Abstract
Background
The association between body mass index (BMI) and migraine in adults has been well established. However, studies in children and adolescents are inconclusive. We aimed to study the association between BMI and migraine using a national dataset that comprises the electronic medical records of more than two million adolescents.
Methods
This study included all Israeli adolescents (57.7% males, 42.3% females; mean age 17 years) who were medically assessed before mandatory military service during 1990-2020. As part of the pre-recruitment medical assessment, all the adolescents were screened for migraine and their height and weight were measured. Diagnoses of migraine were confirmed by board-certified neurologists. Prevalences and odds ratios (ORs) for migraine were computed across BMI subgroups. Spline models were applied.
Results
A total of 2,094,862 adolescents were included, of whom 57,385 (2.8%) had active migraine. Among males, the adjusted ORs for migraine were 1.11 (95% confidence interval, 1.06-1.16), 1.13 (1.08-1.17), and 1.24 (1.19-1.30), for the underweight, overweight, and obesity subgroups, respectively, compared to the reference group of low-normal BMI (5th-49th percentile). Among females, the respective adjusted ORs were 1.12 (1.05–1.19), 1.23 (1.19–1.28), and 1.38 (1.31–1.46). Results persisted in sensitivity analyses accounting for other medical and psychiatric comorbidities and parental history of migraine. Spline models demonstrated a J-shaped relation between BMI and migraine.
Conclusions
Both adolescent obesity and underweight were associated with migraine in a sex-dependent manner. This association peaked in female adolescents with overweight and obesity.
Keywords
Introduction
Adolescent obesity is globally increasing and is associated with numerous adverse sequela (1). The association between obesity and migraine in adults has been well established and was found to be largely dependent on age, sex, and race (2,3). However, studies in children and adolescents have been inconclusive. Several studies in pediatric populations have shown an association between migraine and obesity (4–7), while others have not (8,9). Moreover, risk estimates varied widely across studies due to small sample sizes, heterogenous age groups, different weight categorizations, and differences in the case definition for migraine. Additionally, most studies lacked data regarding comorbidities or family history of migraine, which may confound the association between migraine and body mass index (BMI).
We examined the association between board-certified neurologist diagnoses of migraine and measured adolescent BMI in a nationwide, population-based study of over two million adolescents.
Methods
Study population
This retrospective cross-sectional study included all Israeli male and female adolescents who underwent a medical evaluation in the year before their mandatory military service. One year before mandatory military service, Israeli adolescents undergo a thorough medical assessment by physicians to deem their fitness to serve. This includes a review of medical records, a medical interview, and a physical examination (including measurement of height and weight). In general, any significant medical history or abnormal finding in a physical exam necessitates further workup to determine fitness for military service, such as additional testing or consultation with board-certified specialists. Upon completion of the medical assessment, the conscripts' diagnoses are assigned numerical codes, which are then documented in a central database. During 1990–2020, 2,169,089 Israeli adolescents were assessed at ages 16 to 19 years. Examinees with missing height and weight data (n = 74,227) were excluded, the final study sample comprised 2,094,862 adolescents (Figure 1). This sample is considered nationally representative (10). Notably, Jewish ultra-orthodox males, Jewish orthodox females and Arab people are exempt from military service, and therefore underrepresented in this study. The institutional review board of the Israel Defense Forces Medical Corps approved this study and waived the requirement for informed consent based on strict maintenance of anonymity of the individuals included.
Study design and study sample buildup. Examinees were excluded if data on BMI were missing. BMI was classified according to age and sex-matched percentiles based on criteria of the US Center for Disease Control and Prevention: BMI < 5th percentile (underweight), 5th ≤ BMI < 49th percentiles (low-normal BMI), 50th ≤ BMI < 85th percentiles (high-normal BMI), 85th ≤ BMI < 94th percentiles (overweight), and 95th percentile ≤ BMI (obesity).
