Abstract
Prevalence studies exploring the relation between socio-economic status (SES) and headache have shown conflicting results. This is the first prospective study analysing the relation between SES and risk of headache. A total of 22 685 adults not likely to suffer from headache were classified by SES at baseline in 1984-1986, and responded to a headache questionnaire in a follow-up 11 years later (1995-1997). SES at baseline was defined by educational level, occupation, and income. The risk of frequent headache and chronic headache (> 6, and ≥ 15 days/month, respectively) at follow-up was estimated in relation to SES. When defining SES by educational level or type of occupation, low status was associated with increased risk of frequent and chronic headache at follow-up. The risk of frequent and chronic headache decreased with increasing individual income, but only among men. We conclude that individuals with low SES had higher risk of frequent and chronic headache than people with high SES.
Introduction
Inequalities in morbidity related to differences in socio-economic status (SES) are substantial in western Europe and the USA (1, 2), but studies of SES and headache have shown conflicting results. Clinical studies have reported that migraine occurs more frequently among intelligent and highly educated individuals (3, 4), whereas cross-sectional population-based studies have not confirmed these results (5–18). In contrast, some have reported an increased prevalence of headache and migraine among groups with low income (19–22) and low education (22–25). Such an inverse relationship between headache and SES has not been reported in studies outside the USA.
Cause and effect cannot be distinguished in cross-sectional studies, and a prospective study design is preferable. In this prospective study of a large unselected population, we examined the relation between SES measured between 1984 and 1986, and the subsequent risk of migraine, non-migrainous headache, and frequent headache (>6 headache days/month) at follow-up 11 years later (1995–1997). We also assessed the cross-sectional association between SES and headache prevalence as estimated at follow-up in 1995–1997.
Material and methods
In Nord-Tr⊘ndelag County in Norway, two population-based cross-sectional surveys (the HUNT study) have been performed (26, 27). The first survey (HUNT-1) took place in 1984–1986, and the second (HUNT-2) in 1995–1997.
In HUNT-1, all residents ≥20 years old were invited, and a detailed description of the study population has been given by Holmen et al. (26). Briefly, out of 85 100 eligible individuals, 77 310 (91%) answered the questionnaire that was sent with the invitation. The participants reported their highest achieved level of education (six categories), and based on this information we divided the participants into three categories according to duration of education: ≤9 years (obligatory school), 10–12 years, and ≥13 years (university level).
The participants also reported their occupation (ten categories), and this information has been used to reclassify subjects into an approximation of the international social class schema recommended in a WHO report (Erikson, Goldthope and Portocarero, EGP) (1, 28). The reason for our preference of this class schema for the HUNT population is given elsewhere (29).
The participants were reclassified into three categories: (i) high social class (EGP social class I–II): management positions in public or private enterprise, and professionals (white-collar workers); (ii) medium social class (EGP social class III–IV): routine non-manual workers, small proprietors, farmers and forest owners, and other self-employed workers in primary production; (iii) lower social class (EGP social class V–VII): lower-grade technicians, supervisors of manual workers, skilled, semi- and unskilled manual workers (blue-collar workers).
Data on income in 1984 and 1985 were available from the Norwegian Revenue service, by linking data files using the 11-digit identity number given to Norwegian citizens at birth. The identity number was removed before data were supplied to the investigators. Analyses are based on mean income of the pensionable salary in 1984 and 1985. As a reference, mean income in Nord-Tr⊘ndelag County was £9513 (NOK 109 071) for men and £4173 (NOK 47 851) for women (NOK= Norwegian krone; 1 NOK= £0.09 as for 6 June 2002). As for education and occupation, we divided the participants into three income categories, where low income was defined as an income below the 33 percentile (£8267 (NOK 94 800) among men and £1400 (NOK 16 050) among women). High income was defined as an income above the 66 percentile (£11 416 (NOK 130 900) and £5534 (NOK 63 450) among men and women, respectively).
We wanted to establish a headache-free population at baseline. The HUNT-1 questionnaire did not include headache items, but 59 471 persons responded to a question on use of analgesics (‘How often have you taken pain-relieving medication during the last month?’). A total of 41 581 responded that they ‘never’ used analgesics. For the purpose of the present study, we have assumed that among those who had never used analgesics, the proportion of headache sufferers would be negligible. The ‘never users’ were slightly younger, and had higher income, educational level, and social class by occupation compared with those who used analgesics.
Among the 41 581 ‘analgesic-free’ individuals, 7887 had died or moved out of Nord-Tr⊘ndelag County (demographic characteristics not available). Thus, 33 694 were available for HUNT-2, and of these, 22 718 (67%) responded to a headache questionnaire. Individuals who responded to the headache questionnaire tended to be younger (55.0 vs. 58.6 years, P<0.0001), more likely to be women (47% vs. 44%, P<0.0001), and with higher SES measured by education, occupation and income (P<0.0001) than the non-responders.
