The prevalence and characteristics of migraine in a population-based cohort

Methods.

The GEM Study is embedded within the MORGEN project—a population-based study designed to monitor risk factors for and the prevalence of chronic diseases of public health importance in Dutch adults age 20 to 65 years of age.15 The sample for the MORGEN project was selected randomly within equal-size strata of a 5-year age group and gender from population registries in two municipalities (Doetinchem in the eastern Netherlands and Maastricht in the southern Netherlands). The overwhelming majority of Netherlanders are white, which is reflected in the MORGEN sample. During 1995 to 1996 a total of 6,491 individuals (52.7%) participated in the study. Compared with responders, nonresponders were significantly younger than responders (mean age, 36.7 ± 0.15 years versus 39.8 ± 0.15 years) and more often were men (50.0% men and 44.6% women of those invited did not respond). Respondents signed a general informed consent for the MORGEN project, and a specific informed consent for the GEM Study.

MORGEN participants were mailed an extensive self-administered questionnaire that included questions about sociodemographic characteristics, medical history, and psychosocial functioning. This questionnaire was brought by respondents to the study center, where it was reviewed by a trained field worker. The participant then underwent a clinical examination, had blood drawn, and if assessed as screen-positive, filled in additional detailed questionnaires, including one on headache.

Within this framework, case finding for migraine was conducted in three stages. During the first stage, all respondents answered five screening questions in the mailed self-administered questionnaire. These questions were adapted from Stewart et al.9 On the basis of these questions, screen-positive was defined as those who ever had or had in the last 12 months, a severe headache (excluding those due to hangover or sinus infection) and the pain was 5 or higher on a 10-point pain scale or the participant was diagnosed with migraine by a doctor or used antimigraine medication (including sumatripten and ergotamine compounds). A respondent was also classified as screen-positive if there was a history of severe headache in the last 12 months and the pain was rated between 1 to 4 and one of 10 visual aura symptoms was experienced (questions are available from corresponding author). Trained field workers at the study center reviewed the screening questions and identified screen-positive participants.

During the second stage, screen-positive participants completed a more detailed questionnaire that focused on signs and symptoms of migraine headache and aura as outlined in the IHS criteria.4 To aid in the diagnosis of aura, participants were asked to indicate which of 15 focal (including 10 visual) signs they experienced just before the onset of headache. Those reporting visual symptoms were also asked to draw what they saw. This questionnaire was reviewed by trained field workers for consistency and completeness.

During the third stage of case finding, a random sample of screen-positive and screen-negative participants was selected for a telephone interview. Interviewers were medical students trained by study physicians experienced in migraine (G.T., M.F.). The semistructured interview was intended to clarify the signs and symptoms of migraine headache and aura reported in the stage 2 questionnaire. The interviewer had access to the questionnaire completed by the screen-positives. The interview for the screen-negatives was guided by the screening questions and the detailed questionnaire. Frequency of migraine attacks was ascertained during this interview. A final diagnosis was made after the telephone interview. If symptomology was not clear, the diagnosis was made in consultation with the two study physicians. Diagnoses were recorded as headache and aura in the following categories: definite, possible, unclear symptomology, and none. For this report, migraine aura without headache is classified in the migraine-with-aura group.

Results.

A total of 863 lifetime patients with migraine were identified, 620 of whom had active migraine. This included 17 lifetime (12 active) patients identified in the screen-negative subsample. The sensitivity of this case-finding procedure was 0.93 and the specificity was 0.36. The positive predictive value was 0.65 and the negative predictive value was 0.91. Thus the screening procedure picked up many false-positives and missed some true-positives.

Lifetime prevalence.

