Prevalence and Association of Lifestyle and Medical-, Psychiatric-, and Pain-related Comorbidities in Patients With Migraine

A Cross-sectional Study

Jiu-Haw Yin MD; Yu-Kai Lin MD; Chun-Pai Yang MD, PhD; Chih-Sung Liang MD; Jiunn-Tay Lee MD; Meei-Shyuan Lee PhD; Chia-Lin Tsai MD; Guan-Yu Lin MD; Tsung-Han Ho MD; Fu-Chi Yang MD, PhD


Headache. 2021;61(5):715-726. 

In This Article


Previous research from our group has shown that migraine may be associated with a greater risk of several conditions (e.g., anxiety, depression, restless legs syndrome, and insomnia).[5,7,8] The current study aimed to evaluate the association between migraine and a variety of comorbidities using a larger sample. Therefore, the study was partly a secondary analysis, but the nature of the analysis was pre-planned.


No statistical power calculation was conducted prior to the study. The sample size was based on the available data. The current study expanded the dataset using which we had previously published three articles on patients with migraine in association with sleep quality, restless legs syndrome, and depression/anxiety.[5,7,8] This cross-sectional study included a sample of 1753 participants aged between 20 and 60 years old. A total of 496 healthy, headache-free controls were volunteers recruited from the community through advertisements between June 2014 and October 2019, and 1257 were participants undergoing outpatient monitoring in a headache clinic at the Department of Neurology of Tri-Service General Hospital (TSGH) between November 2014 and October 2019. Of these, 409 of the patients with migraine and 179 of the healthy controls were included in our previous studies.[5,7,8] After excluding patients with concomitant primary and secondary headache disorders, those with insufficient data to determine migraine frequency, and those lacking information on clinical characteristics, a total of 1257 patients were included in the final analysis. Of those, 311 participants were diagnosed with chronic migraine (≥15/month) and 946 were diagnosed with episodic migraine (1–14/month). Migraine patients with and without aura were enrolled according to the criteria of the International Headache Society.[2] In this study, all migraine patients were interviewed and diagnosed by the corresponding author (FCY), a board-certified neurologist and headache specialist. The migraine subtype (episodic migraine, chronic migraine, migraine with aura, migraine without aura) was diagnosed according to the International Classification of Headache Disorders, 3rd edition (beta).[2] The 496 healthy, headache-free controls were recruited from the community and directly interviewed by the specialist (FCY). They had not previously been diagnosed with headache disorders, including primary headache disorders such as tension-type headache, migraine, cluster headache, or other secondary headache disorders, except for infrequent episodic tension-type headaches (<1 day/month, average <12 days/year). In addition, they had no history of degenerative diseases of the central nervous system. All participants signed a written informed consent form after receiving a full written and verbal explanation of the study. The study protocol was reviewed and approved by the Institutional Review Board of the TSGH.


After written consent was obtained, participants filled out a detailed questionnaire, which consisted of a series of questions concerning migraine diagnosis, sleep, headache burden, work, lifestyle, education, marital status, medical conditions, pain conditions, and psychiatric conditions. Demographic data, including sex, age, body mass index (BMI), educational level, marital status, employment status, and lifestyle factors (smoking, drinking, and coffee consumption) were self-reported. All participants then participated in a structured interview performed by a physician. Self-reported medical (peptic ulcer disease, mitral valve prolapse, asthma, history of head injury, and glaucoma) and pain conditions (dysmenorrhea) were confirmed by the corresponding author (FCY) according to previous medical records (peptic ulcer disease, mitral valve prolapse, asthma, history of head injury, glaucoma) as well as physical and laboratory examinations (hypertension, diabetes, and thyroid disease). Hypertension was defined as elevated systolic blood pressure ≥140 mm Hg and/or diastolic blood pressure ≥90 mm Hg, or current treatment with antihypertensive therapy.[15] Diabetes was defined as hemoglobin A1C ≥ 6.5% and/or fasting plasma glucose ≥126 mg/dl, or current therapies to treat diabetes.[16] The diagnostic criteria for fibromyalgia were satisfied if the following three conditions were met: (a) Widespread Pain Index ≥ 7 and Symptom Severity Score ≥ 5, or Widespread Pain Index in the range of 3–6 and Symptom Severity Score ≥ 9, (b) symptoms had been present at a similar level for at least 3 months; and (c) the patient did not have a disorder that would otherwise explain the pain.[17] Psychiatric conditions were confirmed by a board-certified psychiatrist. Subjective cognitive complaints were assessed by the subjective memory complaints scale[18] and the Ascertain Dementia 8 questionnaire.[19,20] These methods were adopted from our previous study.[21] For patients, it took roughly 1 h to complete the questionnaire and 5–10 min to complete the interview, while controls required an average of 30 and 5 min for the questionnaire and interview, respectively.

Statistical Analysis

Outcomes were lifestyle-related variables and medical-, psychiatric-, and pain-related comorbidities. The explanatory variables (exposures) were the status of migraine, frequency of migraine (chronic vs. episodic), aura, and sex. The control variables (confounders) were sex, age, BMI, educational level, marital status, and employment status. Sex and aura were considered as effect modifiers. Sex was a confounder when investigating the association between migraine and outcomes in the whole cohort, an effect modifier when the analysis was stratified by sex, and an explanatory variable when comparing the outcomes between male and female participants.

The descriptive statistics of interval variables were expressed as mean and standard deviation, while nominal variables were expressed as frequency and percentage. The characteristics of the groups (e.g., migraine vs. control; chronic vs. episodic migraine) were compared using a chi-square test for nominal variables or an independent sample t-test for interval variables. The normality of interval variables (age, BMI, and educational level) was tested using the Kolmogorov–Smirnov test, and the results demonstrated that these variables did not violate the assumption of normal distribution (data not shown). The proportion of lifestyle factors and the prevalence of conditions comorbid with migraine between the groups were compared using multivariable logistic regression analysis. The logistic model was adjusted for sex, age (as an interval variable), BMI (as an interval variable), educational level (as an interval variable), marital status, and employment status to rule out possible confounding effects. The other nominal covariates were dummy coded. Based on the preplanned protocol, the multivariable logistic regression analyses were stratified by subgroups of aura and sex. All the covariates and predictors were introduced into the model using a forced entry, and the multicollinearity among covariates and predictors was assessed using the variance inflation index. The result showed that the variance inflation index of all the covariates and predictors ranged from 1.035 to 1.172, which indicated no collinearity (data not shown).

All tests were two-tailed, and p < 0.05 was considered statistically significant. The sample size of this study was moderate (number < 2000); therefore, the effects that were statistically significant were usually clinically significant. No adjustment for multiple testing (multiplicity) was made, because some of the analyses performed in this study were exploratory rather than confirmatory. Data analyses were conducted using SPSS 25 (IBM SPSS Inc, Armonk, NY). The proportion of missing data in this study was less than 10% (number of missing cases: 81; total cases: 1834) and were mainly BMI and migraine frequency values. Any individual missing exposure or confounder data was omitted from the analysis.