PDF
ePub
Citation
Print
INTRODUCTION
Dementia is a global epidemic; the number of people with dementia worldwide was estimated at 47.47 million in 2015 and expected to reach 135.46 million in 2050 [1]. In several studies and meta-analyses, diabetes was estimated to increase the risk of dementia by approximately 50% [2]. This increased risk in subjects with type 2 diabetes mellitus translates into an average earlier onset of dementia of 2.5 years compared with subjects without diabetes [3].
Whether diabetes is an actual risk factor for Alzheimer's disease (AD) has been the focus of many studies [4]; however, the risk factors for dementia in patients with diabetes have been evaluated in relatively few studies [56]. Results from previous studies showed that diabetic participants with poor glycemic control [7], long duration of diabetes [5], pre-existing micro- or macrovascular complications [58], and severe hypoglycemia [8] were at increased risk of dementia. However, in those studies, the number of study participants was small [57], the assessment of confounding factors was limited [68], and no differentiation between specific types of dementia was presented [568]. In only one study in Taiwan, the relationship between diabetes and subsequent AD incidence was investigated using nationwide population data [6]. However, the authors did not have access to data such as metabolic risk factors, smoking, alcohol, and socioeconomic class, because the National Health Insurance Claim data were based only on the diagnosis codes.
Recently, concerns have increased regarding young-onset dementia (YOD), in which symptoms occur before the age of 65 years. YOD has been estimated to comprise approximately 8% of new dementia cases in Australia [9]. Although AD is the most common type of YOD, secondary dementia is more frequent in YOD patients compared with late-onset dementia (LOD) [1011]. Due to its rarity, the diagnosis of YOD is often delayed, but has a significant psychosocial effect because the most productive years of patients' lives are affected [10].
In the present study, the National Health Insurance Service (NHIS) data were used to estimate the effect of diabetes on the development of dementia in general, and specifically on AD or vascular dementia (VaD). Furthermore, the possible risk factors for dementia in diabetes were comprehensively investigated. Lastly, the incidence rate and specific risk factors for YOD in diabetes were examined.
METHODS
Design, study setting, and participants
The NHIS is a government-operated mandatory social health insurance program that covers almost the entire Korean population (approximately 97%). The NHIS in Korea comprises comprehensive sets of health information for approximately 50 million Koreans [1213]. The NHIS, which contains claims and mortality data can be accessed by all researchers whose protocols are approved by both a NHIS Review Committee and an Institutional Review Board (IRB). This study protocol was approved by the IRB of Korea University Ansan Hospital (IRB number: 2019AS0138). Informed consent was waived by the board.
From this database, 2,482,918 individuals 40 years of age or older with diabetes who underwent health checkups provided by the NHIS between January 1, 2009 and December 31, 2012 were selected. Then, patients diagnosed with dementia prior to enrollment (n=20,652), and those with any missing data (n=544,564) were excluded. A total of 1,917,702 participants were finally included as the study population and followed up until dementia was diagnosed or until December 31, 2015 (Supplementary Fig. 1).
Demographic factors
Body mass index (BMI) was calculated by dividing the weight (kg) by height squared (m2). Based on BMI, participants were classified as underweight (<18.5), normal weight (18.5 to 22.9), overweight (23 to 24.9), obese (25 to 29.9), or severely obese (≥30) [14]. Blood samples were taken after an overnight fast and quality control procedures followed the Korean Association of Laboratory Quality Control guidelines.
Detailed information regarding smoking status, alcohol consumption, physical activity, and past medical history of study participants was assessed via standardized self-reported questionnaires. Subjects were categorized as non-smokers, ex-smokers, or current smokers. Heavy drinkers were defined as individuals who consumed ≥30 g of alcohol per day [15]. Physical activity was divided into exercising less strenuously versus more than three times a week for at least 20 minutes per day. The baseline comorbidities of participants were identified based on the combination of past medical history and clinical and pharmacy International Classification of Diseases (ICD-10) codes.
Definitions
Individuals with diabetes were defined as either patients who were both prescribed anti-diabetic drugs and had the presence of ICD-10 codes (E11–E14), or patients with fasting glucose levels >126 mg/dL derived from the health check-up database [16]. Treatment of diabetes was defined as NHIS medication data for drugs prescribed during the same year as the health examination. Anti-diabetic medications were categorized as oral hypoglycemic agents (OHAs) and insulin. The OHAs dispensed in the pharmacy during the study period consisted of six classes (sulfonylurea, biguanide, alpha-glucosidase inhibitor, thiazolidinedione, dipeptidyl peptidase 4 inhibitor, and meglitinide). Duration of diabetes was calculated from the first diagnosis of diabetes to the baseline examination on study entry. The diagnosis of diabetic retinopathy was defined based on the ICD-10 code H36.0.
