Professional Documents
Culture Documents
Longtermeffects
Longtermeffects
A recent cross-sectional study found that subjects reported for both migraine without aura (MO) and
with migraine, in particular migraine with aura MA.4-7 Other studies addressing the same issue
(MA), were more prone to subclinical strokes in the found no differences in cognitive skills between mi-
area supplied by the posterial cerebral artery as graineurs and controls.8-12 Potential problems of sam-
evaluated by brain MR images.1 The study also re- ple size, subject selection, the method used to
ported an increased occurrence of deep white matter establish headache diagnoses, and other method-
lesions in female migraineurs.1 Since such lesions ologic issues may, at least in part, explain the con-
accrue with age and are associated with cognitive flicting results reached by these studies. We sought
decline and dementia this finding has led to renewed to overcome several methodologic shortcomings by
speculation on whether migraine is a progressive conducting a large population-based study among
brain disease.2,3 middle-aged Danish twins in which headache diag-
A number of studies have concluded that migraine noses were established by neurologists according to
has a deleterious effect on cognitive skills, a finding the International Headache Society (IHS) criteria.13
From the Danish Twin Register & Epidemiology (Drs. Gaist, Bak, and Christensen), Institute of Public Health, University of Southern Denmark;
Department of Neurology (Drs. Gaist, Madsen, Tsiropoulos, Bak, and Sindrup), Odense University Hospital; Department of Clinical Epidemiology (Dr.
Pedersen), Aarhus & Aalborg University Hospital, Denmark; Department of Psychology (Dr. McGue), University of Minnesota, Minneapolis; and Department
of Neurology (Dr. Krogh Rasmussen), Hillerød Hospital, Denmark.
Supported by the Danish Medical Research Council (grant 22– 00 – 0451) and the Danish Headache Society.
Received May 27, 2004. Accepted in final form October 25, 2004.
Address correspondence and reprint requests to Dr. David Gaist, Epidemiology, IST, University of Southern Denmark, Sdr. Boulevard 23A, 5000 Odense C,
Denmark; e-mail: dg@dadlnet.dk
Female* Male*
All 24.0 (7.3) 10.9 (3.3) 5.5 (2.4) 47.4 (13.2) 24.5 (7.4) 11.1 (3.6) 4.7 (2.2) 46.2 (12.2)
Age, y
⬍50 25.4 (7.1) 11.5 (3.4) 6.4 (2.4) 53.4 (11.7) 25.2 (7.6) 11.4 (3.8) 5.5 (2.3) 50.8 (10.1)
50–54 24.3 (7.3) 10.9 (3.3) 5.8 (2.5) 49.2 (11.2) 25.0 (7.6) 11.5 (3.6) 4.9 (2.2) 48.8 (11.9)
55–59 24.0 (7.4) 10.8 (3.6) 5.3 (2.3) 47.2 (13.5) 24.8 (7.1) 11.1 (3.3) 4.5 (2.2) 45.9 (11.8)
60⫹ 23.1 (7.1) 10.8 (3.2) 4.9 (2.3) 43.1 (13.5) 23.6 (7.2) 10.6 (3.5) 4.4 (2.0) 42.3 (12.4)
Schooling, y
⬍9 22.4 (6.7) 10.1 (2.9) 4.9 (2.3) 42.9 (13.2) 22.9 (7.2) 10.0 (3.8) 4.2 (2.0) 42.1 (12.1)
9–10 25.4 (7.3) 11.6 (3.6) 6.0 (2.3) 51.7 (11.5) 25.8 (6.6) 12.1 (3.6) 5.2 (2.1) 50.8 (9.9)
11⫹ 28.2 (7.4) 12.9 (3.5) 7.1 (2.3) 55.3 (9.4) 29.1 (7.9) 13.7 (3.7) 5.9 (2.4) 54.3 (9.3)
Vocational training, y
0 21.7 (6.7) 9.8 (3.0) 4.7 (2.3) 41.6 (13.1) 21.6 (7.5) 9.4 (3.2) 3.9 (2.2) 39.0 (12.6)
1–2 24.3 (7.1) 11.2 (3.3) 5.7 (2.3) 49.1 (12.3) 24.3 (6.8) 10.9 (3.3) 4.7 (2.1) 46.1 (11.2)
3–4 27.1 (7.3) 12.0 (3.4) 6.4 (2.6) 52.2 (12.8) 26.8 (7.9) 12.7 (3.6) 5.3 (2.3) 52.3 (12.0)
5⫹ 28.2 (7.0) 13.5 (3.1) 7.1 (2.2) 54.4 (7.8) 29.0 (7.0) 13.6 (3.7) 5.8 (2.4) 53.2 (9.4)
migraine that was not verified at the neurologist interview. We (figure). Participation rate did not vary by sex (78.0% in
therefore repeated the analyses using another control group com- women vs 77.7% in men). We stratified eligible subjects by
prising twins from pairs where both twins had screened negative
for migraine.
