Given the high prevalence of obesity1 and of migraine2 within the US population, there is increasing interest in the relationship of these 2 conditions.3–5 A longitudinal population-based study demonstrated that obesity is a risk factor for the development of chronic daily headache.3 A different population-based study demonstrated that in persons with episodic migraine (<15 days per month), body mass index (BMI) was associated with migraine attack frequency and severity, although not with migraine prevalence.4 In the same population, obesity was associated with transformed migraine (≥15 days per month), and in this group with attack frequency, severity and disability.5 The mechanisms accounting for the link between obesity and migraine attack frequency and severity remain speculative. Migraine has also been linked to psychiatric disorders most prominently depression and anxiety.6 Longitudinal studies have shown that the relationship between migraine and depression is bidirectional, with migraineurs having a 5-fold greater risk of depression, and patients with depression having a 3-fold higher risk of migraine. This bidirectionality suggests a common neurobiology.7 Anxiety disorders, particularly general anxiety disorder, are also strongly associated with migraine.8,9 Coming full circle, disorders of mood and anxiety are strongly associated with obesity.10–20 Given the complexity of these associations, our study was undertaken to discern, within a population of persons seeking care for headache, (1) the effect of depression in modifying the relationship between BMI and migraine features, (2) the effect of anxiety in modifying the relationship between BMI and migraine features.
Objective.—To discern the effects of depression and anxiety on the migraine–obesity relationship.
Background.—Migraine and obesity are highly prevalent conditions and are both independently linked to psychiatric conditions, mainly depression and anxiety.
Methods.—Data are from an ongoing cross-sectional multicenter study on comorbid conditions in clinic patients seeking treatment for headache. The diagnosis of migraine was determined by the examining physician based on the International Classification of Headache Disorders (ICHD)-II criteria. Participants completed a self-administered questionnaire with information on demographics, headache features, and physician-diagnosed comorbid medical and psychiatric disorders. The questionnaire included scales for measuring current depression (PHQ-9), anxiety (BAI), and headache-related disability (HIT-6).
Results.—A total of 721 migraineurs (88% women) from 8 different headache treatment centers were included in this study (mean age = 42 years, SD = 12). Aura was reported in 45% and chronic headache (≥15 headache days/month) in 35% of the participants. Prevalence of obesity in our population was 30% and only 38% had normal weight. Obesity was more common in men (P= .004), African Americans (P= .026), and in lower education (P= .05) and household income (P=.05) groups. Current depression (PHQ-9 score ≥10) was noted in 42% and current anxiety (BAI score ≥8) in 70% of the obese migraineurs. In ordinal logistic regression, obesity was associated with current depression (odds ratio [OR]= 1.86, 95% confidence interval [CI]: 1.25 to 2.78) and anxiety (OR = 1.58, 95% CI: 1.12 to 2.22). A significant effect of depression on the body mass index (BMI) and headache frequency relationship was noted. Obese migraineurs with depression were more likely to have higher headache frequency (OR = 4.16, 95% CI: 1.92 to 8.99) and headache-related disability (OR = 7.10, 95% CI: 2.69 to 18.77) compared to normal weight migraineurs without depression. Similarly, obese migraineurs with anxiety were more likely to have higher headache frequency (OR = 1.96, 95% CI: 1.07 to 3.61) and headache-related disability (OR = 3.59, 95% CI: 1.64 to 7.86) compared to normal weight migraineurs without depression. Compared to migraineurs with either current depression or anxiety, those with both these conditions were more likely to have higher headache frequency (OR = 3.18, 95% CI: 1.86 to 5.43) and headache disability (OR = 6.13, 95% CI: 2.58 to 14.59).