The presence of hypertension was defined as a systolic/diastolic blood pressure ≥140/90 mm Hg or the presence of at least one claim per year for the prescription of antihypertensive agents under ICD-10 codes (I10–I13, I15). The presence of dyslipidemia was defined as total cholesterol ≥6.21 mmol/L or the presence of at least one claim per year for the prescription of antihyperlipidemic agents under ICD-10 codes (E78). Estimated glomerular filtration rate (eGFR) was calculated using the equation from the Modification of Diet in Renal Disease (MDRD) study: eGFR=175×serum creatinine−1.154×age−0.203×0.742 (for women). The eGFR <60 mL/min/1.73 m2 was defined as chronic kidney disease (CKD) [17]. The history of cardiovascular diseases (CVDs), stroke, myocardial infarction, and angina pectoris, was evaluated using standardized questionnaires. Myocardial infarction and angina pectoris were regarded as ischemic heart diseases (IHDs). Depression was defined using ICD-10 codes (F32–34).
Dementia was determined based on prescribed anti-dementia medications (rivastigmine, galantamine, memantine, or donepezil) along with ICD-10 codes (F00, F01, F02, F03, G30, or G31) for medical expense claims submitted to the NHIS. AD was diagnosed with the F00 and G30 codes and VaD with the F01 code. Other dementias were defined based on the F02, F03, and G31 codes, and when dementia subtyping was ambiguous. Two or more dementia diagnosis codes were sometimes registered together. If primary and secondary diagnoses for dementia existed, the dementia subtype was determined based on the primary diagnosis. If all dementia codes were registered as only secondary diagnoses, the decision was delayed until the next visit. If the dementia codes were clarified as primary and secondary diagnoses at the next visit, the dementia subtype was defined from the primary diagnosis code. If dementia codes were continuously registered as only secondary diagnoses, the case was defined as “other dementia.”
Statistical analysis
Data are expressed as mean±standard deviation, geometric mean (95% confidence interval [CI]) or percentage. Differences between groups were identified using the Student's t-test; the chi-square test was used to assess differences in the distribution of categorical variables. The incidence rates were presented as the number of dementia occurrences per 1,000 person-years. Multiple Cox proportional hazards regression analysis was used to evaluate the association between risk factors and development of all types of dementia, AD, or VaD. A competing risk analysis was completed to determine whether accounting for deaths is a competing risk. In addition, the effect of risk factors on outcome based on age <60 years versus ≥60 years was evaluated. Because the usual age cutoff for YOD and LOD is the development of dementia at 65 years of age, the YOD and LOD were defined with a baseline age of 60 years, since average follow-up time was approximately 5 years in this study. To show the association between generalized or central obesity and incident dementia, multivariate adjusted incidence rates and hazard ratios (HRs) for dementia based on BMI and waist circumference (WC) categories were analyzed. All statistical results were analyzed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and a P<0.05 was considered statistically significant.
RESULTS
Development of dementia (all types, AD, VaD)
During an average of 5.1 years of follow-up, 92,758 participants developed dementia; 66,092 developed AD, 12,813 developed VaD, and 13,853 developed other dementia. The incidence rates of all types of dementia, AD, and VaD were 9.5, 6.8, and 1.3/1,000 person-years, respectively, in participants with diabetes. Although the incidence of dementia sharply increased with age, YOD comprised 4.8% of the total dementia occurrence. The incidence rate of YOD was 0.9/1,000 person-years, in contrast to LOD, which was 17.9/1,000 person-years. AD and VaD comprised 55.7% and 26.7% of YOD and 72.0% and 13.2% of LOD, respectively. The incidence rates of total dementia, AD, and VaD per 1,000 person-years based on age groups are shown in Fig. 1.
Baseline characteristics based on the development of dementia
Table 1 shows the baseline characteristics of participants grouped based on future development of dementia. Participants who were older, female, non-smokers, consumed less alcohol or were non-drinkers, and did not exercise had more incident dementia. In addition, subjects who had lower income, plasma glucose levels, BMI and WC, and those with hypertension, dyslipidemia, diabetic vascular complications, and depression developed dementia more frequently. Patients who developed dementia had a longer duration of diabetes and were treated with OHAs and insulin more frequently than subjects who did not develop dementia.
