Database source and study population
This nationwide population-based cohort study retrieved data from the Korean National Health Insurance Service (K-NHIS) Database. A detailed summary of the K-NHIS database has been reported previously [19, 20]. In short, the K-NHIS is the only public insurance system covering almost the entire Korean population. As of 2015, 99% of the total population of Korea was covered by the K-NHIS [21]. Therefore, this database is representative of the entire population of South Korea, which consists exclusively of East Asian people as Koreans are known to be of a single race and ethnicity [22]. Furthermore, the K-NHIS offers biannual health checkups, based on age, for all subscribers and provides demographic information, results of health checkups, structured questionnaires, and history of diagnoses according to the International Statistical Classification of Diseases Related Health Problems, Tenth Revision (ICD-10). Clinical information included body mass index (BMI); basic blood tests; and social habits, such as alcohol consumption, smoking status, and physical activities. Income levels were also collected since health insurance premiums are charged differently as per income levels [23]. We also collected mortality data from Statistics Korea as previously described [24]. The applicability and strengths of the K-NHIS database compared to other databases that include information of various blood tests, demographic findings, and socioeconomic status, have been well described in other studies [25,26,27,28,29]. This distinguishes the current study from other database studies. Use of the K-NHIS database is permitted if study protocols are approved by both the government’s official review committee and review board of the medical institution. The institutional review board of Hallym University Dongtan Sacred Heart Hospital approved this study and waived the requirement for informed consent because of its retrospective nature and anonymized analysis (IRB No. HDT 2022–05-019). This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
Study population
The K-NHIS cohort used for this study consisted of people who underwent a nationwide health checkup between January 1, 2009 and December 31, 2012. Patients who had been diagnosed previously with T2DM and those who were newly diagnosed during health checkups were included in the study. We excluded patients who (1) were < 20 years of age, (2) had a past medical history of MI or stroke, or (3) had claims of migraine diagnosis after health checkups. Patients with missing data during health checkups were excluded. In total, 2,229,598 patients with T2DM were included in the main analysis. The patients were followed upon from January 1, 2009 to December 2018, and the data were analyzed between August 1, 2020 and October 31, 2020.
Definitions of type 2 diabetes mellitus and migraine
T2DM was defined as follows: (1) having at least one claim per year for a prescription of antidiabetic medication under ICD-10-CM codes (i.e., E11–14, which excludes patients with type 1 diabetes) from the insurance claims data; or (2) not under ICD-10 E11-14 codes neither a prescription of oral hypoglycemic agents (OHA) or insulin, but having a fasting blood glucose (FBG) ≥ 126 mg/dL in the general health checkups. OHA included metformin, sulfonylureas, meglitinides, dipeptidyl peptidase 4 (DPP-4) inhibitors, thiazolidinediones, and α-glucosidase inhibitors. This definition of T2DM has been reported in previous studies based on the K-NHIS database [26, 30, 31]. Medication status was assessed at the baseline year, and duration of T2DM was measured from the first diagnosis to the index date. We defined severe diabetes as having: more than 5 years of T2DM, using insulin, and using 3 or more types of OHA.
We identified patients ≥ 20 years of age with migraine using the diagnostic ICD-10 code G43. Migraine with aura was defined with ICD-10 code G43.1 at least once within three years before the health check-up. Migraine without aura was defined with ICD-10 code G43.X at least once within three years before the health check-up. We set a 1-year lag-period to prevent immortal time bias for the outcomes, and a total of 32,764 patients were excluded (Fig. 1).
Definitions of covariates and outcome variables
Demographic data, including age, sex, body weight, height, waist circumference, and previous history of vascular risk factors such as hypertension and dyslipidemia, were obtained. We also collected blood pressure and laboratory data, including liver function tests (aspartate aminotransferase [AST], alanine aminotransferase [ALT], gamma-glutamyl transferase [rGTP]), kidney function tests (glomerular filtration rate [GFR]), and lipid profiles (total cholesterol, high-density lipoprotein [HDL], low-density lipoprotein [LDL], and triglyceride). Data on lifestyle behaviors, including alcohol consumption, physical activity, and smoking habits, were collected using a self-reported questionnaire. Specifically, the average alcohol intake per day (g/day) was analyzed to evaluate alcohol consumption. Regular exercise was regarded as high-intensity physical activity (extreme shortness of breath for > 20 min per session, ≥ 3 days per week) and/or moderate-intensity physical activity (shortness of breath for > 30 min per session, ≥ 5 days per week) [32, 33]. Low-income level was defined as the composite of the lowest quartile of yearly income, as previously reported [30].
We defined newly diagnosed cardiovascular events as MI, IS, CVD, and all-cause death as study endpoints. These endpoints were defined based on ICD-10-CM claim codes with additional conditions. MI was defined as ≥ 1 claim under ICD-10 codes I21 or I22 during hospitalization or ≥ 2 claims under those codes. IS was defined as the presence of ICD-10 codes I63 or I64 during hospitalization with claims for brain imaging (magnetic resonance imaging or computed tomography) according to previous reports [34,35,36]. Data on the date of death were obtained from the K-NHIS database. The final follow-up was conducted in December 2018.
Statistical analysis
Data are presented as means ± standard deviation (SD) or medians with interquartile ranges for continuous variables and numbers and frequencies for categorical variables. For continuous variables, student’s t-tests or Mann–Whitney U tests were used, as appropriate. For categorical variables, chi-square tests or Fisher’s exact tests were used, as appropriate. Multivariate Cox proportional hazard regression analyses were used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the association between migraine and cardiovascular outcomes (MI, IS, CVD, and all-cause death) in patients with T2DM. Four models were used for multivariate analysis—(1) adjusted for age and sex; (2) model 1 plus additionally adjusted for smoking status, alcohol consumption status, regular physical activity, and household income level; (3) model 2 plus additionally adjusted for history of hypertension, dyslipidemia, and BMI (calculated as weight in kilograms divided by height in meters squared); and (4) model 3 plus additionally adjusted for duration of T2DM and use of insulin and/or OHA. All analyses were 2-sided, and a P-value ≤ 0.05 was considered statistically significant. All statistical analyses were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC, USA).
Data and resource availability
All raw data were accessible from the designated terminals approved by the K-NHIS. Upon reasonable request, data are available through approval and oversight by the K-NHIS.