Definition of migraine
As part of the pre-recruitment medical assessment, all adolescents were routinely questioned about the occurrence and frequency of headaches over the past year. If migraine was reported in the medical records provided by the primary physician or a history of frequent headaches in the past year was elicited during the medical interview, the adolescent was referred to a neurologist. All migraine diagnoses were confirmed by board-certified neurologists according to the International Classification of Headache Disorders (11), as previously described (12). Adolescents who were diagnosed with migraine and experienced at least one migraine attack per month in the past year were assigned a diagnostic code that indicated active migraine. Those who experienced at least one migraine attack per week were assigned a specific diagnostic code that indicated a high frequency of migraine attacks. These diagnostic codes were used to deem to adolescents fitness for service. They allign with the International Classification of Diseases (ICD) code for migraine (ICD11/8A80), but also include the frequency of migraine attacks. Adolescents who solely had a history of migraine or experienced less than one migraine attack per month did not receive the above diagnostic codes, because it did not result in significant functional impairment or altered their fitness for military duty. Consequently, these individuals could not be identified in the current study were included in the control group.
Data collection and study variables
Height and weight were measured by medics using a beam balance and stadiometer, with examinees barefoot and wearing only light clothing. Adolescent BMI was calculated as weight in kilograms divided by the square of the height in meters. BMI was classified according to age and sex-matched percentiles based on criteria of the US Center for Disease Control and Prevention, validated for Israeli adolescents (10): BMI < 5th percentile (underweight), 5th ≤ BMI < 49th percentiles (low-normal BMI), 50th ≤ BMI < 85th percentiles (high-normal BMI), 85th ≤ BMI < 94th percentiles (overweight), and 95th percentile ≤ BMI (obesity). Sociodemographic data were obtained based on routine reports on examinees from various governmental ministries. Residential socio-economic status was reported from the Israeli Central Statistical Bureau on a 1–10 scale that was divided into three ordinal groups (low, medium, and high), as previously described (13). Education level was dichotomized by a cutoff of 11 years based on data received from the Israeli Ministry of Education, as this represents the maximum school instruction that could be attained at the time of assessment (14). Cognitive performance was assessed by general intelligence tests, which were shown to correlate over 85% with the Wechsler Adult Intelligence Scale (15). Cognitive performance was classified by a general intelligence score into three groups, low (−1 > SD), medium (−1 to 1 SD), and high (>1 SD). For this study, adolescents were categorized as having unimpaired health if they did not have any medical condition requiring chronic medical treatment or follow-up, and did not have any history of cancer, major surgery, or psychiatric comorbidity. In earlier studies evaluating the association between BMI and several other diseases, we have focused on this group of healthy adolescents in order to minimize potential residual confounding posed by coexisting medical conditions (10,14,16).
Statistical analysis
Categorical variables were presented as n (%) and compared using the chi-squared test. Logistic regression models were used to determine odds ratios (ORs) for migraine, with the 5th–49th BMI percentile group as the reference. Analyses were stratified by sex given the interaction between BMI categories, sex, and migraine. Models were unadjusted or adjusted to pre-specified variables that we have previously included in studies that examined the association between BMI and several other diseases (10,13,14,17). These variables included residential socio-economic status, education, and cognitive performance. Adolescents with any missing sociodemographic data (n= 23,418, 1.1%) were omitted from the multivariable analysis. All the tests used were two-tailed, with P-values <0.05 considered to be statistically significant. We additionally investigated the association between BMI and migraine using natural spline models. These piecewise regression models allowed us to explore the association without pre-assumption regarding linearity or homogeneity within certain BMI groups (such as normal BMI range), which could potentially average the actual effect. (18). Splines with four degrees of freedom were used to assess and visualize odds functions and 95% confidence intervals. Spline models were adjusted as in the primary analysis and included all the adolescents with BMI values between 15 and 35 kg/m2, with the median BMI of the low-normal BMI group used as a reference. All data analyses were performed using R version 4.0.2 (R Core Team, Vienna, Austria).
We calculated the population attributable fraction (PAF) for males and females, with an unverified assumption of causality of adolescent overweight or obesity with migraine. PAF for migraine was calculated for overweight and obesity (≥85th BMI percentile) as follows: PAF = Pe × (OR − 1)/(Pe × (OR − 1) + 1) × 100, in which OR is the unadjusted OR of overweight and obesity (with the 5th-84th BMI percentile as the reference) and Pe is the prevalence of overweight and obesity in the study period (1990-2020), or as projected based on the mean prevalence of overweight and obesity during 2018-2020.
