Study Design, Setting, and Populations
We used a population-based retrospective observational cohort study established by the National Health Insurance Service (NHIS) of Korea between 2009 and 2012, traced to 2017. NHIS is a compulsory single-payer national healthcare coverage system and has covered approximately 50 million population of South Korea since 1989. Detailed information about the NHIS claim database has been made available [21]. Briefly, the NHIS database has access to the medical records of the entire Korean population and they undergo general health screenings every 2 years. The NHIS claims database contains information on diagnoses (International Classification of Diseases, 10th revision (ICD-10) code), prescription records, and demographics. The NHIS claim database includes major risk factor variables, but lacks many medical records and laboratory tests such as body size, brachial blood pressure and fasting blood chemistry data. Therefore, we combined the NHIS claims database with the health screening examination dataset.
From the NHIS claims database, we extracted 2,703,713 people among those who had undergone health check-ups, including those who had a blood sugar level of 126 or higher at their health screening examination or those who were prescribed anti-diabetes medication and were given a diagnosis of T2D (ICD-10 code, E11 to E14) between January 2009 and December 2012. The exclusion criteria are as follows: age < 40, no anti-diabetic medication within 1 yr of the examination, those who had developed dementia before the examination or developed dementia within 1 yr of the examination, and those with missing data (See Figure 1). The study cohort was followed up to 2017.
The study therapies were defined as follows: Met, SU, TZD, DPP-4i, Meglitinide (Megl), α-glucosidase inhibitor (AGI) and insulin because the use of these medication classes was common at the time of the baseline of the study (in contrast to other drug classes such as SGLT-2 inhibitors and GLP-1 agonists). Among the 1,578,322 people who used anti-diabetic medications, there were many differences in baseline characteristics for each group with only an insulin regimen, oral monotherapy, oral dual therapy and triple therapy (Suppl Table 2). The association between anti-diabetic medications and dementia was investigated by stratifying with/without insulin treatment due to different baseline characteristics. Based on Korean guideline of diabetes management, Met has been usually used as first line pharmacotherapy to treat people with T2D. After preliminary analysis, we focused on the association between oral dual therapy and various types of dementia by comparing Met+SU group as reference group.
Study Outcomes
The primary outcome was the first diagnosis of all-cause dementia (ICD-10 codes, F00~03, F107, G30, G310, G311, G318 and a prescription of anti-dementia medication), AD (ICD-10 codes, F00, G30 and a prescription of anti-dementia medication), or VaD (ICD-10 codes, F01 and a prescription of anti-dementia medication).
Covariates
The study covariates included age at index date, gender, duration of diabetes, income, presence of hypertension, presence of dyslipidaemia, smoking, drinking, presence of regular exercise, body mass index, systolic and diastolic blood pressure, total cholesterol, fasting glucose, and estimated glomerular filtration rate.
Statistical analysis
Data are presented as the mean ± SD for normally distributed continuous variables and as proportions for categorical variables. ANOVA was used to analyse the continuous variables, and the differences between nominal variables were compared with the chi-square test. Annual event rates are described as the number of events per 1000 person-years (PY). Hazard ratios (HR) and their corresponding 95% confidence intervals (CI) were calculated using Cox proportional hazard models for investigating the association between dementia and diabetes medication, after adjusting for age, sex, income, hypertension, dyslipidaemia, DM duration, smoking, drinking, exercise, BMI (body mass index), glucose and the estimated GFR (glomerular filtration rate). Subgroup analyses differentiated by multiple cardiovascular risk factors were subsequently performed. All P-values were two-sided, and a value <0.05 was considered statistically significant. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).
Ethical Approval
All subjects provided written informed consent, and this study protocol was approved by the Institutional Review Board of the Sungkyunkwn University School of Medicine/Kangbuk Samsung Hospital (IRB No. 2018-01-036) and the Institutional Review Board of of the NHIS (number: NHIS-2019-1-218) . Informed consent was waived by the IRB.