The Effect of Borderline Diabetes on the Risk of Dementia and Alzheimer’s Disease

The study population was derived from the Kungsholmen Project, a community-based cohort study on aging and dementia, which is fully described elsewhere (22,23). Briefly, all registered inhabitants who were living in the Kungsholmen district of Stockholm, Sweden, and were aged ≥75 years on 1 October 1987 were initially invited to participate in the project. At baseline (1987–1989), a two-phase survey consisting of a screening phase and a clinical phase was implemented. The screening phase included a health interview and administration of the Mini-Mental State Examination (MMSE) for all 1,810 participants. In the clinical phase, all subjects who screened positive (MMSE ≤23), as well as an age- and sex-stratified random sample of subjects who screened negative (MMSE >23), were invited to undertake a comprehensive physical, neurological, and psychiatric examination usually performed in clinical practice. During the clinical phase, 110 subjects were refusals and 225 were diagnosed as having prevalent dementia according to the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria (24). Thus, 1,475 of 1,810 baseline participants were identified as being free of dementia. Of them, 2 subjects had mental disorders and 172 refused to participate in the first follow-up examination (1991–1993) or had moved, and an additional 128 subjects were excluded due to having diabetes at baseline (i.e., history of diabetes, use of antidiabetes medications, or a random blood glucose level ≥11.0 mmol/l) (10) or missing baseline blood glucose data, leaving 1,173 subjects for the current analysis.

At the first follow-up, 914 subjects underwent clinical examination, and 738 were diagnosed as nondemented. Of them, 536 individuals underwent the second follow-up evaluation (1994–1996), and 41 refused to participate in the examination. At the third follow-up (1997–1998), 306 of the 435 nondemented survivors at the time when the second follow-up examination was performed received a dementia work up and 30 dropped out due to refusal or moving. Medical records and death certificates were available for all participants who died during the first (n = 259), second (n = 161), and third (n = 99) follow-up periods.

Informed consent was received from all participants or from informants when the person was cognitively impaired. The ethics committee at the Karolinska Institutet approved all phases of the Kungsholmen project.

At baseline, data on age, sex, and education were collected from the subjects by trained nurses following standardized protocols (22,23). Global cognitive functioning was assessed with MMSE. Weight and height were measured with a standard scale in light clothing and no shoes. BMI was calculated as weight in kilograms divided by the square of height in meters. Arterial blood pressure (i.e., systolic Korotkoff phase I and diastolic phase V) was measured by nurses with the subjects in a sitting position after at least a 5-min rest. Information on history of diabetes (ICD-8 and ICD-9 code 250), heart disease (ICD-8 and ICD-9 codes 410–414, 427, and 428), and stroke (ICD-8 and ICD-9 codes 430–438) until the baseline survey was derived from the computerized inpatient register, which encompasses all inpatient admissions to hospitals in the Stockholm area from 1969 on (25). Data on medical drug use for the 2 weeks before the baseline interview were collected from the subjects and verified by inspecting drug prescriptions and containers. Medical drugs were coded following the Anatomical Therapeutic and Chemical classification system (26). Antihypertensive drugs were defined as all medicines potentially used for lowering blood pressure (Anatomical Therapeutic and Chemical codes C02, C03, and C07). Genomic DNA was prepared from peripheral blood samples that were taken at baseline, and APOE allelic status was determined following a standard procedure.

Blood samples were taken at baseline, and blood glucose level was measured using a glucose oxidase procedure (2). Borderline diabetes was considered to be present if the random blood glucose level was ≥7.8 mmol/l but <11.0 mmol/l (27,28).

At each follow-up, all participants underwent a comprehensive clinical examination and cognitive tests (22,23). Cognitive functions were tested by asking for facts of general knowledge and past personal information (semantic and episodic memory), by object naming and comprehension of commands and sentences (language), by problem solving and interpretation of proverbs (abstract thinking), by copying figures (visuospatial ability), and by calculation and solving mathematical problems (calculation). Dementia was diagnosed on the basis of clinical judgement following the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria, in which a validated 3-step diagnostic procedure was used as previously reported (29). In brief, two examining physicians independently made a preliminary diagnosis, and in the case of disagreement, a third opinion was sought to reach a concordant diagnosis. The diagnosis of Alzheimer’s disease was similar to the international criteria for probable Alzheimer’s disease (30) and required gradual onset, progressive deterioration, and lack of any other specific causes of dementia. For subjects who died during each follow-up period, two physicians made a diagnosis of dementia or Alzheimer’s disease by thoroughly reviewing medical records and death certificates.