Conclusion.—Depression and anxiety were common in obese migraineurs. The relationship of obesity with migraine frequency and migraine-related disability is modified by depression and by anxiety, with the strongest effect observed in migraineurs with both depression and anxiety.
body mass index
International Classification of Headache Disorders
Headache Impact Test
Patient Health Questionnaire
Beck Anxiety Inventory
Patient Selection.— This study is from the data of an ongoing multicenter study of the Women's Issues Section research consortium of the American Headache Society, on comorbid medical and psychiatric conditions in individuals with headache. The recruitment of the cross-sectional survey of headache clinic patients began in February 2006 at 8 outpatient headache centers, after each center separately obtained approval from the Institutional Review Boards (IRB). Participants were examined by a headache specialist, who determined the patient's eligibility and obtained informed consent per the IRB protocol. Participation in the study was offered to consecutive women using the following inclusion criteria: women with primary headache disorder as defined by the International Classification of Headache Disorders (ICHD)-II criteria,21 18 years and older, willingness to complete a self-administered electronic questionnaire on a personal digital assistant (PDA), eg, the Palm® handheld device. Exclusion criteria included the following: not physically well enough to complete an electronic questionnaire on a PDA, not literate in English. Patients were given opportunity to ask questions. The physician or the study personnel provided the subjects with verbal instructions and a brief demonstration to complete the survey. After the patient completed the survey, the physician entered the primary headache diagnoses based on the ICHD-II criteria and the average monthly headache frequency over the prior 3 months. The physician also recorded whether the headaches had transformed from episodic to chronic. Table 1 details the migraine diagnoses of the study population. The analysis in this study includes all persons with migraine with aura, migraine without aura, and probable migraine, whether episodic or chronic.
|Diagnosis (International Classification of Headache Disorders [ICHD]-II code)||n (%)|
|Migraine (1.0)||721 (100)|
|Migraine without aura (1.1)||375 (52)|
|Plus chronic migraine (1.5.1)||127 (17.6)|
|Migraine with aura (1.2)||321 (44.5)|
|Plus chronic migraine (1.5.1)||108 (15)|
|Probable migraine (1.6)||25 (3.5)|
|Plus chronic migraine (1.5.1)||16 (2.2)|
Measures.— In keeping with recent recommendation for research in headache and psychiatric comorbidity, we have used ICHD-II diagnostic headache criteria and reliable, validated questionnaires with clinically relevant cut-off scores to classify headache-related disability using Headache Impact Test™ (HIT-6),22 current depression (prior 2 weeks) using patient health questionnaire (PHQ)-9,23 and current anxiety (prior week) using Beck's anxiety inventory (BAI).24
HIT-6.— The HIT-6, a 6-item scale,22 which correlates well with headache severity, has been determined to be reliable and valid in evaluating the impact of headache on health-related quality of life in patients seeking primary25 and headache subspecialty care.26 HIT-6 produces a score ranging from 36 to 78. Severity of impact was graded as “little or no impact” for score ≤49, “some impact” for 50 to 55, “substantial impact” for 56 to 59, and “very severe impact” for a score ≥60.
PHQ-9.— The PHQ-9 is a self-reported diagnostic and severity measure for current (in the prior 2 weeks) depression using criteria from the Diagnostic and Statistical Manual of Mental Disorders (DSM) IV. It has been demonstrated to have superior validity criterion for the diagnosis of “major depressive disorder” compared to 2 other established depression screening questionnaires, and it can detect changes over time, and detect sub threshold depression.23 The 5 levels of depression severity based on the PHQ-9 scores were “minimal” for scores 0 to 4, “mild” for scores 5 to 9, “moderate” for scores 10 to 14, “moderately severe” for scores 15 to 19, and “severe” for scores 20 and above. A study has shown that PHQ-9 scores of ≥15 are associated with 68% sensitivity and 95% specificity in diagnosing “major depressive disorder” using DSM-IV criteria. PHQ-9 scores between 10 and 14 were associated with “other depressive disorder” and scores ≤10 represented “no depressive disorder.”23 Other depressive disorder is defined as a depressive disorder whose criteria encompass fewer symptoms than are required for any specific DSM-IV diagnoses. PHQ-9 was derived from the self-administered PHQ of the PRIME-MD (Pfizer Inc., New York, NY) used for making criteria-based diagnoses of mental disorders.27
BAI.— The BAI assesses the severity of patient anxiety. The questionnaire consists of both physiological and cognitive components of anxiety addressed in the 21 items describing subjective, somatic, or panic-related symptoms.24 A person is asked to rate how much he or she has been bothered by each symptom over the past week on a 4-point scale. Total scores range from 0 to 63 with 4 levels of anxiety: minimal (0 to 7), mild (8 to 15), moderate (16 to 25), and severe (26 to 63).