sex and migraine screening status and found that in these
Within-pair analyses. We also performed within-pair compar- strata participants were generally younger (mean differ-
isons of the cognitive scores of twins with migraine, MO or MA, ence 0.6 to 1.1 years) than nonparticipants, with the excep-
with that of their nonmigraineur co-twins. Within-pair differences tion of screening positive men (mean difference ⫺0.3
in mean cognitive scores were adjusted for potential confounders years) (table 2). Participants had completed more years of
through an ANOVA model that a priori included educational at-
tainment and depressive symptom score. Other variables were schooling and vocational training, and achieved higher
chosen for inclusion as described above. Within-pair comparisons scores on all cognitive tests vs nonparticipants (see table
for cognitive scores were also performed stratified by age (⬍55, 2). However, with the exception of schooling, the magni-
55⫹ years), cumulative number of attacks (⬍50, 50⫹), and zygos- tude of these differences did not vary substantially by screen-
ity (monozygotic vs same sex dizygotic).
To account for the non-independence of the observations on
ing status (see table 2). In a separate analysis comparing the
twins, subjects from pairs where both twins participated (intact combined group of migraineurs and screening-positive non-
pairs) were analyzed as clusters of two in all multivariate models. participants vs the combined group of non-migraineurs and
Our primary goal was to study whether migraine through its screening-negative nonparticipants (worse case scenario
attacks had a detrimental effect on cognitive skills. We viewed analysis), the two groups achieved highly comparable cogni-
educational attainment as a potential confounder and adjusted for
its effect in the multivariate analyses. However, some mi- tive scores (data not presented).
graineurs have attacks at an age where they might interfere with A lifetime diagnosis of migraine (IHS 1) was established
their schooling or vocational training. Furthermore, according to in 536 subjects henceforth referred to as migraineurs. Mi-
some researchers migraine may be a neurodevelopmental disor- graineurs comprised 347 subjects with MO (IHS 1.1), 157
der26,27 in which case poor educational attainment may be linked
with migraine. We therefore also present data on educational
subjects with MA (IHS 1.2), and 85 twins with migrainous
attainment for both the entire dataset and the intrapair disorder (IHS 1.7). Twins with migrainous disorder ful-
comparisons. filled all but one of the criteria for a diagnosis of MO (83
In the within-pair comparisons we also estimated ORs for edu- twins) or MA (2 twins). A diagnosis of chronic tension-type
cational attainment using conditional logistic regression. In order headache (IHS 2.2) was established in 68 subjects, and a
to provide more robust estimates we collapsed the data into two
groups for this analysis (schooling: ⬍9, 10⫹ years; vocational further 5 subjects had episodic cluster headache (IHS 3.1).
training: 0 to 2, 3⫹ years). In all, 64 subjects were diagnosed with secondary head-
aches. Although 17 of these subjects also had migraine, the
Results. The cognitive scores of participants in the base- debut of the primary and secondary headaches was sepa-
line interview varied across strata of sex, age, and educa- rated by more than a year in all cases. Finally, 51 mi-
tional attainment in a predictable manner (table 1). Of the graineurs and 5 nonmigraineurs had had a headache that
4,314 twins participating in the baseline survey, 1,789 they could not recall in sufficient detail to permit a diagno-
were eligible for inclusion in the neurologist interview, and sis to be established and were therefore classified under
1,393 (77.9%) subjects aged 45 to 67 years participated IHS 13, ”Headache not classifiable.”
