Abstract
Background
Several potential dietary trigger factors for migraine have been proposed. However, few studies have examined the intake pattern of these dietary items compared to adequate control populations and whether intake levels may vary by migraine aura status or attack frequency.
Methods
We conducted a cross-sectional study among participants in the Women’s Health Study. We used logistic regression to evaluate the association between migraine and headache status and low intake of foods commonly reported to affect migraine.
Results
A total of 25,755 women reported no history of migraine or headache, 5573 reported non-migraine headache and 7042 reported any migraine. Those with non-migraine headache or any migraine were more likely to have low intake of total alcohol (OR = 1.22, 95% CI:1.14–1.29 and OR = 1.17, 95% CI:1.11–1.24, respectively). Migraineurs with aura were more likely to have low intake of chocolate, ice cream, hot dogs, and processed meats. Those who experience migraine at least once per week were more likely to have low intake of skim/low-fat milk and white and red wine.
Conclusion
Intake of most suggested migraine dietary triggers differs by migraine aura status and attack frequency, a pattern not found for non-migraine headache.
Keywords
Introduction
Approximately one-fifth of women experience migraine and up to one-third of migraineurs experience migraine aura. Since migraine is a chronic-intermittent disorder and attack frequencies vary among migraine patients, there has been interest in determining what factors may trigger migraine attacks, including dietary factors. Several potential dietary trigger factors for migraine have been proposed, including alcohol, cheese, and chocolate (1). Most studies have only examined dietary triggers for overall migraine and are often lacking appropriate control populations. Moreover, data on dietary patterns by migraine aura status and migraine attack frequency are sparse and have mainly focused on alcohol consumption (2–5). One study that examined several specific food triggers observed that beer, citrus fruits, and vegetables were significant potential triggers for migraine with aura (MA) but not migraine without aura (MO) (6).
Using a non-headache and a non-migraine control group allows the evaluation of whether certain dietary intake patterns are specific to migraine. In addition, we can compare among different migraine subtypes and migraine attack frequencies. While we would expect all migraineurs to avoid similar food items in hopes of preventing the onset of migraine headache pain, differences in intake for those with different aura status or different attack frequency of migraine may suggest that avoidance patterns differ depending on disease severity.
Using a large population of initially healthy women, we examined whether low intake of foods commonly considered to be migraine triggers differed by migraine and headache status as well as by aura status and migraine attack frequency. We are examining low intake because we hypothesized that food avoidance would be associated with these various headache forms in this observational study of middle-aged women.
Methods
Study population
The Women’s Health Study (WHS) was a large, randomized, double-blind, placebo-controlled trial on the effect of low-dose aspirin and vitamin E on cardiovascular disease and cancer. The design and results of the trial have been previously reported (7–9). At baseline, 39,876 United States female health professionals aged 45 years or older free of any major illnesses were randomized to receive aspirin, vitamin E or their placebos in a factorial design. The WHS was approved by the institutional review board at Brigham and Women’s Hospital and informed consent was obtained from all participants.
Migraine and headache assessment
Prior to randomization, the women were asked “Have you ever had migraine headaches?” and “In the past year, have you had migraine headaches?” Women who reported experiencing migraine headaches within the past year were additionally asked about whether they experienced aura and about the frequency of their migraines (daily, weekly, monthly, every other month, and less than six times per year). The baseline questionnaire did not ask about non-migraine headache, therefore information from the six-month questionnaire was used to assess headache. The six-month questionnaire asked women if they had experienced any headache since the baseline questionnaire. Based on the women’s responses to these questions, we classified women as any migraine (women who reported ever experiencing migraine headache), non-migraine headache (women who reported experiencing headache but did not report migraine headache) and no history of headache (women who reported no history of migraine or non-migraine headache). In secondary analyses, we classified the any migraine group into those who reported migraine within the past year versus those who did not report migraine within the past year. We further classified women who reported migraine within the past year by aura status (MA versus MO) and by migraine frequency (at least once per week, monthly and less than monthly).
A previous study in the WHS has shown good agreement between self-reported MO and classification of MO based on the second version of the International Classification of Headache Disorders (10) criteria.
Food intake assessment
We limited our analysis to particular food items that have been previously hypothesized to be possible migraine triggers. These food items were chocolate, caffeinated coffee, decaffeinated coffee, total dairy products, cottage cheese, cream cheese, cheese (not cream or cottage cheese), ice cream, sherbet/ice milk/frozen yogurt, sour cream, yogurt, skim or low-fat milk, whole milk, cream, bacon, hamburger, hot dogs, processed meats, liquor, beer, white wine, red wine, and total alcohol consumption.