Several sub-analyses were conducted. (i) To minimize residual confounding by coexisting morbidities, the analysis was restricted to examinees with unimpaired health. (ii) To facilitate comparison of our findings with other studies, we analyzed the association using standard categories of BMI. This was done either according to US CDC percentiles (4,6): BMI < 5th percentile (underweight), 5th ≤ BMI < 85th percentiles (normal weight), 85th ≤ BMI < 94th percentiles (overweight), and 95th percentile ≤ BMI (obesity), or (iii) according to the World Health Organization (WHO) categories for adults (2,3); underweight, BMI < 18.5 kg/m2, 18.5 kg/m2 ≤ BMI < 25 kg/m2 (normal weight), 25 kg/m2 ≤ BMI < 30 kg/m2 (overweight), and BMI ≥ 30 kg/m2 (obesity). (iv) Both migraine and obesity are complex diseases influenced by genetic and environmental factors (19). To assess the contribution of parental history of migraine to the investigated association, we limited the analysis to adolescents for whom both parents' medical history as adolescents could be identified in earlier years of the database. In this analysis we added to the multivariable model any parental history of migraine as a dichotomic variable. (v) We conducted a subgroup analysis based on the world region of the adolescent’s origin, according to the birthplace of the adolescent’s father or grandfather (if the father was born in Israel): Native Israeli, Western world countries (including Europe, the United States, Canada, South Africa, Australia, and New Zealand), former United Soviet Socialist Republic countries, North Africa, and West Asia. These regions were chosen because they encompass the main origins from which Jewish immigrants arrived in Israel, as previously described (10,16,17) (vi) We conducted a sensitivity analysis in which the outcome was migraine of at least one attack per week.
Results
Baseline characteristics of the 1,208,149 males and 886,713 females included in the study are presented in Table 1. The mean age at the time of medical evaluation was 17.3 ± 0.5 years. In total, 57,385 (2.8%) persons had migraine: 31,533 males and 25,852 females. There was an interaction between migraine, BMI, and sex (pinteraction = 4.8 × 10−8).
Characteristics of the study cohort at baseline, according to BMI groups.
*The BMI range refers to a mean age of 17.3 years. Note that for each of the 48 months between ages 16 and 19 years, different sex- and age-specific BMI ranges determine the US Centers for Disease Control and Prevention percentiles.
Full education: either a higher school student at the time of the examination or completed 12 years of formal education.
Among females, the unadjusted ORs for migraine were 1.15 (95% CI, 1.09–1.22), 1.26 (95% CI, 1.21–1.31), and 1.44 (95% CI, 1.37–1.52) for the underweight, overweight, and obesity subgroups, compared to the reference group of low-normal (5th-49th percentile) BMI. These associations persisted in a multivariable model adjusted for residential socioeconomic status, education status, and cognitive performance. The adjusted ORs were 1.12 (95% CI 1.05–1.19), 1.23 (95% CI, 1.19–1.28), and 1.38 (95% CI, 1.31–1.46), respectively (Figure 2). Among males, the unadjusted ORs for migraine were 1.11 (95% CI, 1.06-1.16), 1.12 (95% CI, 1.08-1.17), and 1.22 (95% CI, 1.17-1.27) for the underweight, overweight, and obesity subgroups, respectively, compared to the reference group of low-normal (5th-49th percentile) BMI. These point estimates overall persisted in a multivariable analysis (Figure 2) and in a sensitivity analysis limited to individuals with unimpaired health (Online Supplementary Figure S1).
(a) Odds ratio for migraine and (b) Odds ratio for migraine with at least one attack per week.
Similar findings for both sexes were observed in other BMI metrics (Online Supplementary Figures S2 and S3). In a sensitivity analysis that included 702,778 adolescents (33.5%) for whom pre-recruitment medical records of both parents as adolescents were available (Online Supplementary Figure S4), the association between migraine and BMI remained stable after adjusting for the history of parental migraine and sociodemographic variables (Online Supplementary Figure S5). Additionally, stratification of the adolescents to five world regions, according to their fathers’ or grandfathers’ birthplace yielded ORs that were consistent with the main findings (Online Supplementary Figures S6 and S7).
Results of the spline models confirmed the main findings and demonstrated a J-shaped relation with a monotonic increase in ORs for migraine from BMI above 20 kg/m2; the relation was more pronounced in females than males (Figure 3).