602 NEUROLOGY 64 February (2 of 2) 2005
migraine (see table 3). Comparison of schooling and voca-
tional training between twins with migraine or one of the
migraine subtypes and nonmigraineurs did not reveal any
consistent patterns for migraine and MO (table 4). How-
ever, subjects with MA of both sexes had higher schooling
than nonmigraineurs, and female twins with MA had re-
ceived more vocational training than female nonmi-
graineurs, although none of these differences were
significant. Sex-stratified cognitive scores of migraineurs
and nonmigraineurs were highly comparable, although
subjects with migraine, MO or MA, consistently scored
marginally better than nonmigraineurs (see table 4).
Stratifying by age produced similar results (data not pre-
sented) as did adjusting for potential confounders (see ta-
ble E-1 on the Neurology Web site at www.neurology.org).
Consecutive inclusion of variables in the multivariate mod-
els as described in Methods had little impact on the results
(data not presented). Repeating the analyses with the
2,599 screening-negative twins as a control group produced
highly comparable results (data not presented).
Cognitive scores showed little variation when stratified
by the cumulative number of attacks of MO a subject had
had. Although not significantly so, subjects who reported
having had more than 100 attacks of MA had lower scores
on the digit span (mean 11.1 [SD: 3.4]), delayed word recall
(5.4 [2.4]), and symbol digit substitution test (47.7 [11.6])
than both subjects with 50 to 100 lifetime attacks (12.7
[4.9]; 6.0 [1.9]]; 48.2 [9.7]) and less than 50 lifetime attacks
(11.8 [3.9]; 5.9 [2.0]; 49.3 [12.1]) of MA. These differences
were less marked in women and were attenuated after
adjustment for potential confounders (data not presented).
No association was found between the age at migraine
Figure. Flow chart of study participation. *Baseline in-
debut or the number of years with migraine and cognitive
terview comprised 4,314 twins. However, in all, 22 twins
skills (data not presented).
emigrated (3), died (16), or were lost to follow-up due to
Within–pair comparisons. We identified a total of 139
secret address (3) after completion of baseline interview,
twin pairs (61 monozygotic and 78 same sex dizygotic) that
but before selection for present study. A further three twins
were discordant for migraine, MO or MA. Comparisons of
did not respond to migraine screening questions and were
mean cognitive scores revealed only minor differences of
also excluded. One twin died after random selection but
similar magnitude and direction as the individual-based
before contact was established.
analyses described above, in spite of the fact that the twins
with migraine had fewer years of schooling and vocational
A total of 75 twins participated from the random sample training than their co-twins (table 5). Adjusted within-pair
of 100 pairs where both twins were screening negative. A comparisons and analyses stratified by age and sex, and
diagnosis of MO was made in one twin, MA in none, and cumulative number of lifetime attacks produced similar
migrainous disorder in two twins. The negative predictive results (data not presented).
value of the screening questions for migraine including Twins with migraine had completed fewer years of
migrainous disorder (IHS 1) was 96.0% (95% CI [88.8% to schooling (10⫹ years OR: 0.31 [0.14 to 0.74]), and this
99.2%]), for MO it was 98.7% (95% CI [92.8% to 99.9%]), difference was also apparent in twin pairs discordant for
and for MA 100% (one sided 97.5% CI: 95.2%). MO and MA (see table 5). Twins with migraine, MO, or
Compared with nonmigraineurs of the same sex, mi- MA were not significantly different with regard to voca-
graineurs were of similar age, but less frequently single, tional training compared with their co-twins (see table 5).
current smokers, or heavy alcohol drinkers, and these dif-
ferences were more marked in men (table 3). Male mi-
Discussion. We found that 536 middle-aged Danes
graineurs also had a higher current depressive symptom
score compared with nonmigraineurs (p ⫽ 0.08). A neurol-
with a lifetime diagnosis of migraine or one of its
ogist diagnosis of chronic tension-type headache was 1.7 main subtypes achieved cognitive scores that were
times more common in women (p ⫽ 0.05) and 4.5 times highly comparable with those of 857 nonmigraineurs
more common in male migraineurs (p ⫽ 0.001) vs nonmi- in the same age group. Our results were not influ-
graineurs of the same sex. Female migraineurs were more enced by several potential confounders. Stratifica-
frequently current users of medications with a potential tion by number of lifetime attacks of MO or MA had
negative influence on cognition than nonmigraineurs no major impact on our results. Our study suggests
(13.6% vs 9.4%; p ⫽ 0.07) (see table 3). Less marked differ- that a lifetime history of migraine or one of its prin-
ences were found for a number of other potential confound- cipal subtypes has little impact on cognitive function.
