We conducted a case-control study within two groups of 40–80-years old subjects, one with and without type 2 diabetes in Ulaanbaatar. Sample size calculation was based on the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) guidelines for reporting case-control studies [10]. We estimated, a priori, that the prevalence of dementia would be 15% in nondiabetic subjects and 30% in diabetic subjects. We used a significance level of 5% and a power of 80%. Finally, the sample size was 131 subjects in each group. The patients with type 2 diabetes were randomly selected during their outpatient routine clinical examinations in the Mongolia-Japan Teaching Hospital from 1st of June to 30th of August 2020. The inclusion criterion was if they were currently aged 40 years or over. The control group consisted of individuals with a normal glucose tolerance level confirmed by an endocrinologist and matched 1 : 1 by age and gender. The following pre-existing conditions constituted exclusion criteria: stroke, neurodegenerative disease, multiple sclerosis, cranial trauma, cancer, paraneoplastic cancer, neurosurgery, alcohol-related depression, Parkinson’s disease, and depression determined by Tsung test $>$70 scores [11].

According to the Helsinki Declaration, the study was conducted, and the medical ethical committee approved it of the Mongolian National University of Medical Sciences (METc No13-03/1A). Furthermore, all participants provided their written informed consent.

Cognitive function assessment: Cognitive function was evaluated with the Mini-Mental State Examination (MMSE) [12]. The MMSE consists of 11 cognitive function domains, including: orientation (2), retention (1), attention and calculation (1), recall (1), language (5), and copying (1), with a total score of 30 points. Dementia was defined as the participants scoring 24 points or below. We used a Tsung test with 20-question as an additional measure to assess the level of depression. Each of the 20 questions has answer options, including not at all, ranging from slightly disagree, slightly agree, definitely agree, and the depression was defined as scoring 70 points or more.

Covariates and other variables: The education level was categorized as low, medium, and high. Participants in the lower level of education group included no formal education and graduated elementary school. Suppose participants who graduated middle school, high school and college with no degree were considered medium education groups. High levels of education included bachelor’s degree, master’s degree and Ph.D. degree. Current smoking status was categorized as non-smokers and smokers. Participants were classified as married, cohabiting or single as a dichotomous variable. Family income was categorized as low, medium, and high. Information on the disease was derived from patient records. Daily fruit and vegetable intakes were evaluated using the Food Frequency Questionnaire [4]. A fruit and vegetable intake of more than 450 gr per week per person was referred to as sufficient, and an intake of less than 450 gr per week was referred to as insufficient intake Physical activity was assessed using the Global Physical Activity Questionnaire [13]. The physical activity and intake of fruits and vegetables were categorized as sufficient and insufficient, separately. Furthermore, we asked questions about diabetes history, diabetes duration, glycemic control, medication, insulin use and complications of diabetes.

Bodyweight, height, waist circumference, and blood pressure were measured using a standardized protocol. Body Mass Index (BMI; kg/m${}^2$) was subsequently calculated. A semiautomatic device was used to measure blood pressure in a half-sitting position. Fasting plasma glucose (FPG) was measured by the hexokinase method, and HbA1c was measured using high-performance liquid chromatography. Total cholesterol was measured using an enzymatic colorimetric method.

The study characteristics were expressed as means with a standard deviation (SD) for normally distributed variables and as numbers with percentages in cases of categorical data. The differences between groups were compared using the Student’s t-test for continuous variables. The frequency distributions of categorical variables were analyzed using the Pearson Chi-Square test and general linear analyses. Furthermore, Pearson Chi-Square test and two-way ANOVA were used to compare the difference between means of MMSE according to groups such as age (60 and $\ge$60 years of age), diabetes duration (1–5 years, 6–10, 11–15 and above 16 years), glycemic control (good and poor controlled as defined by glycated hemoglobin below and above 7.5%), presence of diabetes complications (yes/no), hypoglycemic occurrence (never, rare and regular).

Binary logistic regression analysis was performed to evaluate the association between diabetes and dementia. In univariate regression analysis, associations of age, gender, socioeconomic characteristics, and diabetic characteristics with dementia were tested. In addition, linear regression analysis was used to evaluate the associations of age with MMSE score. Furthermore, the interaction effect was tested with age in the association of diabetes with dementia. Unstandardized Beta-coefficients and Odds ratio (OR) are reported with a 95% confidence interval (CI). Analysis was adjusted for age, education, marital status and family income, diet and physical activity.

All statistical analyses were performed using IBM SPSS V.27.0 (Chicago, IL) and GraphPad Prism V.4.03 (San Diego, CA). A two-sided statistical significance was set at P 0.05 for all tests.