The lifetime prevalence in women is 33% (95% CI, 30.8 to 37.3) and in men is 13.3% (95% CI, 11.7 to 16.2). The prevalence of migraine increased with age (table 1, figure 1). Among men the prevalence increased from 5.5% (95% CI, 4.2 to 8.0) at 20 to 24 years of age to 23.1% (95% CI, 17.0 to 35.0) at 50 to 54 years of age. Among women, the lifetime prevalence increased from 17.9% (95% CI, 15.9 to 21.7) at 20 to 24 years of age to 40.7% (95% CI, 30.2 to 61.2) at 50 to 54 years of age. After age 25, the female-to-male ratio of prevalence declined steadily from 5.6 to 1.8. The prevalence of migraine was significantly higher in women compared with men, but did not differ by age or socioeconomic status (table 2).

View this table:

• View inline
• View popup

Table 1.

Prevalence of migraine: the Genetic Epidemiology of Migraine Study

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Lifetime prevalence of migraine by sex and age: the Genetic Epidemiology of Migraine Study.

View this table:

• View inline
• View popup

Table 2.

Sociodemographic characteristics of migraineurs and nonmigraineurs: the Genetic Epidemiology of Migraine Study

Prevalence estimates were notably affected when the patients found in the screen-negative group were accounted for. This effect was particularly evident in the prevalence estimates for those older than 40 years and for women (see figure 1, table 1). For instance, in 50-year-old women, the lifetime prevalence was 40.7% when the screen-negative patients were accounted for, and was 25% when these patients were ignored.

One-year prevalence.

The 1-year prevalence is 7.5% (95% CI, 6.8 to 8.9) for men, and 25.0% (95% CI, 23.0 to 28.8) for women. In men, the 1-year prevalence of migraine ranged from 3.0% (95% CI, 2.0 to 5.3) at age 20 to 24 years to 15.7% (95% CI, 10.5 to 25.7) at age 50 to 54 years (see table 1, figure 1). For women in the same age strata, the prevalence was 17.0% (95% CI, 14.9 to 20.7) and 24.0% (95% CI, 15.9 to 39.8) respectively, peaking at 33.2% (95% CI, 27.8 to 43.7) at age 35 to 39 years (see table 1, figure 2). There were no differences in prevalence by age or socioeconomic status (see table 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. One-year prevalence ratio of migraine and subtypes by age. Total = all migraine; MO = migraine without aura, MA = migraine with aura, and migraine both with aura and without aura.

Among the 620 active patients with migraine, 63.9% had MO, 17.9% had MA, and 13.1% had migraine both with and without aura (MOMA). The remaining 5% could not be subtyped. Presence of aura did not differ significantly by sex, age, or socioeconomic level (see table 2).

Migraineurs experienced a median of 12 attacks per year; 25% of the migraineurs had five or fewer attacks and 25% had two or more attacks per month. There were significant differences in attack frequency between men and women. There were no significant differences in attack frequency by age, socioeconomic status (see table 2), or subtype of migraine (data not shown).

Discussion.

Our data show that the prevalence of migraine is high, particularly among women of child-bearing age. Aura accompanied migraine headache in 31% of the sample of active migraineurs. Of these, almost half also experienced MO. The burden of migraine is quite high, with more than 25% of migraineurs experiencing an attack at least two times per month.

There are several aspects of our case-finding methods that need to be noted. This is one of the few studies during which an in-depth telephone interview was conducted.10,16 This telephone interview is similar to conducting a face-to-face clinical interview, which in the absence of any biological markers for migraine remains the gold standard for diagnosis. This process helps to clarify signs and symptoms of migraine and aura, and provides an extra degree of certainty about the diagnosis above and beyond that provided by a self-administered questionnaire used in other studies.9,17

Second, we not only conducted diagnostic interviews on screen-positive individuals, but also on a sample of screen-negative individuals. The screen-negative patients were incorporated into our prevalence estimates, which was not done in other studies. Our uncorrected 1-year prevalence estimates are similar to those estimated for other community samples.5 For instance, in our study the 1-year prevalence of migraine that ignores screen-negative cases is 6.7% in men and 18.6% in women. In the American Migraine Study9 it was 6.0% for men and 17.7% for women. When screen-negative patients are taken into account, our prevalence estimates are higher (25% in women and 7.5% in men) than those reported by other studies. It should be noted that studies by Rasmussen et al.6 in a Danish sample and Stewart et al.16 in a multiethnic sample of Americans administered a diagnostic interview to the total sample. These studies report 1-year prevalence estimates in white men similar to ours. The estimates for women differ: In the Danish study,6 the 1-year prevalence was 15%, in the American sample it was 20.4%,16 and in our sample it was 25.2%. This could be due to differences in the data available for the final diagnosis coupled with gender differences in how headache history is reported, or the characteristics of nonresponders. Furthermore, although the age range covered in the samples was similar, the distribution per 5-year age strata may have differed, and this might affect the overall prevalence given the increase with age in patients with migraine.