ers of the association between cognitive skills and A number of studies provide evidence of cognitive
February (2 of 2) 2005 NEUROLOGY 64 603
Table 2 Eligible subjects stratified by sex, migraine screening status, and participation in present study (n ⫽ 1,789)
Female Male
Age, y, mean (SD) 56.2 (6.1) 58.3 (5.9) 56.2 (6.2) 58.2 (6.0) 56.5 (6.3) 56.2 (6.5) 56.9 (6.3) 57.5 (6.9)
Schooling, y, n (%)
⬍9 288 (50.5) 101 (68.2) 130 (54.2) 57 (70.4) 156 (51.7) 51 (62.2) 156 (55.5) 57 (67.1)
9–10 219 (38.4) 42 (28.4) 91 (37.9) 18 (22.2) 109 (36.1) 26 (31.7) 90 (32.0) 17 (20.0)
11⫹ 63 (11.1) 4 (2.7) 19 (7.9) 6 (7.4) 37 (12.3) 5 (6.1) 35 (12.5) 11 (12.9)
Vocational training, y,
n (%)
0 148 (26.0) 73 (49.3) 76 (31.7) 42 (51.9) 51 (16.9) 26 (31.7) 44 (15.7) 28 (32.9)
1–2 314 (55.1) 59 (39.9) 124 (51.7) 26 (32.1) 177 (58.6) 47 (57.3) 167 (59.4) 46 (54.1)
3–4 89 (15.6) 14 (9.5) 33 (13.8) 13 (16.0) 41 (13.6) 5 (6.1) 43 (15.3) 7 (8.2)
5⫹ 18 (3.2) 2 (1.4) 7 (2.9) 0 (0.0) 33 (10.9) 4 (4.9) 27 (9.6) 4 (4.7)
Cognitive test scores,
mean (SD)
Fluency* 24.4 (7.0) 21.6 (8.0) 24.4 (6.4) 21.9 (7.5) 24.9 (7.1) 22.0 (8.0) 25.3 (7.1) 23.2 (6.6)
Digit span† 11.0 (3.5) 9.8 (3.5) 11.2 (3.2) 10.0 (3.3) 11.3 (3.6) 10.5 (3.7) 11.4 (3.4) 10.4 (3.2)
Delayed word 5.8 (2.4) 4.8 (2.6) 5.7 (2.2) 5.0 (2.5) 4.8 (2.3) 4.5 (2.2) 4.8 (2.2) 3.8 (2.0)
recall‡
Symbol digit 48.9 (12.5) 43.6 (13.9) 48.2 (13.7) 42.7 (13.1) 46.9 (12.9) 44.3 (13.7) 45.7 (11.0) 42.9 (13.9)
substitution§
deficits in migraineurs, in particular in subjects with left-handed and had lower performances at age 3 to
MA,5-7,28 involving psychomotor skills,4,6,7 memory,4,5,7 13 on tests of verbal ability, but not on other tests of
and dysfunctions in the early stages of visual pro- IQ.27 Migraineurs in that study had also performed
cessing.29 However, in other studies of this issue no less well on school examinations at the age of 15 to
evidence of interictal cortical dysfunction was found in 17 and had not achieved bachelor degrees as fre-
subjects with migraine or MA.8-12 Several authors have quently as headache-free subjects.27 Also, a number
suggested that these inconsistencies in the literature of cross-sectional studies, in particular from the
may be due to methodologic issues such as sample size, United States, have reported an inverse relationship
highly unrepresentative samples, and lack of adequate between migraine prevalence and socioeconomic sta-
comparison groups.6,9,27 Interestingly, our results are in tus as measured by education or income,30-33 al-
line with the only other study of the subject that was though this association was less pronounced in a
carried out in a large population-based setting. This more recent study.34 In our study, we found no evi-
study identified 99 middle-aged subjects with self- dence of a link between educational attainment and
reported current migraine within the cohort of the migraine in the overall analyses. However, in the
Maastricht Aging Study and found that they fared within-pair comparisons, which are particularly in-
equally well on the letter-digit substitution test and teresting since these analyses control the large num-
delayed recall tests compared with 1,753 nonmi- ber of factors the twins share including childhood
graineurs from the same cohort.11 socioeconomic status, a tendency toward lower num-
Some researchers believe that migraine may be a ber of years of schooling in subjects with migraine,
neurodevelopmental disorder, in which case cogni- and in particular MA, was found. However, as the
tive function would be expected to be affected even confidence limits indicate, our results concerning ed-
before the onset of migraine attacks.26,27 A cohort of ucational attainment in the within-pair analysis
980 3-year-old children from New Zealand were fol- should be interpreted with caution.
