Abstract
Background
Some mechanisms were proposed to explain a putative association between obesity and migraine, such as proinflammatory and prothrombotic states present in both conditions, shared genetic and environmental features and similar comorbidities (1).
Most epidemiological knowledge about the association between obesity and migraine is based on studies that used self-reported data on one or both conditions. There is evidence that this may be particularly biased. Obese migraineurs underestimate their body weight and body mass index (BMI) (2), and the proportion of individuals with migraine who remain undiagnosed in the general population is high (3).
Studies focused on the relationship between migraine and obesity have found conflicting results. Brown et al. (4) analyzed self-reported data from 14,779 women aged 18–23 years who participated in the baseline survey of the Australian Longitudinal Study on Women's Health. They found an increased odds ratio (OR) for migraine in participants with obesity (BMI > 30 kg/m2), compared to those of normal weight (18.5 to 24.9 kg/m2). In a multivariate analysis of the National Health and Nutrition Examination Survey (NHANES) 1999–2002 data (5) of individuals ≥ 20 years, those with BMI >30 kg/m2 reported severe headaches during the past three months compared to individuals with normal BMI. In another analysis based on NHANES surveys (from 1999 to 2004), and stratified by age and sex, Peterlin et al. (6) reported an association between obesity and migraine in men and women aged 55 years or less, but a lack of association in older individuals. Recently, Peterlin et al. (7), evaluating individuals aged ≥ 18 years in the National Comorbidity Survey Replication, a population-based national survey in the United States (188 with episodic migraine and 3674 controls), found episodic migraine to be associated with obesity in individuals aged 50 years or less, but not in older subjects. They also report that female sex and white race were positively associated with episodic migraine in that study. In the Nord-Trøndelag Health (HUNT) study, Robberstad et al. (8) describe a positive association between migraine and markers of unfavorable lifestyle, including overweight, in individuals aged 12–18 years.
However, in a telephone interview study of 30,215 participants ≥ 18 years (3791 with migraine), Bigal et al. (9) found a positive association between higher self-reported BMI and a high frequency of migraine episodes, but they found no association with migraine diagnosis. Keith et al. (10) analyzed data from 11 datasets in the United States (US) to analyze the association between BMI and the occurrence of headaches. These analyses yielded conflicting results. For most of the datasets, a positive association between BMI and headaches was observed. However, data from the Women's Health Initiative (WHI) was the sole dataset that focused exclusively on migraine diagnosis. In analyses restricted from that cohort, they found a slightly negative association between migraine and obesity in full-adjusted models. Mattsson et al. (11) studied the association between migraine and BMI in 684 women aged 40 to 74 years in Sweden using a structured questionnaire, BMI was measured in all participants. They found no association between BMI and migraine diagnosis, frequency or severity.
Studies on the association between migraine and abdominal obesity (AbO) are scarce. The same NHANES data analysis from Peterlin et al. (6) found a higher chance of migraine diagnosis in women with AbO younger than 55 years, which was independent of BMI. An inverse result, however, was found for older women.
Brazil is a country with high prevalence rates of migraine. A nationwide population-based study found an adjusted prevalence rate for migraine and probable migraine of 15.2% and 26.3%, respectively (12). The Brazilian Longitudinal Study of Adult Health (ELSA-Brasil) is a non-population-based, multicenter cohort study that included as part of the baseline assessment anthropometric measurements and a validated headache questionnaire, based on International Headache Society (IHS) criteria. We aimed to evaluate cross-sectionally the association between migraine (diagnosis and episode frequency) and obesity (measured by BMI) or AbO (measured by waist circumference) using data from ELSA-Brasil. We also hypothesized that the presence of AbO could influence the putative association between migraine frequency and obesity.
Methods
Study design
The ELSA-Brasil design and concepts have been detailed elsewhere (13). Briefly, it is a cohort study of 15,105 voluntary civil servants, aged between 35 and 74 years, from six cities. It focuses on cardiovascular diseases, diabetes and related disorders. Baseline assessment took place from August 2008 to December 2010, and included in-person interviews conducted by trained personnel and clinical and laboratory measurements.