Data Collection.— The electronic questionnaire was designed with Pendragon® Forms 5.0 computer software (Pendragon Software Corporation, Libertyville, IL) and addressed the following topics: age, height, weight, gender, race, household income, highest educational level attained, age of onset of headaches, impact of headaches on daily life, severity of current depression and anxiety, self-reported physician-diagnosed history of depression and anxiety, and treatment (current and past) for depression and anxiety. There were no personal health identifiers in the questionnaire. The questions were formatted as forced-response items (each item requires a response in order to proceed to the next item) to minimize data entry errors and missing items. The average time taken for completion of the survey was 25 minutes. The physician or the study personnel entered a security code to indicate completion of the survey, and for secure data transmission. Data were uploaded to secure central database by synchronizing the PDAs with an internet-enabled computer using Pendragon® SyncServer computer software (Pendragon Software Corporation, Libertyville, IL). The database was maintained with technical support at the University of Toledo-Health Science Campus, the primary site for the study.
A total of 721 surveys were collected from all 8 centers, which recruited participants during periods ranging from 6 weeks to 9 months. Names of the individual centers and number of surveys contributed from each are as follows: The University of Toledo, Toledo, OH (244 surveys), University of Medicine and Dentistry of New Jersey, Stratford, NJ (205 surveys), Nashville Neuroscience Group, Nashville, TN (117 surveys), Women's Medical Group of Irvine, Irvine, CA (48 surveys), The University of Cincinnati, Cincinnati, OH (37 surveys), Loyola University Medical Center, Chicago, IL (33 surveys), Swedish Medical Center, Seattle, WA (21 surveys), and John Muir Medical Center, Walnut Creek, CA (16 surveys). BMI was not available for 6 individuals as information on either height or weight was missing. HIT-6 score was not available in 11 surveys as responses to 1 or more of the 6 items were missing.
Statistical Analysis.— Based on the height and weight information, BMI was calculated. Individuals were classified as underweight (<18.5 kg/m2), normal weight (18.5 to 24.9 kg/m2), overweight (25 to 29.9 kg/m2), and obese (≥30 kg/m2) according to the International Classification of the World Health Organization. Less than 2% (n = 12) of individuals were underweight, and were not included in the analyses. Categorical data analysis was done using χ2 or Fisher's exact test and evidence of trends in binomial proportions was examined using Cochran–Armitage test statistic. For data with deviation from normality, including analysis of disability, depression, and anxiety scores, nonparametric tests, including Spearman's rho correlation test, to examine the relationship between the scores were utilized. Multivariate ordinal logistic regression analysis for survey data (PROC SURVEYLOGISTIC) was performed to measure the associations of depression and anxiety with BMI, controlling for age, gender, race, education, household income, and antidepressant or anxiolytic medication use. Based on the average monthly headache frequency over the prior 3 months, we generated groups in the order of <5, 5 to 9, 10 to 15, and ≥15 headaches days. For the analysis examining effect modification, PHQ-9 score of ≥10 was considered as a cut-off level for current depression, and BAI score ≥8 as a cut-off level for current anxiety. Ordinal logistic regression models (PROC SURVEYLOGISTIC) were employed to examine the effect of BMI and depression on headache frequency and headache-related disability. An ordinal independent variable was created by crossing the levels of BMI (normal, overweight, obese) with current depression (present, absent), and normal weight with no depression as the reference level. The effect of anxiety and BMI on headache frequency and disability were examined in similar ordinal regression models as used above for depression. For all models, odds ratios (OR) and 95% confidence intervals (CI) were used to measure the strength of the associations. The significance of the ORs was examined using Wald's χ2 test statistic, and ORs of >1.00 indicate an increased likelihood of being in the higher order of the dependent variable. The multicenter unequal sampling frame and the survey design of this study was taken into account in the analysis and adjusted P values and confidence intervals are reported.28 All statistical hypotheses were tested at 0.05 level of significance, and the analyses were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC).