Age- and sex-adjusted risk factor analysis
The effects of variables on the development of dementia after adjusting for age and sex were examined (Supplementary Table 1) and similar to the results shown in Table 1, except for the effects of smoking; current smoking was a risk factor for dementia in this model. When dementia was specified as AD or VaD, the effects of old age and female sex were stronger for incident AD than VaD; however, current smoking, hypertension, stroke, and CVD had higher HRs for VaD than AD. The other variables had similar effects on both AD and VaD (Supplementary Table 1).
Multivariate analysis
Compared with the age- and sex-adjusted model, multiple Cox proportional hazard analysis after adjusting for all variables together did not significantly modify the associations between risk factors and dementia, except for mild alcohol intake, which became a protective factor, and for dyslipidemia, which had a null effect (Table 2). When comparing AD with VaD, female sex was a risk factor only for AD. Current smoking, hypertension, and stroke showed higher HRs for VaD than AD. Fasting plasma glucose (FPG) levels ≥126 mg/dL were a protective factor for AD but were not significant for VaD. BMI had a divergent effect on dementia; low BMI (≤18.5 kg/m2) was a risk factor but high BMI (≥23 kg/m2) was a protective factor for both AD and VaD, with stronger effects on AD than VaD. However, in a competing risk analysis where death was considered a competing risk, female sex was also a risk factor for VaD as well as AD. Low BMI was not a risk factor for VaD but very high BMI (≥30 kg/m2) was a protective factor for VaD.
Risk factors for YOD and LOD
When the same analysis was repeated to examine the moderating effect of age (Table 3), most risk factors had a greater influence on YOD than on LOD, especially low BMI, history of stroke, and depression. However, female sex was a stronger risk factor for the development of AD in LOD than in YOD. FPG levels ≥126 mg/dL were a protective factor for AD in YOD but a risk factor for VaD in LOD.
The effects of BMI or WC on dementia based on age were compared (Fig. 2). After adjusting for several confounding factors, consistent negative associations were observed between BMI and dementia in YOD and LOD patients, with a higher impact of underweight on YOD compared with LOD (P for interaction <0.001) (Fig. 2A). A slight U-shaped association was observed between WC and dementia; lower WC (males ≤85 cm, females ≤80 cm) was associated with a higher risk of dementia. Although the effect was weaker, higher WC (males >100 cm, females >95 cm) also increased the risk of dementia regardless of age (Fig. 2B).
DISCUSSION
In the present study, the incidence and risk factors for dementia were observed in participants with diabetes using Korean representative data. The overall incidence of dementia in diabetes was 9.5/1,000 person-years and increased exponentially with age. Participants at higher risk of dementia were females, smokers, had low income, long duration of diabetes, were taking more hypoglycemic agents, and had micro- and macrovascular complications, hypertension, depression, or were underweight. Mild alcohol intake, regular exercise, and being overweight were protective factors for dementia in people with diabetes. The incidence rate of YOD was 0.9/1,000 person-years, and risk factors were similar to LOD; however, the HRs of low BMI, low WC, a history of stroke, and depression were significantly higher for YOD than LOD.
The difference between the incidence and risk factors for AD and VaD in subjects with diabetes was compared in this study. In developed countries, the ratio of AD/VaD is higher than in developing countries [18]. According to a meta-analysis of epidemiologic studies on dementia in Korea, the AD/VaD ratio increased from 1.96 in the early 1990s to 4.13 in the 2010s [19]. In the present study, the ratio was approximately 5.2 in people with diabetes, and vascular risk factors such as smoking, hypertension, and stroke showed higher HRs for VaD than AD. Because diabetes is a well-known risk factor for VaD [20], the improvement of vascular risk factor management in diabetes may be responsible for this relatively lower incidence of VaD compared with AD. In this study, the ratio of AD/VaD was 2.1 for YOD compared with 5.5 for LOD, indicating the incidence of VaD in YOD is higher than in LOD.