Headache questionnaire and migraine diagnosis
All of the participants who answered “yes” to the question, “In the last 12 months, did you have a headache?” at the baseline evaluation were invited to answer a detailed headache questionnaire based on the IHS criteria (14). This questionnaire is validated in Brazil (15) and has been used in previous studies (16). Migraine was defined as the presence of definite migraine (IHS reference codes 1.1/1.2) or probable migraine (IHS reference code 1.6). The no-migraine group (
For this paper, we subdivided individuals with migraine in groups according to the following question: “In the past 12 months, at which frequency did you have headaches?” For better comparison, we also present here an approximate equivalence to weekly frequencies. Possible answers were (1) less than one per month, (2) from one per month to one per week (one to four per month), (3) two to six per week (five to 24 per month) and (4) daily (>24 per month). In our article, the term “non-daily migraine” is related to individuals who (a) fulfilled criteria for migraine or probable migraine and (b) answered the question about headache frequency with option (1), (2) or (3). The term “daily migraine” is related to individuals who (a) fulfilled criteria for migraine or probable migraine and (b) answered the question about headache frequency with option (4).
Anthropometric measurements
Anthropometric measurements were assessed using standard equipment and techniques, with high reproducibility (13,17). BMI was calculated as weight (in kilograms) divided by the square of height (in meters) and organized in strata based on the World Health Organization (WHO) criteria. A normal BMI was considered as a BMI of 18.5 to 24.9; overweight was defined as a BMI of 25.0 to 29.9; and obesity was defined as a BMI ≥30.0 kg/m2. AbO was defined as a waist circumference >88 cm in women and >102 cm in men (18).
Other variables
Race, monthly income and educational level were self-reported, and categorized accordingly. Age was classified in strata of 10-year intervals (35–44, 45–54, 55–64, 65–74). Information on medication use in the 15 days before the baseline assessment was obtained from ELSA-Brasil participants. Patients were categorized according to smoking status as never, past or current smokers. We defined diabetes as a medical history of diabetes mellitus or diabetes treatment, a fasting serum glucose ≥ 126 mg/dl, HbA1c levels ≥ 6.5% and/or a two-hour oral glucose tolerance test ≥ 200 mg/dl. Hypertension was defined as a medical history of hypertension or hypertension treatment, a systolic blood pressure ≥ 140 mmHg or a diastolic blood pressure ≥ 90 mmHg. Subjects were classified as using prophylactic medications for migraine if using medications with evidence levels A or B, according to the 2012 guidelines of the American Academy of Neurology (19).
Study sample
In this article, we excluded from the study sample (1) 194 individuals with missing information on the headache questionnaire, anthropometric measurements and/or medication use; (2) 204 participants using one or more drugs that are available in Brazil for weight loss treatment (20); or (3) 141 subjects because of BMI less than 18.5 kg/m2, because of the small number. Thirty-six (25.5%) of these 141 subjects had non-daily migraine and only one (0.7%) of them had daily migraine.
Statistical analysis
Categorical variables were presented as proportions. Chi square, Fisher's exact and Kruskal-Wallis tests were used whenever applicable. A Chi-squared trend test was performed to analyze whether the proportion of obese individuals changed with higher migraine episode frequency.
We built multinomial logistic regression models to determine whether BMI and/or AbO were associated with all migraine, non-daily and/or daily migraine. In all multinomial models, the dependent variable was categorized as “no migraine,” “non-daily migraine” and “daily migraine.” “No migraine” was the reference group, and it included two subgroup of individuals: (1) those who answered “no” to the question “In the past 12 months, at which frequency did you have headaches?” and (2) those classified with headache types other than migraine exclusively. All models included both AbO diagnosis and BMI strata as independent variables. Models are presented (1) as crude rates, (2) adjusted for age and sex and (3) full-adjusted. To determine which variables should be included in the full model, we considered other confounding variables associated with obesity (race, educational level, monthly income, smoking status, diabetes diagnosis, hypertension diagnosis) with a
We ran post-hoc analyses analyzing the association between migraine and obesity in subgroups according to age/gender (men <50 years old, men ≥50 years old, women <50 years old, women ≥50 years old), and race (white, brown, and black). In addition, we ran full-adjusted post-hoc models restricted to individuals with migraine, to determine whether daily migraine was independently associated with obesity compared to those with non-daily migraine frequency. We used interaction models to study the influence of body fat distribution on the association between BMI and migraine in all individuals, in those <50 years old and in those ≥50 years old. All models reached convergence.