Among the 22 718 analgesic-free responders, a total of 22 187 reported headache frequency. Among the 22 718 individuals assumed to be free of headache in HUNT-1, 6317 (28%) reported to suffer from headache in HUNT-2 (migraine 7% and non-migrainous headache 21%).
In HUNT-2 (1995–1997), all inhabitants ≥20 years old were invited, and details of the study have been reported elsewhere (27). Briefly, the participants answered two questionnaires. Classification according to education and occupation was identical to that in HUNT-1. Income data in 1995 were available from the Norwegian Revenue service. As a reference, mean income in 1995 in Nord-Tr⊘ndelag County was £13 293 (NOK 152 414) and £8103 (NOK 92 909) for men and women, respectively. We divided the participants into three categories, where income below the 33 percentile was classified as low (£6471 (NOK 74 200) among men and £1143 (NOK 13 100) among women), and income above the 66 percentile was classified as high (£18 446 (NOK 211 500) among men and £11 914 (NOK 136 600) among women).
In addition, individuals ≤69 years old had to answer the question ‘Have you during the last year had trouble in handling expenses for food, transport, or residence, etc.?’. In the questionnaire, four response options were given: never, seldom, sometimes, and often.
The second questionnaire (Q2) in HUNT-2 included 13 questions on headache (27), and individuals who answered ‘yes’ to the question ‘Have you suffered from headache during the last 12 months?’ were classified as ‘headache sufferers’.
Individuals who reported suffering from migraine in the questionnaire were diagnosed as migraine sufferers. In addition, individuals who fulfilled the following criteria were also diagnosed as suffering from migraine: (i) headache attacks lasting from 4 to 72 h (≤72 h for those who reported frequent visual disturbances before the attacks); (ii) headache had at least one of the following three characteristics: pulsating quality, unilateral location, or aggravation by physical activity; (iii) during headache, at least one of the following was present: nausea, photophobia or phonophobia.
Our criteria for migraine were a modified version of the migraine criteria of the International Headache Society (IHS), the most notable modification being that severity of pain was not included among the pain characteristics (item no. (ii)). As a consequence, our migraine criteria were less rigorous regarding number of pain characteristics required for diagnosis. Also, our migraine criteria differed from the IHS criteria, for example, by not considering the number of previous attacks experienced over the lifetime. The discrepancy between our migraine criteria and the IHS criteria has been discussed previously (27).
Headache that did not satisfy the criteria for migraine was classified as non-migrainous headache. The headache diagnoses were mutually exclusive.
Migraineurs and non-migrainous headache sufferers were also categorized according to headache frequency. Since there was no significant difference between the diagnostic categories (migraine and non-migrainous headache) as to the association with SES, these groups were combined. Frequent and chronic headache were defined as headache >6 and 14 days/month, respectively.
Out of 92 566 invited individuals, a total of 49 948 subjects (54%) completed the headache questionnaire where information on income was available.
Ethics
The study was approved by the Regional Committee for Ethics in Medical Research, and by the Norwegian Data inspectorate.
Validity and reliability of the headache diagnosis
The criteria for headache diagnosis have been validated (27) by comparing the diagnoses based on information in the questionnaire with diagnoses made in a clinical interview in a sample of participants. For migraine, the positive predictive value of a questionnaire-based diagnosis was 84%, and the chance-corrected agreement (κ) was 0.59 (95% CI 0.47–0.71), which is considered good. For non-migrainous headache, and for frequent and chronic headache, κ values were 0.43, 0.50, and 0.44, respectively, which indicate moderate agreement (30).
The 1-year prevalence of migraine, non-migrainous headache, frequent, and chronic headache in HUNT-2 were 12%, 26%, 8%, and 2%, respectively (29).
Statistical analysis
Differences between proportions were analysed by χ2 test. P-values <0.05 were considered statistically significant. In multivariate analyses, using multiple logistic regression, we used information about SES (years of education, occupational status, and income) in HUNT-1 to estimate the relative risk of headache, as registered at follow-up in HUNT-2. Relative risk was calculated for migraine, non-migrainous headache, and frequent and non-frequent headache.
We evaluated potential confounding by age (5-year categories), body mass index (BMI), current smoking (yes/no), alcohol consumption (three categories), and physical activity (three categories).
The precision of the relative risks was assessed with 95% confidence intervals (CI). Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS), version 8.0 (SPSS Inc., Chicago, IL, USA).