In several studies, uncontrolled diabetes was shown associated with dementia in diabetic subjects [72122]. Similar to results reported in many previous publications [56], duration of diabetes and insulin treatment were risk factors for dementia in the present study; however, we found high glucose did not lead to dementia. The association of high glucose with dementia remains controversial. Bruce et al. [5] showed that glycosylated hemoglobin and fasting glucose were not associated with an increased risk of dementia, whereas Xu et al. [7] found that uncontrolled diabetes was associated with an increased risk of dementia. Poor glycemic control was associated with greater cognitive decline at 70 to 79 years of age [23], but with less decline over 80 years of age in participants with diabetes [24]. Although severe hypoglycemia is a risk factor for dementia [8], whether the lower fasting glucose levels of participants in this study indicate that more hypoglycemic events were associated with more frequent prescription of OHAs or insulin is unclear.
In the present study, being underweight was shown a risk factor for dementia. Higher BMI was a protective factor for AD and VaD in both age groups. Controversy exists surrounding the association between obesity and dementia. In a meta-analysis, the harmful effects of higher BMI was evident in long-term follow-up, although a protective effect was evident in short-term follow-up [25], indicating that even if obesity is associated with a high risk of dementia, weight loss during the preclinical period appears a protective factor (a reverse causality). This concept was also confirmed in the Whitehall II study, in which the BMI trajectory was shown in incident dementia for 28 years [26]. The authors explain the inconsistency between studies was due to insufficient follow-up time, which did not take preclinical weight loss into account. Therefore, the 5.1-year follow-up period in the present study is insufficient for accurate differentiation between a real effect of obesity and weight loss before the onset of disease. In addition, using BMI does not distinguish between changes that are due to alterations in lean mass from adipose mass. Body weight tends to decrease as part of normal aging, which might result from a combination of muscle and adipose mass loss in late life [2728]. This age-associated change in body composition and related hormonal change may promote AD and vascular pathology. Although the mechanism related to the development of YOD has yet to be determined, underweight condition due to sarcopenia in midlife might explain the strong association between YOD and low BMI in the present study.
In several reports the increasing problem of YOD has been demonstrated [102930]; however, the incidence and risk factors have been evaluated in only a few studies. The YOD incidence rate in the present study was 0.9/1,000 person-years, and traditional cardiovascular risk factors such as smoking, hypertension, CKD, stroke, and IHD were associated more strongly with YOD than LOD in patients with diabetes. In addition, a strong association between YOD and vascular disease including stroke, transient ischemic attack, CKD, or hypertension was shown in previous studies [313233]. These results indicate that many cases of YOD in individuals with diabetes are preventable by adequate management of vascular causes.
A history of depression or stroke had a stronger influence on the development of YOD than LOD in people with diabetes. The frequency of depression was also higher in subjects with YOD than with LOD in the Registry of Dementia of Girona [30]. In several longitudinal studies, depression was a major risk factor for incidence of dementia [34]. Depression was shown to injure neurons by several different mechanisms such as inflammation [35], elevated glucocorticoid production [36], amyloid deposition [37], or dysfunctional immune activity [38], which may lead to cognitive impairment. However, in a recent study, depression was a significant risk factor for YOD only if occuring within 10 years from YOD onset [32], indicating that depression might be a prodromal consequence of YOD rather than an etiological risk factor. Further research is needed to reach a definite conclusion whether the two conditions are linked as risk factors or are part of a continuum. The mechanism by which stroke confers an increased risk of YOD remains unclear. Stroke may trigger a neurodegenerative process by disrupting amyloid clearance [39] or by activating autoimmune responses [40] in the brain. Ongoing exposure to existing vascular risk factors, immune processes, and pathogenic mechanisms may contribute to risk of dementia after stroke in younger people.
The present study had several limitations. First, since the Korean insurance claim data was used based on diagnosis codes, instances of misdiagnosis or misclassification of diabetes and dementia including its subtypes may have occurred. Second, except for the duration of diabetes and controversial glycemic control, modifiable factors specific to the patients with diabetes compared with known dementia risk factors in the general population were not found. Comparison with people without diabetes may clarify diabetes-specific risk factors for dementia. Nonetheless, to date, this is the largest study in which the incidence and risk factors for dementia and its subtypes were evaluated using a nationally representative cohort study of diabetes.
In conclusion, the incidence and risk factors associated with dementia in Korean participants with diabetes were presented in this study. Because diabetes is closely associated with dementia and many vascular risk factors for dementia are modifiable, the results may aid in creating effective preventive strategies to decrease the rising dementia epidemic and its related health and socioeconomic burden.
XML Download