The baseline characteristics of subjects with and without borderline diabetes were compared using χ² tests for categorical variables and t tests for continuous variables. The incidence rate was calculated as the number of events occurring during the entire follow-up period divided by person-years of follow-up. Cox proportional hazards models were used to estimate the hazard ratio (HR) and 95% CIs of dementia and Alzheimer’s disease. For nondemented subjects, the follow-up time was calculated from the date of baseline interview to the date of the last follow-up examination or death. For the demented subjects, the follow-up time was estimated as the full time during which the subjects were free of dementia plus half of the follow-up time during which dementia developed (25). The proportional hazards assumption was confirmed by the tests based on Schoenfeld residuals, which showed that the hazards were proportional over the follow-up period among groups of subjects with and without borderline diabetes. The combined effect of two factors was assessed by creating dummy variables based on the joint exposures to both factors. We examined statistical interaction by incorporating the independent variables and cross-product term in the same model. Age, sex, education, baseline MMSE score, BMI, and vascular factors (i.e., heart disease, stroke, antihypertensive drug use, and blood pressure) were considered as potential confounders. As borderline diabetes is related to elevated mortality (31) and it is likely that dementia was under diagnosed in deceased individuals, we also adjusted for survival status at follow-up. Further, the development of diabetes during follow-up and its interaction term with borderline diabetes were also included in the model. All dementia, Alzheimer’s disease, and deaths were used as separate outcomes in the Cox regression analyses.

At baseline, 47 (4.0%) of the 1,173 dementia- and diabetes-free subjects were identified as having borderline diabetes. The prevalence of borderline diabetes was 4.2% in women and 3.4% in men (χ² = 0.34, P > 0.50). Subjects with borderline diabetes were more likely to have a lower MMSE score and higher blood pressure than those without this condition, but the two groups did not differ significantly in age, sex, education, heart disease, stroke, antihypertensive drug use, APOE ε4 status, or BMI (Table 1).

During the 9-year period, which covered a total of 6,076 person-years of follow-up (mean per person, 5.0 years; maximum, 10.5 years), 397 subjects were diagnosed with dementia, including 307 with Alzheimer’s disease. As shown in Table 2, subjects with borderline diabetes had a higher incidence rate of dementia and Alzheimer’s disease. The borderline diabetes-associated HRs were 1.61 (95% CI 1.02–2.58) for dementia and 1.68 (1.03–2.86) for Alzheimer’s disease after controlling for age, sex, and education. Further Cox regression analyses suggested that the association between borderline diabetes and elevated risk of dementia and Alzheimer’s disease was present independent of multiple potential confounders (Table 2). We repeated the analyses among subjects who survived until the time when dementia status was determined (n = 653, 348 dementia and 281 Alzheimer’s cases). Similar results were obtained after controlling for potential confounders; the borderline diabetes-associated HRs were 2.02 (95% CI 1.23–3.34) for dementia and 2.23 (1.32–3.67) for Alzheimer’s disease.

We examined the joint effect of borderline diabetes and severe systolic hypertension (≥180 mmHg) (32). Borderline diabetes in combination with severe systolic hypertension significantly increased the risk of dementia and Alzheimer’s disease (Table 3). The adjusted HRs related to the interaction term of borderline diabetes by severe systolic hypertension were 3.77 (95% CI 0.84–16.91; P = 0.08) for dementia and 4.89 (1.07–22.42; P = 0.04) for Alzheimer’s disease. We also conducted stratified analyses by borderline diabetes and APOE ε4. Among APOE ε4 noncarriers, borderline diabetes was significantly associated with an increased risk of Alzheimer’s disease, whereas in the ε4 carriers, borderline diabetes was related to a decreased risk of dementia. However, only a few borderline diabetic subjects were APOE ε4 carriers at baseline. Borderline diabetes showed no statistical interaction with APOE ε4 and no joint effect with obesity (BMI ≥30 kg/m²) on the risk of dementia and Alzheimer’s disease.

During the 9-year follow-up period, 17.0% of subjects with borderline diabetes and 1.1% of those without the condition developed diabetes (χ² = 68.5, P < 0.001). Among 47 borderline diabetic subjects, 22 died. Multiadjusted Cox regression analysis showed that borderline diabetes was significantly associated with elevated mortality (HR 1.48 [95% CI 1.06–2.06]). These results supported the need of adjustment for the potential confounding of follow-up survival status and the development of diabetes in the analyses that related borderline diabetes to dementia and Alzheimer’s disease.