The significance level was set at 0.05. The analysis was carried out using R for Windows version x64 2.15.1 (22) and nnet package (23).
Results
Baseline characteristics of ELSA-Brasil participants (women and men), according to migraine diagnosis and frequency.
ELSA-Brasil: The Brazilian Longitudinal Study of Adult Health; IQ: interquartile; USD: United States dollars.
Use of prophylactic medications for migraine evidence level A or B.19
Figure 1 shows, for the group of individuals with migraine, the proportions of individuals from each BMI category according to the frequency of episodes. The proportion of obese individuals increased as migraine frequency also increased ( Proportion of participants within each body mass index (BMI) stratum (in kg/m2), according to migraine diagnosis and frequency.
Odds ratios and 95% confidence intervals for all migraine, non‐daily and daily migraine according to the presence of overweight, obesity and abdominal obesity.
Abdominal obesity was defined as a waist circumference >102 cm in men and >88 cm in women. Reference level for BMI strata was a BMI between 18.5 and 25 kg/m2. Full-adjusted models are adjusted for age, sex, race, educational level, monthly income, smoking status, diabetes diagnosis, hypertension diagnosis and the use of medications classified as having evidence level A or B for migraine prophylaxis.19
We found a positive association between obesity and daily migraine in full-adjusted models. Excluding from the reference group those individuals with other headaches yielded similar results (online Supplementary Appendix 1). In addition, models comparing individuals without any headache to those with definite migraine showed higher ORs for the association between daily migraine and higher BMI strata. In this model, the previous trend toward a positive association between overweight and daily migraine reaches statistical significance (OR 2.21; 95% confidence interval (CI) 1.18–4.14) (online Supplementary Appendix 2). Post-hoc analyses, restricted to individuals with migraine, revealed that daily migraine remained significantly associated with obesity in full-adjusted models (
Although AbO was not independently associated with migraine in full-adjusted models (Table 2), models to study the interaction among BMI strata, AbO and age found significant interaction terms. We found that the association between higher BMI and daily migraine was affected by the presence of AbO (
Odds ratios and 95% confidence intervals for daily migraine according to presence of overweight or obesity in individuals with and without abdominal obesity.
Abdominal obesity was defined as a waist circumference >102 cm in men and >88 cm in women. Reference level for BMI strata was a BMI between 18.5 and 25 kg/m2. Full-adjusted models are adjusted for age, sex, race, educational level, monthly income, smoking status, diabetes diagnosis, hypertension diagnosis and the use of medications classified as having evidence level A or B for migraine prophylaxis.19
Discussion
We found a strong association between higher BMI levels and daily migraine in ELSA-Brasil participants independent of age, sex, race, educational level, monthly income, smoking status, diabetes diagnosis, hypertension diagnosis and use of prophylactic medications for migraine. However, BMI was not associated with migraine when migraine diagnosis, regardless of frequency, was studied as the dependent variable. AbO was not associated with migraine diagnosis or frequency in the multivariate analyses. Interaction and stratified models, however, showed that the positive association between obesity and daily migraine remained strong only in the absence of an AbO diagnosis. AbO influence was more prominent in individuals aged 35–49 years.
Despite negative results in studies with smaller sample sizes (11,24), our findings are consistent with studies that included a high number of participants, such as those conducted by Bigal et al. (9), with the advantage of using structured strategies to assess anthropometric measurements and migraine diagnosis in a very large sample.