A total of 721 patients, aged 18 years and older (mean = 42 years, SD = 12), from 8 different headache treatment centers, completed the survey. Approximately 10% of those asked did not participate, usually because of acute headache, or lack of time following the visit. No data were collected on those declining participation. Very few persons (<1%) who started the survey did not complete it. All the participants in this study had migraine, and migraine with aura was diagnosed in 45% (n = 321) of the study population. Table 2 summarizes the demographic features, the BMI distribution, and the headache characteristics of the study population. The majority of participants were women (88%) and Caucasian (90%). In the study population, frequency of overweight and obese individuals was similar. A significant proportion of individuals had chronic headache (35%) and severe disability (70%).
|Age, years (mean ± SD)||42 ± 12|
|Male: Female||87 (12): 634 (88)|
|African American||44 (6)|
|Not a high school graduate||22 (3)|
|High school graduate||192 (27)|
|College graduate||253 (35)|
|Post graduate||131 (18)|
|$20,000 to $50,000||174 (24)|
|$50,000 to $100,000||281 (39)|
|Body mass index|
|Underweight (<18.5)||12 (<2)|
|Normal (18.5 to 24.9)||273 (38)|
|Overweight (25 to 29.9)||212 (29)|
|Obese (30 to 35)||218 (30)|
|Headache onset age, years (mean ± SD)||20 ± 10|
|Headache frequency (days per month)|
|10 to 14||123 (17)|
|5 to 9||204 (28)|
|Headache-related disability (HIT-6 score)|
|Very severe (≥60)||495 (70)|
|Substantial (56 to 59)||88 (12)|
|Some (50 to 55)||78 (11)|
|Little or none (≤49)||49 (7)|
There were demographic differences across the BMI groups. Prevalence of overweight (40% vs 29%) and obesity (37% vs 30%) were higher in men compared to women (χ2= 10.83, P= .004). Obesity was more common in African Americans (48%) compared to Caucasians (30%) or any other race (33%) (χ2= 11.0, P= .026). Persons with lower education (P= .05) and household income level (P= .053) had a tendency of being overweight and obese.
Forty-six percent (n = 326) of the study population reported a history of physician-diagnosed depression. Mean age at diagnosis of depression was 28 years. Prevalence of depression was highest (52%) in the obese migraineurs, and lowest in the normal weight group (40%) (Table 3). Prevalence of current depression, based on the PHQ-9 scores, was also significantly higher in the obese group. Major depressive disorder was recorded in 22%, and other (mild to moderate) depressive disorder in 20% of the obese individuals. History of anxiety, reported by 35% of the total population (n = 247), was not significantly associated with BMI. Frequency of current severe anxiety, however, was highest in the obese group (13%). In separate analyses, values of BMI linearly correlated with increasing PHQ-9 scores (r= 0.14, P < .001) and BAI scores (r= 0.13, P < .001). Antidepressant and anxiolytic use did not differ between the BMI groups (Table 3).
|Body Mass Index|
|Normal n (%)||Overweight n (%)||Obese n (%)|
|History of depression*||106 (40)||98 (48)||111 (52)||P= .034|
|Major depressive disorder†||32 (12)||27 (13)||47 (22)||P= .008|
|Other depressive disorder‡||45 (16)||39 (18)||45 (20)|
|None||196 (72)||146 (69)||126 (58)|
|Currently on antidepressants||68 (25)||61 (29)||68 (31)||NS|
|History of anxiety*||90 (34)||84 (41)||64 (31)||P= .08|
|Severe||20 (7)||23 (11)||29 (13)||P < .001|
|Moderate||59 (22)||29 (14)||43 (20)|
|Mild||71 (26)||58 (27)||83 (38)|
|Minimal||123 (45)||102 (48)||63 (29)|
|Currently on anxiolytic medication||46 (17)||39 (18)||33 (15)||NS|
Table 4 illustrates the association of increasing BMI with increasing severity of depression and of anxiety. In ordinal regression models adjusting for the demographic information and antidepressant or anxiety medication use, obesity was associated with a higher degree of depression (OR = 1.86, 95% CI: 1.25 to 2.78) and anxiety (OR = 1.58, 95% CI: 1.12 to 2.22), regardless of headache characteristics.