Natural spline models are shown for the odds ratios and 95% confidence intervals of the associations of BMI with migraine in males and females. Spline models were adjusted as in the primary analysis and included all the adolescents with BMI values between 15 and 35 kg/m2, with the median BMI of the low-normal (5th–49th percentile) BMI group used as a reference.
When the outcome was migraine of at least one attack per week, the adjusted ORs among females were 1.33 (95% CI, 1.16–1.44), 1.16 (95% CI, 1.06–1.27), and 1.44 (95% CI, 1.27–1.62) for those with underweight, overweight, and obesity, respectively, compared to adolescents with low-normal (5th–49th percentile) BMI as the reference group. Among males, this association between migraine and BMI was ablated (Figure 2).
The PAF of adolescent overweight or obesity for migraine was 2.5% (2.0–3.2%) in males and 4.7% (4.1–5.9%) in females, based on mean prevalences of 17.4% in males and 15.6% in females during the study period. The projected PAF of adolescent overweight or obesity for migraine was 3.3% (2.6–4.1%) in males and 6.5% (5.8–7.3%) in females, based on prevalences of 23.0% in males and 22.0% in females during 2018–2020.
Discussion
We examined the association between migraine and measured adolescent BMI in a nationwide, population-based study of over two million adolescents. We found associations of both obesity and being underweight with migraine. The associations were stronger among females than males; the odds were increased by 38% for adolescent females with obesity compared to females with low-normal BMI. Being underweight was linked to mildly increased odds of migraine in both sexes. These associations were materially unchanged following adjustments for sociodemographic confounders.
Data from various populations around the world show striking differences in prevalence rates of migraine in children and adolescents (4–24%), predominantly due to differences in migraine definition across studies and population characteristics, such as geography, ethnicity, culture, and methodological differences across studies (20–22). The migraine prevalence of 2.8% reported herein was markedly lower due to the stringent case definition of migraine used in our study. Comparatively with other population studies in which migraine was diagnosed based on self-report/questionnaires, all migraine cases in the current study were confirmed by neurologists, and only examinees who had at least one attack per month in the preceding year were classified as having migraine. The available data on the association between BMI and migraine in childhood and adolescence is inconclusive. Risk estimates varied widely between studies due to small sample sizes, heterogeneous age groups, and major differences in the case definition for migraine (4–9). In a study from Norway that included 5847 students aged 13–18 years, 392 of whom were diagnosed with migraine according to questionnaires and subsequent structured interviews with nurses, the age and sex-adjusted OR for migraine among adolescents who were overweight was 1.6 (95% CI 1.4–2.2) (6). A study of 273 patients aged 9–17 years, who were evaluated in a tertiary care center in Israel reported an almost fourfold increased risk of headaches among overweight females, and no such association in males (4). On the contrary, in a study that included 245 children and adolescents with migraine treated in a specialized headache clinic, no association was observed between being overweight and migraine (9). Two multi-center studies from the US also reported a null association between obesity and migraine (5,8). Our findings in adolescents are comparable with the current evidence in adults (2,3). In a recent meta-analysis that included 41 studies, encompassing data from 792,500 individuals, the age and sex-adjusted pooled risk estimates of migraine in those with obesity and underweight were increased by 28% (95% CI 15–43%) and 21% (95% CI 9–34%), respectively, compared with those of normal weight (2). The association was largely dependent on age, sex, and race, which is in line with the associations reported here. Results of the spline models echoed the dose-response relation between BMI and migraine found in the previous meta-analysis.
In the current study, BMI was associated with active migraine of at least one migraine attack per month. When the outcome was at least one migraine attack per week, the association persisted among women but not among men. Current data on the association between BMI and frequency of migraine attacks are limited. In a large population study of adults, overweight and obesity were associated with higher migraine attack frequency (23). However, no association was observed between underweight and migraine attack frequency. In a longitudinal study that aimed to identify risk factors for the progression from episodic migraine to chronic migraine, BMI was not associated with increased migraine attack frequency or progression from episodic to chronic migraine (24). However, these studies did not stratify by sex, and BMI was analyzed as a continuous variable, which may ablate the sex-specific associations that we observed among those with underweight and obesity.