Prior to randomization, women were asked to complete a semiquantitative food frequency questionnaire (11). The validity and reproducibility of this questionnaire has been previously demonstrated (12). The women were asked how often they consumed a serving of each food item on average over the previous years. The possible response options were never or less than once per month, one to three times per month, once per week, two to four times per week, five to six times per week, once per day, two to three times per day, four to five times per day, and ≥6 times per day. These responses were converted into the average number of servings per day. Citrus was defined as the sum of oranges, orange juice, grapefruit, and grapefruit juice. Total dairy products was defined as the sum of cottage cheese, cream cheese, cheese (not cream or cottage cheese), ice cream, sherbet/ice milk/frozen yogurt, sour cream, yogurt, skim or low-fat milk, whole milk, and cream. Processed meat was the sum of bacon, hot dogs, and other processed meats. Total alcohol consumption was the sum of liquor, beer, white wine, and red wine.
Since the distribution of most food items was skewed toward low consumption, we dichotomized intake into 0 servings per day (no intake) versus more than 0 servings per day. Citrus and dairy products were more normally distributed so we categorized women into quintiles of intake for each food item. Our outcome was no intake or intake in the lowest quintile (referred to a “low intake”) compared to those who consumed the item or intake in the four higher quintiles (referred to as “non-low intake”).
We excluded women with more than 70 blanks on the semiquantitative food frequency questionnaire or with implausible total daily energy intake (≤600 or ≥3500 kcal/day) similar to previous studies in the WHS (13,14).
Data analysis
We first excluded 119 women missing information on migraine status at baseline and next we excluded 1387 women without dietary data. This left 38,370 women for our analyses. Using logistic regressions, we calculated odds ratios (OR) and 95% confidence interval (CI) for low intake for each food item for women who experience any migraine and women who experience non-migraine headache compared to those who have no history of headache. We examined low intake because low intake may represent a food avoidance pattern as a result of common beliefs about the food being a migraine trigger. Separate logistic regression models were constructed for each food item. We also performed additional analyses in which we excluded those with no history of headache and just compared women who experience any migraine to women who experience non-migraine headache.
Next, we calculated OR and 95% CI for low intake for each food item for women who experienced migraine within the past year compared to women who reported experiencing migraine in the past but had no attacks in the past year.
Finally, using those women who experienced migraine within the past year, we examined the association between MA status or migraine frequency and low intake of each food item using logistic regressions to evaluate whether the food pattern in these subgroups differed from the food pattern of any migraine.
All analyses adjusted for total caloric intake (continuous), age (continuous), body mass index (BMI) (<25 kg/m2, 25 to <30 kg/m2, ≥30 kg/m2), exercise (rarely/never, <1 time per week, one to three times per week, ≥4 times per week), smoking status (never, past, smoke <15 cigarettes per day, smoke ≥15 cigarettes per day), history of diabetes (yes/no), history of hypertension (yes/no), treatment for high blood pressure (yes/no), history of high cholesterol (yes/no), and treatment for high cholesterol (yes/no). Fewer than 90 women were missing information on any of our covariates and were assigned to the reference category or to the past user category (smoking).
All analyses were performed using SAS 9.1.3 (Cary, NC). We did not use all 131 food items available from the semi-quantitative food frequency questionnaire, but used only those items that have been previously suggested to be migraine triggers. Since all food items were selected a priori, we used a two-tailed p value of <0.05 when determining if our results were statistically significant.
Results
Of the 38,370 women eligible for inclusion in our study, 5573 reported non-migraine headache and 7042 reported any migraine. Of the 7042 women who reported any migraine, 4946 reported experiencing migraine within the past year. Of the women who experienced migraine within the past year, 2972 reported MO in the past year and 1974 reported MA within the past year.
Baseline characteristics of women according to migraine aura status in the Women’s Health Study (n = 38370).
y: years; BMI: body mass index.
Multivariate-adjusted a odds ratio of low intake of food items by migraine and headache status (N = 38084).
Adjusted for age, body mass index, exercise, smoking status, average daily calorie consumption, history of diabetes, history of hypertension, treatment for high blood pressure, history of high cholesterol and treatment for high cholesterol. Percentages indicate the number of people in that particular headache or migraine category who report low intake.