lowed up for 23 years, serially subjected to neuropsy- Our study has a number of strengths. The sample
chological tests, and had migraine diagnoses is large and population-based, and the participation
established according to the IHS criteria at age 26. rate was high. The cognitive tests we used are sensi-
Compared with headache-free individuals, subjects tive to even mild cognitive impairment as demon-
who later developed migraine were more frequently strated by the analyses of the entire baseline sample
604 NEUROLOGY 64 February (2 of 2) 2005
Table 3 Characteristics of study participants stratified by sex and migraine diagnosis (n ⫽ 1,393)
Female Male
Age, y, mean (SD) 56.4 (6.2) 56.1 (6.1) 0.48 56.3 (6.1) 56.8 (6.4) 0.31
Married or cohabiting 301 (80.1) 338 (77.9) 0.45 147 (91.9) 347 (82.0) 0.003
Alcohol,† drinks per week
Abstainers 46 (12.2) 48 (11.1) 0.57 4 (2.5) 21 (5.0) 0.09
0–9 274 (72.9) 312 (71.9) 94 (58.8) 211 (50.0)
10–19 42 (11.2) 57 (13.1) 42 (26.3) 114 (27.0)
20⫹ 10 (2.7) 13 (3.0) 20 (12.5) 76 (18.0)
Smoker
Current 140 (37.2) 178 (41.0) 0.32 52 (32.5) 183 (43.3) 0.05
Past 87 (23.1) 83 (19.1) 61 (38.1) 146 (34.5)
Never 149 (39.6) 173 (39.9) 47 (29.4) 94 (22.2)
Current depressive symptom score†‡
⬍22 103 (27.4) 98 (22.6) 0.30 53 (33.1) 179 (42.3) 0.08
22–24 131 (34.8) 161 (37.2) 54 (33.8) 137 (32.4)
25⫹ 142 (37.8) 174 (40.1) 53 (33.1) 107 (25.3)
Ever self-reported physician diagnosed
Depression 80 (21.3) 74 (17.1) 0.13 17 (10.6) 45 (10.6) 0.99
Hypertension (drug treated) 75 (19.9) 74 (17.1) 0.29 24 (15.0) 62 (14.7) 0.92
Other disease with potential influence 26 (6.9) 41 (9.4) 0.19 8 (5.0) 17 (4.0) 0.60
on cognition§
Ever head-blow with loss of 76 (20.2) 83 (19.1) 0.71 53 (33.1) 126 (29.8) 0.45
consciousness†
Current drug use with potential cognitive 51 (13.6) 41 (9.4) 0.07 9 (5.6) 32 (7.6) 0.41
influence
Chronic tension-type headache§ 31 (8.2) 21 (4.8) 0.05 10 (6.3) 6 (1.4) 0.001
* Comparison between migraineurs and nonmigraineurs, 2 test for proportions and t-test for means.
† Information was missing for nine twins on alcohol, one on depressive symptom score, two on head-blow with LOC.
‡ Assessed by an adapted form of the Camdex Depression Scale (see ref. Johnson 2003).