The differences in the corrected and uncorrected sex-age values ranged from 0 to 40%. The largest differences were in the lifetime estimates in persons older than 40 years of age and in women. The significant differences between the corrected and uncorrected prevalence in women older than 50 years of age suggest recall bias. Such bias has been examined in studies of psychiatric disorders, where it has been found that lifetime history is underreported when the illness is milder, less recurrent, or not treated.18 This is the first study to estimate the potential magnitude of the underestimation of lifetime prevalence estimates of migraine. However, the sample of screen-negatives that we interviewed, although equal across age and sex strata, was relatively small. Thus the weighted sex- and age-specific estimates in the screen-negative group are based on a small number of patients. This has likely introduced a certain degree of instability in the estimates, reflected by relatively wide CIs in some strata.

The response rate to the MORGEN project as a whole was moderate, although not substantially different from other population-based studies designed specifically to investigate migraine.9,10 The critical issue in assessing bias from nonresponse is whether migraine status was related to response. A tendency toward overrepresentation of migraineurs in participants was reported by Rasmussen et al.6; however, as they note, respondents were informed that the research was designed to study headache. Invitees to the MORGEN project were told the research was designed to investigate general health status. A small subsample of nonresponders to the MORGEN project (n = 56) answered the headache screening questions as a part of the nonresponse procedure. Overall, the percent of screen-positives was slightly lower in the nonresponse sample compared with the participants (19.6% versus 27.1%; p = 0.3). The percent of screen-positives for headache in the last year was the same between nonresponders and responders (18.6% versus 17.8%; p = 0.9). However, without a detailed headache history, it is not possible to ascertain the degree of selective response and whether it resulted in an over- or underestimate of prevalence.

The 31% of migraineurs with aura is slightly lower than the 38% reported in the population-based study of 40-year-old Danish adults.8 We find equal proportions of aura in men and women, in contrast to the Danish study that found more aura in women.8 Among those we identified with aura, 42% also had attacks without aura. In the Danish study, 13% of patients with aura were reported to have attacks with no aura.19 The difference in results could arise if respondents from the two studies recalled the aura differentially, which does not seem likely given the similarity in the overall proportion of patients with aura. Possibly, study differences in how MO and MA were ascertained or classified may explain these differences. MA is thought to have a stronger genetic basis then MO, and some have proposed the two types of migraine have different etiologies.19,20 However, the existence of a relatively large number of mixed cases raises the possibility of common etiology, including genetic factors. This mixed group may also be another distinct subgroup of migraineurs. Perhaps the aura in mixed patients has a different basis than that in migraineurs who always have an accompanying aura. To investigate further the differences in MO and MA, it will be important to identify in a standard way the group with a mixed history of aura.

As in other studies9,10 we found a peak in 1-year prevalence among women near 40 years of age. We found no difference in migraine prevalence by socioeconomic status, similar to that reported for the Canadian10 and Danish studies.21 In contrast, the American Migraine Study9 found that persons in the lowest socioeconomic strata had the highest prevalence. It is of interest to note that in Canada, Denmark, as well the Netherlands, there is virtually universal coverage of health care, which you do not find in the United States.

Frequency of migraine attacks is a measure of migraine-related disability. We found that 50% of patients with active migraine had at least one attack per month. This compares with data reported for the Canadian study22 and the French study.12 However, these other studies showed women to have more frequent attacks then men, which we did not find.