While our study is non-mechanistic in nature, several potential mechanisms may underly the association between abnormal weight and migraine. Obesity is a complex disease, reflecting the interactions of environmental and genetic variables, with and increasingly recognized involvement of the central nervous system (25). Excessive adipose tissue stimulates the release of inflammatory mediators and reduces the production of adiponectin, predisposing to a pro-inflammatory state. Inflammatory peptides can sensitize the central nervous system, causing permanent damage to the periaqueductal grey matter, which is an area related to the pathophysiology of migraine (26). Dysregulation of orexin-A, orexin-B, and calcitonin gene-related peptides, secondary to obesity, was also linked to neurogenic inflammation and consequently to migraine attacks (19). Differences between males and females in the association of BMI with migraine could be attributed to ovarian hormone modulation in persons of reproductive age, and to differences in fat tissue distribution among adolescents (27). Alternatively, migraine could affect obesity status. Central and peripheral pathways regulating feeding and adipose tissue function overlap extensively with pathways implicated in migraine pathophysiology (28). Regardless of the exact mechanism, our findings are of public health importance in highlighting the association between two prevalent medical conditions, especially since adolescent obesity is rising (1), and could be a potential modifiable risk factor for migraine (29,30). We also found that being underweight was associated with migraine in both sexes. The association was comparable to that observed in previous meta-analyses in adults. However, the pathophysiology behind the association between being underweight and migraine is less understood.
This study has limitations. First, due to the observational nature of our study, a causal relationship between BMI and migraine cannot be inferred. Second, our study captured only active migraine cases with at least one migraine attack per month during the year prior to the medical assessment. However, our findings underline the important association between BMI and migraine diagnoses accompanied by functional disability. Third, we lacked clinical data on migraine, such as the occurrence of aura and pharmacological treatment. Fourth, we did not have longitudinal data on the cumulative exposure to obesity. Fifth, we lacked adiposity measures other than BMI, such as body composition, waist circumference, and waist-to-hip ratio. Nevertheless, BMI is considered the preferable method for screening according to the US Preventive Services Task Force (31). Lastly, we had no information on lifestyle factors such as diet, smoking, physical activity, and sleeping habits. The strengths of our study include a large national study sample of adolescents of a heterogenous ancestry, representing a variety of genetic origins (16,32); systematic data collection of anthropometric measures over a narrow age range; confirmed diagnoses of migraine by neurologists, with a distinction of migraine that is functionally disabling; and strict control of health condition at age 17 years.
In conclusion, both obesity and underweight were associated with migraine in otherwise apparently healthy adolescents, in a sex-dependent manner, with accentuation of the findings regarding migraine of at least one attack per week. Future studies are needed to determine whether maintaining healthy weight during childhood and adolescence could alleviate migraine symptoms and prevent disability.
Article highlights
Among two million adolescents, body-mass index was significantly associated with migraine in J-shaped manner, with the size effect being greater for females who were overweight and obese. The association between body-mass index and migraine was comparable to that observed in previous meta-analyses in adults. Among female adolescents, being overweight and obese was significantly associated with higher migraine attack frequency.
Supplemental Material
sj-pdf-1-cep-10.1177_03331024231209309 - Supplemental material for Body mass index and migraine in adolescence: A nationwide study
Supplemental material, sj-pdf-1-cep-10.1177_03331024231209309 for Body mass index and migraine in adolescence: A nationwide study by Yair Zloof, Avishai M. Tsur, Maya Simchoni, Estela Derazne, Dorit Tzur, Asaf Honig, Maya Braun, Esther Ganelin-Cohen, Gil Amarilyo, Orit Pinhas-Hamiel, Arnon Afek and Gilad Twig in Cephalalgia
Footnotes
Authorship contribution statement
YZ was responsible for conceptualisation, statistical analysis and writing original draft. AMT and MS statistically analyzed and interpreted the data, and critically revised the manuscript. ED and DT curated the data and critically revised the manuscript. AH, MB, ECG, GA, OPH, and AA contributed to the discussion and critically revised the manuscript. GT designed and supervised the study, analyzed and interpreted the data.
Acknowledgment
We dedicate this manuscript to the memory of Dr. Yair Zloof, a brilliant physician and gifted researcher, who was killed in recent war in Israel.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
The current data set is subject to military restrictions, and therefore its availability is limited. Data request or queries may be addressed to the corresponding author.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
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.