Our results contrast with the findings of Mattsson et al. (11), who reported a non-association between BMI and migraine frequency. Two major differences may explain these conflicting results. First, in their report, differences in migraine frequency among obese and non-obese women were tested comparing the median number of attacks in each group. Because the proportion of individuals with very high episode frequency is small, this method of analysis may mask a real association between obesity and this specific subset of migraineurs. Second, although both studies used consistent methodologies for measurements, in the present work, we have studied a much larger sample. Although there is no description of how many daily migraineurs they had in their work, it is reasonable to presume that this number would be too small to allow conclusions on the association between obesity and daily migraine. In Keith et al.'s study (10), when analyzing WHI data, they found a slight but significant negative association between migraine and obesity. However, WHI data are limited to women aged 50 to 79 years, an age strata remarkably different from the majority of migraine patients. In most of the other publicly available epidemiologic datasets described in that article, we observe a significant positive association. However, for those studies, questionnaires were focused on headache diagnosis instead of migraine, which reduces comparability to our results.
Migraine diagnosis was not significantly associated with BMI in the present study. We detected an association with obesity only in individuals with daily episode frequency. It is possible to speculate that this could happen because obesity may influence the progression of episode frequency in individuals with migraine, but may not actually induce migraine. This result contrasts with the results from studies conducted by Brown et al. (4), Ford et al. (5) and Robberstad et al. (8). Some differences may be pointed out between those studies and ours. The most important ones are related to samples and populations in each study. Brown et al. analyzed women aged 18–23 years using data from the baseline survey of the Australian Longitudinal Study on Women's Health. Ford et al. analyzed men and women, aged 20 years or older using data from the NHANES Survey, a population-based national study in the US. Robberstad et al. studied junior and high school students from the HUNT study in Norway. In our study, we focused on analyzing cross-sectionally middle-aged male and female civil servants using ELSA-Brasil baseline data. It is reasonable to consider these differences as partially responsible for different study results. However, in the study by Robberstad et al., a clear association between headache frequency and poor or very poor lifestyle status was present for those with monthly or weekly/daily headaches, but not for those with less-frequent headaches. Regarding individuals with migraine in particular, the association between weekly/daily migraine and overweight was the strongest among headache subtypes (OR 2.2; 95% CI 1.5–3.3), consistent with our findings. Unfortunately, they do not describe in detail the putative association between other levels of migraine frequency and BMI.
Recently, Peterlin et al. (7) studied a population-based sample of 3862 individuals and concluded that episodic migraine was independently associated with obesity, with higher ORs in individuals younger than 50 years of age, in women and in those of white race. We did not find any significant associations between obesity and all or non-daily migraine in these subgroups in our sample. Some differences between their study and ours may also explain these conflicting results. First, Peterlin et al.'s study included individuals aged 18 years and older, while the ELSA-Brasil cohort study included participants aged 35–74 years. Additionally, individuals with remitted migraine, who had did not have any migraine episodes during the year before the ELSA-Brasil assessment, were classified as having no migraine in our study. In addition, race may be more difficult to define in the Brazilian population, a society with predominantly European ancestry, but still with a high rate of racial blending (25).
It could be argued that the association between BMI and daily migraine is solely mediated by the use of prophylactic medications, as some directly induce weight gain (antidepressants, valproate) or are associated with the development of depression (beta-blockers), which may ultimately promote obesity (21). According to our results, however, this explanation appears to be unlikely. A very low proportion of daily migraineurs in this study were on medications used for migraine prophylaxis. Furthermore, BMI remained associated with daily migraine in the logistic regression models after adjustment for the use of these drugs.
We also did not find a significant association between AbO and migraine diagnosis or frequency. Peterlin et al. (6) have described opposite results for the influence of AbO on migraine, as they found a positive association for young women and a negative association for older women. This may reflect differences in the study populations. However, although their study used a similar methodology to assess AbO, migraine assessment was based solely on one question used by NHANES. As mentioned by those authors, this may have led to some inaccuracy in their study.
Our results suggest that the association between BMI and daily migraine may also be influenced by the presence of AbO. Independent associations have been described between body fat distribution and, for example, coronary heart disease (26) and some types of cancer (27,28). Age and adipose tissue distribution (29,30) affect systemic metabolism and adipocyte function. Although our data do not enable us to indicate causality, we could speculate that some nuances in adipocyte function according to body fat distribution and age may influence the relation of higher BMI with the frequency of migraine episodes. Along with the differences noted above, comparing our study and Peterlin et al.'s study (7), we also verify differences in the age of included participants between our study and those of Robberstad et al. (8) (12–18 years) and Brown et al. (4) (18–23 years). As a corollary to our results, it is plausible to admit that the relationship between migraine and obesity may vary with age, and this may be an explanation for different results.