Results
Whereas non-frequent headache showed no relation to SES, low SES defined by educational level and occupation at baseline was associated with increased risk of frequent and chronic headache. This was evident for both sexes and for both migraine and non-migrainous headache. Since there was no difference between the diagnostic categories (migraine and non-migrainous headache) as to the association with SES, these groups were combined, as displayed in Table 1. Among individuals with <10 years of education, the relative risk of both frequent and chronic headache was 1.8 (95% CI 1.5–2.4 and 1.2–1.8, respectively) compared with individuals with ≥13 years of education. This higher risk of frequent headache among lower-educated individuals was most evident among individuals <60 years old (Fig. 1). Low social class defined by own occupation was also associated with higher risk of frequent headache in both sexes and chronic headache in women (Table 1), evident in all age groups. Low SES defined by income was also related to higher risk of frequent and chronic headache, but only among men (Table 1). Adjustment for BMI, smoking, alcohol consumption, or physical activity did not change the results mentioned above.
Age-adjusted risk ratio (RR) of non-frequent (1–6 headache days/month), frequent headache (>6 headache days/month), and chronic headache (≥15 headache days/month) by education, occupation∗, and income
∗The HUNT occupational classification reclassified into the Eriksen, Goldthorpe and Portocarero (EGP) social class schema (28).
†CI, Confidence interval.
‡Seventy-three men and 1187 women reported that they had never had paid work (full-time household workers, students, or persons receiving social security).
Prevalence of frequent headache (%) in HUNT-2 by age in individuals with duration of education <13 years (○) and ≥13 years (▪) in HUNT-1 (n=20 627). ∗P<0.05.
Also in the cross-sectional analysis (Table 2), low SES was associated with higher prevalence of frequent and chronic headache, evident for both migraine as well as for non-migrainous headache. Overall, individuals with <10 years of education had a 90% higher prevalence of frequent (OR = 1.9, 95% CI 1.7–2.1) and 70% higher prevalence of chronic headache (OR = 1.7, 95% CI 1.4–2.0) compared with those with high education (≥13 years). Similarly, low social class defined by type of occupation and income was associated with high frequency of headache, evident for both sexes. A high prevalence of frequent and chronic headache was also found among those who were unclassified by occupation (students, full-time household workers or those who received social security) (Table 2). Further analysis revealed that a strong association between disability pension and frequent headache may at least partly explain this result (data not shown). Finally, the prevalence of frequent and chronic headache was twice as high among individuals who reported economic problems compared with those who ‘seldom or never’ had such problems (Table 2).
Cross-sectional analyses of age-adjusted odds ratio (OR) of non-frequent (1–6 headache days/month), frequent headache (>6 headache days/month), and chronic headache (≥15 headache days/month) by education, occupation∗, income, and self-reported economic problems
∗The HUNT occupational classification reclassified into the Eriksen, Goldthorpe and Portocarero (EGP) social class schema (28).
†CI, Confidence interval.
‡Thirty-seven men and 1979 women were full-time household workers; 340 men and 570 women students; 3844 men and 5129 women received social security; and 442 men and 781 women had missing data.
Discussion
In this prospective study, low SES defined by education and occupation at baseline was associated with increased risk of frequent and chronic headache 11 years later. This was evident for both sexes and for migraine as well as for non-migrainous headache. The risk of frequent and chronic headache decreased with increasing individual income, but only among men. Also in the cross-sectional part of the study, low SES was associated with frequent and chronic headache.
No previous studies concerning SES and headache have had a prospective design. Therefore, only the cross-sectional part of our study can be compared with previous studies. The present results differ from several of those studies, which have found a fairly uniform headache prevalence across categories of education (8–11, 14, 15, 19), occupational class (8–10, 15, 19), or income (5–7, 16–18). Our main finding is, however, in agreement with studies from the USA. Scher et al. (25) reported that frequent headache defined as 180 or more headaches per year was more prevalent in those with the lowest level of education, and Stewart et al. (20) found that headache was more frequent among migraineurs in lower-income groups.
Whereas low SES was associated with increased risk of frequent and chronic headache, we found no such relation between non-frequent headache and SES. In agreement, Schwartz et al. (31) reported increased prevalence of chronic tension-type headache with decreasing level of education, whereas the prevalence of episodic tension-type headache decreased. Our results indicate that SES is related to headache frequency rather than to type of headache. The lack of difference between our two diagnostic categories may reflect that these two headache types share common risk factors. However, it may also reflect limited diagnostic accuracy (27), and the fact that many migraineurs also have tension-type headache.
In cross-sectional studies, the relation between cause and effect cannot be distinguished, since the information about education, occupation type or income is reported at the same time as the information on headache is obtained. Accordingly, an inverse relation between headache prevalence and SES may be explained by headache interfering with scholarly achievements and job career, as suggested in the social selection model (32, 33). In our prospective study the information about SES was obtained on a presumably headache-free population. Therefore, it was less likely that headache had interfered with education or job career as registered at baseline. Our results indicate that other factors associated with low SES, such as stress, poor diet, or poor medical care, may influence headache risk (the social causation model) (32, 33). Against this, one may argue that adjustment for lifestyle factors such as smoking and physical activity did not change our results. However, the questionnaire-based information about smoking, alcohol consumption, and physical activity must be considered with caution, because these data were not validated. Furthermore, other unmeasured factors related to SES may have confounded the association.