Those who experience non-migraine headache were also more likely to experience low intake of caffeinated coffee (OR = 1.10, 95% CI: 1.03, 1.18) but this was not seen for those who experience any migraine (OR = 1.00, 95% CI: 0.94, 1.07). Additionally, those who experience non-migraine headache were less likely to have low intake of chocolate and hamburger, which suggests that they consume more of those items than participants with no history of headache. Those who experience any migraine were more likely to have low intake of total dairy, cheese, sour cream, and skim or low fat milk but this was not seen for those with non-migraine headache. For other food items, we did not observe differences in food intake for those with non-migraine headache and those with any migraine compared to those with no history of headache.
We also performed additional analyses comparing those who experience any migraine to those who experience only non-migraine headache. Women who experience any migraine were less likely to have low intake of caffeinated coffee (OR = 0.91; 95% CI: 0.84, 0.99) and whole milk (OR = 0.85; 95% CI: 0.77, 0.95) and more likely to have low intake of hamburger (OR = 1.12; 95% CI: 1.02, 1.24) compared to those with non-migraine headache. We did not observe significant results for any other food item (results not shown).
Next, we compared those with active migraine within the last year to those who reported a past history of migraine. Women who reported active migraine were more likely to report low intake of white wine (OR = 1.13; 95% CI: 1.01, 1.27), red wine (OR = 1.24; 95% CI: 1.09, 1.43), and total alcohol consumption (OR = 1.16; 95% CI: 1.04, 1.29) compared to women with a past history of migraine. This suggests that alcohol consumption patterns we observed in Table 2 may be driven by lower alcohol consumption among women who experience active migraine and that women who no longer experience active migraine may no longer avoid the consumption of alcohol. We did not observe significant results for any other food item (results not shown).
Multivariate-adjusted a odds ratio of low intake of food items by migraine aura status (N = 4946).
Adjusted for age, body mass index, exercise, smoking status, average daily calorie consumption, history of diabetes, history of hypertension, treatment for high blood pressure, history of high cholesterol and treatment for high cholesterol.
MO: migraine without aura; MA: migraine with aura; CI: confidence interval. Percentages indicate the number of people in that particular migraine category who report low intake.
Multivariate-adjusted a odds ratio of low intake of food items by migraine frequency (N = 4881).
Adjusted for age, body mass index, exercise, smoking status, average daily calorie consumption, history of diabetes, history of hypertension, treatment for high blood pressure, history of high cholesterol and treatment for high cholesterol.
Discussion
Results of this large cross-sectional study in women suggest that, with the exception of alcohol consumption, the dietary intake pattern of food items previously suggested to be migraine triggers is specific for migraineurs. Secondary analysis observed that the intake of certain food items like chocolate, some processed meats, some dairy products, and wine varies based on aura status and migraine attack frequency. As this pattern was specific for these subgroups, specific food intake patterns appear to be characteristic for MA and for higher frequency of migraine attacks.
One of the hypotheses behind why some food items have been reported to be migraine triggers proposes hidden food allergies as a cause. A previous study in adults showed that testing for immunoglobulin (Ig)E-specific food allergy and restricting diets based on the results of these tests may help to decrease migraine frequency (15). Observational studies and a randomized controlled trial have shown that diet restriction based on IgG antibodies may reduce the frequency of migraine attacks (16–18). However, another study found no difference in IgE and IgG4 titers in responses among adults with dietary migraine and non-dietary migraine (19). Additionally, a small duodenal biopsy study also did not find differences in plasmacyte populations with anti-IgG, anti-IgM and anti-IgE immunohistochemistry between migraineurs with food-induced migraine and those without food-induced migraine (20). These studies refute the idea that food allergies may underlie the association between food items and migraine attacks. A review article suggested that non-allergic mechanisms may also be involved in food-related migraine since various chemicals like ethanol, sodium nitrate, caffeine, phenylethylamine, tyramine, monosodium glutamate, sodium metabisulfite, theobromine, and benzoic acid may trigger migraine attacks (21).
The most commonly studied trigger in previous studies is alcohol consumption, but results are mixed (3–6,22). We observed that both those who experience migraine and those who experience headache were likely to have low intake of alcohol. This suggests that alcohol may be avoided not just by those with migraine. We did not find significant differences in the risk of high intake of alcohol between MA and MO. However, those who experienced a migraine attack within the past year (either MA or MO) were more likely to have low intake of alcohol than those with a past history of migraine attacks. Additionally, we observed evidence that those with frequent migraine attacks were more likely to be in the lowest category of alcohol consumption. This suggests that alcohol consumption may be avoided by those who experience frequent migraine, which is in line with the observation that those with cluster headache consume less alcohol than the general population (23).