§ According to International Headache Society criteria (2.2) and established through neurologist interview.
of 4,311 twins where poorer cognitive scores were were screening negative. Our validation study indi-
achieved with increasing age, even within the rela- cated that this is not a major concern, since the
tively narrow age-bands employed. Headache diag- negative predictive value of the screening instru-
noses were established by experienced neurologists ment was very high. Nonparticipants clearly had
according to the IHS criteria. The cognitive tests and lower cognitive scores and educational attainment
the questions screening for migraine were adminis- than participants, which could have biased our re-
tered as part of the baseline interview, at which time sults. However, the participation rate was high and
both the trained interviewers and the subjects were most of the observed differences were equally pro-
unaware of our research hypothesis. The extensive nounced among screening positive and screening
baseline interview information allowed us to gauge negative eligible subjects. Furthermore, cognitive
the effect of, and if necessary control for a multitude scores of migraineurs and nonmigraineurs remained
of confounders. Finally, the twin design enabled us highly comparable even after we included nonpartic-
to perform within-pair comparisons where we were ipants according to screening status (worse case
also able to adjust for any potential confounding ef- scenario analysis). We therefore believe that nonpar-
fects of childhood socioeconomic environment. ticipation had no major impact on our results. Our
Our study also has potential limitations. We used choice of control group could be questioned, mostly
a screening instrument to identify potential mi- owing to the high proportion of unconfirmed self-
graineurs. This approach could have resulted in false reported migraine. However, repeating the analyses
negative cases, i.e., subjects who had migraine but with the use of a large control group unrelated to the
February (2 of 2) 2005 NEUROLOGY 64 605
Table 4 Education and cognitive test scores of 1,393 middle-aged Danish twins by sex and migraine diagnosis
* Migraineurs (migraine, MO, or MA) compared with nonmigraineurs of same sex; t-test for means and 2 test for proportions. Men
with MA had better score on delayed recall (p ⫽ 0.03) compared with nonmigraineurs. All other tests produced p values greater than 0.10.
† Test not performed by one female and one male nonmigraineur.
‡ Test not performed by 11 female and 3 male migraineurs and 13 female and 13 male nonmigraineurs.
migraineurs and consisting entirely of subjects from some subjects who had problems recalling lifetime
pairs where both twins screened negative for mi- headache characteristics. We cannot overrule that
graine did not influence our results. It could be ar- this is the case for some migraineurs, in particular
gued that the age group studied may have included those with milder or less frequent headache attacks.
Table 5 Cognitive scores and educational attainment of 139 middle-aged Danish twin pairs discordant for a lifetime diagnosis of
migraine
Migraine, n ⫽ 139 pairs Migraine without aura, n ⫽ 94 pairs Migraine with aura, n ⫽ 30 pairs
Cognitive tests,
mean (SD)
Fluency 24.4 (6.8) 24.5 (7.0) 0.3 (⫺1.2 to 1.8) 24.7 (6.9) 24.3 (7.3) 0.9 (⫺1.1 to 2.9) 24.1 (8.1) 24.9 (7.0) ⫺0.4 (⫺5.7 to 4.9)
Digit span 11.2 (3.5) 10.6 (3.0) 0.9 (0.2 to 1.6) 11.0 (3.5) 10.4 (2.9) 0.9 (⫺0.1 to 1.8) 11.7 (3.5) 11.5 (3.0) 0.4 (⫺1.2 to 2.0)
Delayed word 5.6 (2.4) 5.5 (2.3) 0.3 (⫺0.2 to 0.7) 5.7 (2.4) 5.4 (2.3) 0.5 (⫺0.1 to 1.1) 5.6 (2.4) 5.5 (2.2) 0.4 (⫺0.9 to 1.7)
recall
Symbol digit 48.2 (11.2) 46.3 (13.0) 3.0 (0.6 to 5.4) 47.7 (10.1) 45.0 (13.0) 3.5 (0.5 to 6.5) 48.4 (11.3) 47.8 (13.8) 2.9 (⫺3.9 to 9.6)
substitution†
Schooling, y, n (%) OR OR OR
(95% CI)‡ (95% CI)‡ (95% CI)‡
* Difference in cognitive score (95% CI) adjusted for schooling, vocational education, depressive symptom score, and smoking.
† Missing information on four migraineurs and three nonmigraineurs (seven migraine pairs and three migraine without aura pairs).
‡ OR (95% CI) estimated through conditional logistic regression.
§ One pair not included due to missing information on one subject with migraine.
Updated Information & including high resolution figures, can be found at:
Services http://www.neurology.org/content/64/4/600.full.html
Neurology ® is the official journal of the American Academy of Neurology. Published continuously since
1951, it is now a weekly with 48 issues per year. Copyright . All rights reserved. Print ISSN: 0028-3878.
Online ISSN: 1526-632X.