|Body mass index||Depression (PHQ-9) OR (95% CI)*||Anxiety (BAI) OR (95% CI)†|
|Overweight||1.10 (0.72 to 1.68)||0.83 (0.59 to 1.17)|
|Obese||1.86 (1.25 to 2.78)‡||1.58 (1.12 to 2.22)‡|
Univariate analysis indicated a positive correlation of headache frequency with BMI (r= 0.12, P= .002) and with depression scores (r= 0.29, P < .0001). Significant differences in the median headache frequency were observed between persons with major depression compared to those without depression, at all levels of BMI (Figure). Logistic regression analysis revealed a significant effect of depression on the strength of relationship between BMI and migraine frequency (Table 5). Obese persons with concomitant depression were over 4 times more likely to have a higher frequency compared to normal weight persons without depression. Logistic regression analysis also demonstrated a significant effect of anxiety on the relationship between BMI and migraine frequency (Table 5). Overweight and obese groups with anxiety were more likely to have higher headache frequency, compared to normal weight persons without anxiety. Similarly, in separate regression models, both depression and anxiety modified the relationship between BMI and headache-related disability (Table 5).
|Headache Frequency||Headache-Related Disability|
|Body mass index||Depression*||OR (95% CI)†||OR (95% CI)†|
|Overweight||−||1.24 (0.65 to 2.39)||0.30 (0.07 to 1.27)|
|Obese||−||1.51 (0.85 to 2.68)||0.74 (0.20 to 2.72)|
|Normal||+||2.63 (1.46 to 4.75)‡||4.19 (1.51 to 11.63)‡|
|Overweight||+||3.26 (1.53 to 6.91)‡||6.68 (2.45 to 18.26)§|
|Obese||+||4.16 (1.92 to 8.99)§||7.10 (2.69 to 18.77)§|
|Body mass index||Anxiety*||OR (95% CI)†||OR (95% CI)†|
|Overweight||−||0.91 (0.47 to 1.76)||0.89 (0.42 to 1.87)|
|Obese||−||1.07 (0.58 to 1.97)||1.29 (0.52 to 3.19)|
|Normal||+||1.84 (0.98 to 3.45)||2.24 (1.01 to 5.01)¶|
|Overweight||+||2.17 (1.12 to 4.19)¶||6.05 (2.28 to 16.03)§|
|Obese||+||1.96 (1.07 to 3.61)¶||3.59 (1.64 to 7.86)§|
About 31% of the study population (n = 222) had both current depression and anxiety. The headache frequency and disability scores within each of the BMI groups, were highest in persons with both current depression and anxiety (P < .01 for all comparisons). Compared to migraineurs with either current depression or anxiety, those with both these conditions were more likely to have higher headache frequency (OR = 3.18, 95% CI: 1.86 to 5.43) and headache disability (OR = 6.13, 95% CI: 2.58 to 14.59).
Our results suggest 3 important findings. The first is that, within the population of persons seeking treatment for headache, both depression and anxiety were more common in obese persons. The second is that both depression and anxiety modified the strength of the relationship of obesity with migraine frequency and disability. Lastly the association with higher migraine frequency and migraine-related disability was strongest in those with both depression and anxiety. The link of obesity with depression and anxiety, which we observed in the headache clinic population, is supported by population-based studies.10–16 Data from the Third National Health and Nutrition Examination Survey, for example, showed that obesity was associated with past-month depression, predominantly among persons with severe obesity.12 Although not a universal finding,29 similar links of depression and obesity were found in adolescents,17 US Military active duty personnel,18 and a nationally representative sample of US adults.13 In the longitudinal Northern Finland 1966 Birth Cohort Study the depression–obesity link in adulthood was restricted to women.30 Our subspecialty clinic-based sample included nearly 90% women, and we were unable to detect difference based on gender. The Northern Finland 1966 Birth Cohort Study14 suggests that antidepressant use may be a factor, but our data showed similar proportions of antidepressant and anxiolytic use between the BMI groups. Longitudinal studies have shown obesity in adolescence predisposing to depression in adulthood,14 and depression in adolescence predisposing to obesity in adulthood.15 Our cross-sectional study queried patients on the age of diagnosis of anxiety and depression, but the BMI information was current, so we cannot comment on the temporal association of these conditions.