Our study has some strengths. ELSA-Brasil is a multicenter study that includes civil servants from six cities in three regions in Brazil (South, Southeast and Northeast), which is a country with a high prevalence of migraine (12,31). We used a structured, validated questionnaire for migraine diagnosis according to the IHS criteria, which was administered by trained staff. Anthropometric measurements were also assessed by trained staff and had high reliability coefficients (32). The study also has limitations. First, one important limitation is the impossibility of classifying headache as episodic or chronic since we do not have information as to whether the participant presents with more than 15 days of pain per month. Therefore, we may expect that some individuals with chronic migraine but who do not experience migraine pain every day are classified in the non-daily migraine group. However, in a general sample like ours, we expect this number to be small compared to the total number of individuals with episodic migraine in this group. Second, as the ages of ELSA-Brasil participants are 35–74 years and a substantial proportion of individuals have migraine at an earlier age, we cannot state that our findings are also applicable to individuals younger than 35 years old. As some individuals may have remitted migraine at this age, findings of non-association in the all-migraine group may have been influenced by this fact. Third, although the association between migraine and obesity can be adequately studied, it is not possible to infer causality or prevalence rates from these data because of study design. Fourth, it is also important to note that our data in a sample of Brazilians cannot be generalized to other populations because of differences in obesity and migraine prevalences among countries. However, ELSA-Brasil is a great opportunity to evaluate this association in an upper-middle income country. Fifth, the number of individuals with daily migraine in our sample was small. We acknowledge this can influence our ability to produce accurate OR estimates, especially when dividing participants in subgroups. However, we still had power in this sample to detect significant interaction terms and associations between daily migraine and obesity in stratified analysis. This suggests that age and AbO strongly influence the association between obesity and daily migraine.
In conclusion, using a structured protocol for anthropometric measurements and migraine diagnosis among ELSA-Brasil participants, we found daily migraine to be associated with higher BMI levels, but not with AbO. The association of BMI and daily migraine was influenced by the presence of AbO in individuals aged 35–49 years. On the other hand, when analyzing migraine regardless of episode frequency, we found no associations between this condition and obesity or AbO measurements. Our results suggest that higher BMI levels may be associated with higher episode frequency in individuals with migraine, especially in those without AbO.
Clinical implications
Most studies assessing the association between migraine and obesity use self-reported data. In this study, using validated headache questionnaires and anthropometric measurements in a large sample, body mass index had a significant positive association with daily migraine. Abdominal obesity was not independently associated with daily migraine. Our results indicate that abdominal obesity may influence the association between bodymass index and daily migraine, mainly in individuals aged 35–49 years.
Footnotes
Ethical considerations
The study protocol conforms to the Declaration of Helsinki and was approved by each institutional ethics committee (Hospital Universitário-USP, Fundação Oswaldo Cruz, Instituto de Saúde Coletiva da UFBA, UFMG, Centro de Ciências de Saúde da UFES, Hospital de Clínicas de Porto Alegre). All individuals provided informed consent.
Funding
The ELSA-Brasil baseline study was supported by the Brazilian Ministry of Health (Science and Technology Department) and the Brazilian Ministry of Science and Technology (Financiadora de Estudos e Projetos and CNPq National Research Council) (grant numbers 01 06 0010.00 RS, 01 06 0212.00 BA, 01 06 0300.00 ES, 01 06 0278.00 MG, 01 06 0115.00 SP and 01 06 0071.00 RJ).
Conflicts of interest
VMP is a research fellow of the State of Minas Gerais Agency for Research and Technology (FAPEMIG). PAL and IMB are recipients of a fellowship from the Brazilian National Research Council (CNPq).
Acknowledgements
The authors thank the ELSA-Brasil participants who agreed to collaborate in this study. The authors also thank the research team of the ELSA-Brasil study for their contribution.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