The mean age for DM and non-DM groups was 61.3 $\pm$ 8.5 and 61.0 $\pm$ 8.7 respectively, and 35.9% were male. Age and gender were matched (Table 1). General characteristics of the DM group were low levels of education, high BMI and low fruit and vegetable intake which was statistically not significant (P $>$ 0.05). Alcohol and tobacco use was relatively low in DM patients. In addition, fasting glucose level and total cholesterol findings in DM patients have statistically shown high significance (P 0.05).

The distribution of the MMSE score is shown in Fig. 1. In non-diabetic subjects, the distribution was relatively normal, with a trend of frequency of participants dependently decreasing with MMSE score. The majority of patients scored more than 24 on the MMSE score. There was no similar trend in people with DM. The MMSE scores were significantly different throughout the study groups: 26.1 $\pm$ 3.7 and 27.5 $\pm$ 2.6 for people with and without diabetes, respectively. According to the domains of the MMSE, patients with diabetes have scored less on functions in registration, recall, language, three-stage command, which was statistically significant (Supplementary Table 1).

Fig. 1.

Distribution of MMSE score, according to study groups. Note: MMSE, Mini mental state examination. The distribution of MMSE is shown in people with and without diabetes. Blue bars indicate MMSE in people without diabetes. Green bars indicate MMSE in people without diabetes. In non-diabetic subjects, the distribution was relatively normal, with a trend of frequency of participants dependently decreasing with MMSE score. The majority of patients scored more than 24 on the MMSE score. There was no similar trend in people with DM.

Cognitive impairment was defined as if the patients scored $\le$24 was 29% (n = 38) and 11.5% (n = 15) in people with and without diabetes. We compared CI and non-CI groups; education level was significant in people with and without diabetes, and age was significantly different in people without diabetes (Table 2).

MMSE was examined within the diabetic group to function diabetes duration, glycemic control, diabetic complications, and hyperglycemia. Of these measures, a longer duration of diabetes and more inadequate glycemic control were significantly associated with more unsatisfactory MMSE performance. As shown in Fig. 2, the MMSE score estimation was decreased with increasing diabetes duration, suggesting that the risk of dementia might be higher at long duration of diabetes (P 0.05). The means of MMSE score were 27.0 $\pm$ 2.1 and 25.4 $\pm$ 4.2 in people who are good and poorly controlled, as defined by glycated hemoglobin (P 0.05).

Fig. 2.

Mean of MMSE score, by diabetes duration. Note: MMSE, Mini mental state examination. The chart shows association between MMSE score and diabetes duration. The MMSE score was decreased with increasing diabetes duration.

In the logistic regression analysis, older age is significantly associated with a higher risk of dementia in people without diabetes, but not in people with diabetes. In both groups of diabetic and no-diabetics, higher education level decreased the risk of dementia by an OR of 0.36 (0.20–0.64) and 0.30 (0.12–0.71) for people with and without diabetes, respectively, as compared with lower education levels (Table 3). The association between education and dementia remained significant and materially unchanged after adjusted for age in both groups. Increasing diabetes duration and a higher level of glycated hemoglobin, insulin use, hypoglycemic occurrence and the presence of diabetic complications seem to increase the risk of dementia in people with diabetes. However, those were not significant (P $>$ 0.05, Table 3).

Since CI is an aging-related disease, we performed stratified analyses. In the diabetic group, MMSE scores were not significantly different in middle-aged and older people, while it was significantly different in people without diabetes (Fig. 3). Therefore, we tested it using linear regression analysis between age and MMSE in age groups in diabetic and non-diabetic subjects separately.

Fig. 3.

Mean of MMSE, according to age and diabetes status. Note: MMSE, Mini-mental state examination. The mean score of MMSE is shown in people with and without diabetes according to age group. According to age groups, black and grey bars indicate mean MMSE scores in people without/with diabetes respectively. The mean MMSE score in older people (60 years and older) was significantly lower than younger people (40–60 years) regardless of diabetes. MMSE score tends to be lower in people with diabetes. The interaction for the association between age and MMSE score through diabetes was not significant.

In the linear regression analysis, age was not associated with age groups, even in older people with diabetes. This suggests that age may not be the significant risk in CI development rather than diabetes itself in people with diabetes. In contrast, age was associated with the presence of dementia in people without diabetes. Significantly, the association was stronger in older age people without diabetes. Furthermore, this was independent of gender and education (Table 4). Further association of age and MMSE was independent of other socioeconomic characteristics such as marital status, family income, smoking and alcohol, remaining materially unchanged in people without diabetes (data not shown).