In our study, worsening headache attack frequency and disability were associated with increasing BMI, similar to the findings reported within the migraine cohort from an ongoing population based study.4 We found, moreover, that the relationships of obesity with higher headache frequency, and with higher headache-related disability were significant only in those with concomitant depression. Although similar associations of obesity with frequency and disability were observed in persons with concomitant anxiety, the associations were stronger in the overweight group. Further analysis revealed high mean anxiety score in the overweight group and this could possibly explain the stronger associations noted with this group.
Our earlier work in a different clinic-based headache population highlighted the strong association of major depression with chronic, disabling headache, but did not assess BMI.31 The recent population-based study4 that examined the relationship of headache and obesity adjusted for depression symptoms, but a formal depression diagnosis was not obtained. Anxiety was not assessed.
Our findings that obesity correlates with depression and anxiety, which in turn, correlate with chronic, disabling migraine, are significant for the potential implications regarding the pathogenetic link of these conditions. It has been hypothesized that obesity may be associated with chronic migraine, perhaps causally, through proinflammatory mechanisms.3–5 There is growing evidence that there are common brain monoamines and peptides (including serotonin, dopamine, norepinephrine, neuropeptide Y, and corticotrophin releasing hormone) influencing depression, anxiety, migraine and regulation of body weight, which may serve as a neurobiological link.15,32 Altered function of hypothalamic–pituitary–adrenal axis has been implicated in the pathophysiology of all 4 of these conditions.33–40
Strengths of this study include the large sample size and the geographically distinct locations of the participating clinics. Analysis was adjusted for important confounders (age, gender, race, education, income) and for the multicenter sampling of this study. Although the results were obtained in a clinic population, the mean age of participants in this study corresponds to the age of peak female headache prevalence in the general population. Diagnoses of headache type and frequency were determined by headache specialists using ICHD-II criteria. Extensive history taking, physical examination, and ancillary testing done in the course of patient care, decreased the likelihood that patients had secondary headache. Evaluation of depression and anxiety was performed using instruments with well-demonstrated validity and reliability. The BAI was specifically designed to reduce the overlap between depression and anxiety scales by measuring anxiety symptoms shared minimally with those of depression. Patients found the PDA technology easy to use, and the rapid paperless transfer of data to a central database eliminated errors in data entry. The advantages of using PDA technology in data collection have been acknowleded,41,42 and compared to telephone responses, electronic survey responses contain fewer random and systematic errors (as demonstrated by notably higher reliability coefficients).43,44
Our study has certain limitations. As in the population-based research examining obesity in migraine,3–5 in our study BMI was calculated based on self-reported height and weight. Research demonstrates that self-reported heights and weights tend to be more accurate with younger adults, and have greater limitations in older adults, ie, >60 years in age.45,46 Our median age was 43 years, and less than 5% were over age 60 years. This is, however, a shortcoming of our study and these data should be collected objectively in the future. A variety of medications may have an impact on headache, weight, and mood, and while we collected data of antidepressants and anxiolytics, we did not collect data on headache preventive or acute medications. Limitations, which may limit generalizability of our findings, stem from clinic-based study sample. In our population, the female to male ratio of 7:1 is considerably higher than the 4:1 reported in the migraine–obesity population-based study.4 The proportion of migraineurs diagnosed with aura (45%) in our study was higher than reported in population studies (36%).47
In summary, depression and anxiety were common in obese migraineurs. The relationship of obesity with migraine frequency and migraine-related disability is modified by depression and by anxiety, with the strongest effect observed in migraineurs with both depression and anxiety. These findings further refine knowledge of clinical expressions of migraine and expand hypotheses of pathogenetic mechanisms of migraine and its chronification.
Acknowledgments: This work was in part supported with a section grant from the American Headache Society. The authors thank Brian Szabo, in the Center for Creative Instruction, University of Toledo-Health Science Campus, for technical support and database management. The authors also thank Gina Dawe and Neha Pandya for their assistance in data collection at the University of Toledo-Health Science Campus.
Conflict of Interest: None