Some participants may have changed their SES during the 11-year follow-up, i.e. completed higher level of education (especially among the youngest persons), or changed income or occupational status. Since information on SES was available in both HUNT-1 and HUNT-2, one might consider adjusting the risk estimates for change in SES during the follow-up period. This is, however, incorrect, since the information about SES in HUNT-2 was collected at the same time as the headache questionnaire was answered. In order to be relevant to the question of causality, it is an absolute necessity that the information about exposure be obtained before the development of headache.
Headache questions were not included in HUNT-1, and headache status at baseline had to be determined indirectly by using information on the use of analgesics. Thus, we had to assume that ‘never users’ of pain-relieving medication during the last month in HUNT-1 were unlikely to suffer from headache. It has, however, been reported that some headache patients do not relieve their pain with medication (34–37). If we included a substantial number of individuals at baseline who suffered from headache, this could have influenced our results, but it is difficult to ascertain in which direction. However, only individuals with frequent headache were of particular interest, since no relation between SES and non-frequent headache was found. Identification of a completely headache-free population is difficult, since most individuals will have had headache during the last year (38). However, it seems reasonable to assume that our ‘analgesic-free’ population at baseline had relatively minor headache problems compared with the general population.
We have no information why patients used analgesics in HUNT-1. Since we had to exclude individuals with all pain conditions resulting in analgesics intake, one may speculate whether exclusion of headache-free analgesic users from our study population has influenced the results. However, also among analgesic users, low SES at baseline was associated with increased risk of frequent and chronic headache (data not shown). Thus, most likely, inclusion of headache-free analgesic users in our ‘at risk study population’ would not have altered the findings.
In HUNT-1, 70% of the invited population responded to the question on use of analgesics, and similarly, nearly 70% of individuals invited for HUNT-2 responded to the headache questionnaire. Although the attendance rate was high, we cannot rule out the possibility of selection bias. Individuals who responded to the headache questionnaire were younger, more likely to be women, and had higher SES than the non-responders. Thus, generalization of our results to those who did not participate must be done with caution.
The risk of frequent headache increased among lower-educated people for both sexes, most prominently in individuals <60 years old. Thus, education level does not seem to be a strong indicator for SES among people born before the World War II, or alternatively, SES does not influence prevalence of frequent headache in elderly people.
There was a higher risk of frequent headache associated with low individual income. In the cross-sectional analysis, this was true for both sexes, but in the prospective part of the study it was found only among men. A more appropriate income measure for SES is household income adjusted for number of household members (39). In particular, this may be a more sensitive indicator for women, but unfortunately this information was not available.
There was also an association between self-reported economic problems and frequent headache. Thus, it seems reasonable to assume that low income reflects real financial problems. One may speculate that such problems induce a higher psychosocial stress causing frequent headache in lower-income groups.
The prevalence of frequent and chronic headache was high among individuals who were unclassified according to SES defined by occupation, most evidently in the group of individuals who received disability pension. This finding was surprising, since very few subjects received disability pension in Nord-Tr⊘ndelag because of headache, and it suggests that other disorders than headache may be emphasized when an application for disability pension is made.
There is some evidence that it is better to define the occupational status of women on the basis of their male partners' occupation (40, 41). In the present study, however, low social class defined by own occupation was associated with higher risk of frequent headache also among women.
Because SES seems to be a relatively strong predictor of frequent headache, it is important to consider this factor in studies of patients with frequent and chronic headache. If SES differs between groups, differences in headache frequency may be expected and results should be adjusted for SES. The three types of SES measures applied in the present study all have strengths and weaknesses, but they probably all reflect somewhat different aspects of SES (42).
In conclusion, low socio-economic status at baseline was associated with higher risk of frequent headache 11 years later, evident for both migraine and non-migrainous headache. Assuming that the participants were relatively headache-free at baseline, it is unlikely that headache had interfered with educational or occupational functions. It would seem important to identify more closely which factors related to low SES are responsible for the increased risk of frequent headache.
Footnotes
Acknowledgements
The Nord-Tr⊘ndelag Health Study (The HUNT study) is a collaboration between The HUNT Research Centre, Faculty of Medicine, Norwegian University of Science and Technology (NTNU), Verdal, The National Institute of Public Health, The National Health Screening Service of Norway, and Nord-Tr⊘ndelag County Council. This study was partially financed by a research fellowship to K.H. from the Norwegian Research Council, and partly by GlaxoSmithKline, Norway.