant correlations with any mild cognitive dysfunction (5, 10, 13).
The slowing of simple visual RT seems to be an early marker of a subclinical neuropsychological dysfunction in both MA and MoA that correlates significantly with the frequency of the headache attacks and interictal interval. The interval between a migraine attack and task performance is probably an important independent variable. In an attempt to control the postictal functional and cognitive effects of a migraine episode (38), we assessed the patients at least 2 days after the latest attack, but their performance may have been influenced by fading prodromic symptoms, about which there is still insufficient knowledge, and their importance is still being debated (39). Moreover, the delay between the day of neuropsychological evaluation and the subsequent ictal episodes was not recorded in our present investigation. Considering the wide range of frequencies in our patients, the children may well have been investigated in different stages of the continuing pre or post effects of a migraine episode. As a consequence, varying degrees of interictal abnormality may have been detected. For the time being, it is hard to say whether the slower RTs, particularly evident in the subjects with a high frequency of attacks and a consequently shorter symptom-free interictal interval, were signs of a chronic dysfunction due to the migraine rather than to any transient and reversible impairment induced by a recent migraine episode, or the prodromic effects of a coming attack.
Finally, we wish to emphasize that several methodological and statistical limitations prevent any extension of our findings in a clinical sample to the general population of migraineurs. These limitations include: patient selection bias due to the fact that we studied a small sample of children and adolescents with migraine collected at our institute to infer general features of the heterogeneous population of migraineurs of developmental age; the lack of a matched headache-free control group selected from the same population; and the presence of type I errors as a result of multiple testing without adjusting the nominal α level.
Further studies on larger samples compared with a control group of normal children, or children with other pain diseases, are needed to confirm the sensitivity and specificity of the RT task, and its relationship to frequency of attacks and interictal periods in childhood migraine.
Footnotes
Acknowledgements
The authors thank Frances Coburn for her help with the English. Thanks are also due to Professor C. Umiltà for his useful suggestions.