In contrast to a previous study (6), we did not observe differences in intake of citrus or beer but did observe some differences in the risk of low intake for chocolate, cheese, ice cream, hot dogs, and processed meats.
Some of the differences in our findings compared to previous studies may be explained by the differences between the design of our study compared to previous studies. Most previous research is retrospective and asks participants to indicate whether the food items have triggered attacks and potentially how consistently the attacks are triggered by the items. In contrast, the present study examines overall intake of food commonly reported as migraine triggers. Another reason for the discrepant findings is the matter of the association under study. If food items indeed trigger migraine, they would be avoided more over time. Thus, depending on the timing and setup of the study, results may indicate an increased or decreased consumption pattern.
In general, we would expect migraineurs overall to have different intakes of foods commonly reported as migraine triggers compared to those without migraine. Our study allows us to directly examine whether dietary food intake patterns do differ between those with migraine or non-migraine headache compared to a headache-free population. Interestingly, migraineurs do not have lower intake of many foods commonly reported to be migraine triggers. This may indicate that these food items may in fact not be migraine triggers or that the association of food in general as a migraine trigger is more complex. Further stratifying by aura status and migraine frequency allows us to determine if intake patterns might differ by migraine features. Differences suggest that those food items may be more commonly avoided by those with a particular type of migraine potentially because of disease severity or because the food item triggers a specific type of migraine attack.
Some limitations should be noted when interpreting our results. First, we did not specifically assess food as a triggering factor in our study and instead examined average food intake patterns. Although we did observe some differences in food intake by migraine aura status and migraine attack frequency, we cannot directly conclude that these food items are directly related to migraine. Second, this is a cross-sectional study with only one assessment of migraine status and food intake. This prevents us from examining the temporal association between migraine status, migraine frequency, and food intake. For example, we cannot strictly determine if migraineurs consume less of certain foods to avoid triggering migraine, if migraineurs avoid certain foods only after a migraine attack, or if low intake of these foods may play a role in migraine or headache development. Future prospective studies with specifically designed food ascertainments in large samples are needed to further understand the association between migraine, headache, and food intake. Third, migraine and headache status was self-reported so some misclassification of migraine status is possible. However, we do not believe that migraine status reporting was related to reporting of food intake (random misclassification). Although our cohort is larger than many previous migraine food trigger studies, power may have been limited in some subgroups of food categories with generally low intake. Finally, as all participants were female health professionals, our results may not generalize to other populations. Despite these limitations, this is one of the largest studies of migraine and food intake to date. We enrolled a population free of major diseases at baseline, which should help to limit the impact of comorbidities on dietary habits and the food assessment was not specific to participants with migraine. While migraine was self-reported, previous studies have shown good agreement with standardized criteria (10). Additionally, we used a validated semiquantitative food frequency questionnaire to assess food intake.
In conclusion, the intake of certain food items, particularly chocolate, wine, processed meats, and some dairy products, differs based on migraine aura status or migraine frequency. As this dietary intake pattern was not seen for overall migraine, our results suggest that these items may be involved in the initiation of particular subtypes of migraine. Future studies need to be performed to further elucidate the mechanisms by which food items may trigger migraine attacks and to determine if food triggers vary based on the characteristics of the migraine attacks.
Clinical implications
With the exception of alcohol consumption, differences in food items commonly suspected to be migraine triggers were not observable in women with non-migraine headache. Intake patterns of foods commonly reported to be migraine triggers differs by migraine frequency and migraine aura status. This suggests that some food factors are involved in the initiation of particular subtypes of migraine.
Footnotes
Funding
The WHS is supported by grants from the National Institutes of Health (HL-043851, HL-080467, HL-099355, and CA-047988).
Conflicts of interest
The authors report no conflicts of interest with regard to the specific matter of this study but list full disclosures for the past two years:
Dr Rist has received funding from the Rose Traveling Fellowship Program in Chronic Disease Epidemiology and Biostatistics from the Harvard School of Public Health.
Dr Buring has received investigator-initiated research funding and support as principal investigator from the US National Institutes of Health and research support for pills and/or packaging from Bayer Heath Care and the Natural Source Vitamin E Association.
Dr Kurth has received investigator-initiated research funding from the US National Institutes of Health and the French National Research Agency (ANR, Agence Nationale pour la Recherche). He has received honoraria from the British Medical Journal and Cephalalgia for editorial services.
All authors had access to the data and a role in writing the manuscript.