This follow-up study on a community-based cohort of older adults suggests that borderline diabetes is associated with an increased risk of dementia and Alzheimer’s disease, and the risk effect on dementia and Alzheimer’s disease is independent of future development of diabetes risk. Borderline diabetes appears to have a multiplicative interaction with severe systolic hypertension (i.e., ≥180 mmHg) on the risk of Alzheimer’s disease.

The natural history of type 2 diabetes is preceded by impaired glucose regulation, which may last for years before it is clinically manifested (33). Our data showed that during 9 years of follow-up, the incidence of diabetes in subjects with borderline diabetes is much higher than in individuals without the condition. Clinical and cross-sectional studies have suggested that impaired glucose regulation is related to a decrease in cognitive performance (20,21,34). A prospective study showed that pre-diabetes increased the risk of developing cognitive impairment in elderly women (19). No prospective cohort study has investigated the relationship between borderline diabetes and the risk of dementia. We found that borderline diabetes was associated with an ∼70% increased risk of developing dementia and Alzheimer’s disease after controlling for major potential confounders. Diabetes was previously found to be associated with an increased risk of subsequent dementia and vascular dementia, independently of other vascular disease, and marginally with Alzheimer’s disease in this population (10). We were now able to expand our previous findings by showing that borderline diabetes is also associated with Alzheimer’s disease.

It is well known that diabetes increases the risk of vascular disease, which is one of the components for the diagnosis of vascular dementia. It is not surprising, therefore, that an increased risk of vascular dementia is linked to diabetes. It is likely that demented individuals who are suffering from diabetes tend to be diagnosed with vascular dementia. However, in the current study, it is unlikely that borderline diabetes played a role in dementia diagnosis because neither examining physicians nor subjects per se were aware of this condition when the diagnoses of dementia and its subtypes were made.

There are several mechanisms whereby borderline diabetes might influence the initiation and promotion of the underlying pathologies associated with dementia (35). First, increased blood glucose may have direct harmful effects on vascular endothelium or atherosclerotic plaques. Glucose dysregulation may develop in a cluster of risk factors including obesity, insulin resistance, atherogenic dyslipidemia, and hypertension. These factors constitute the metabolic syndrome, which are reported to be predictors of cerebrovascular disease, ischemic stroke, and accelerated cognitive decline and dementia (11,15,36,37). Second, toxic effects of higher blood glucose could lead to slowly progressive functional and structural abnormalities in the brain. Chronically hyperglycemic rodents have been found to express cognitive impairments and abnormalities in synaptic plasticity (38). These processes could affect brain tissue directly but can also lead to microvascular changes. The glucose-mediated effects on cognition and brain structure could be referred to as “accelerated brain ageing” (35,37). Third, glucose dysregulation may be linked to alterations in amyloid metabolism through changes in insulin and its receptor in brain and through the formation of advanced glycation end products (39). Insulin appears to stimulate amyloid-β secretion and inhibits the extracellular degradation of amyloid-β by competing for insulin-degrading enzyme.

We assessed the joint effects of borderline diabetes with other vascular or genetic factors. First, borderline diabetes appeared to have a multiplicative effect with severe systolic hypertension (≥180 mmHg) on the risk of Alzheimer’s disease, which is in line with the previous findings (10,40) that comorbid type 2 diabetes and hypertension exacerbated cognitive decline and had a much higher risk of developing Alzheimer’s disease. By contrast, a longitudinal study on a random sample of Medicare recipients aged ≥65 years found that individuals with diabetes had a lower risk of Alzheimer’s disease among subjects with hypertension (12). However, hypertension was defined in this study with the reference cutoff of systolic pressure ≥140 mmHg or diastolic pressure ≥90 mmHg or use of antihypertensive medications. Second, borderline diabetes and obesity (BMI ≥30 kg/m²) showed no joint effect on dementia risk in our elderly population. Third, we found that, among noncarriers of APOE ε4, borderline diabetes led to a higher risk of Alzheimer’s disease, which is in accordance with a recent report that the association between diabetes and incident Alzheimer’s disease was present only in APOE ε4–negative individuals (41). One explanation is that the strength of association between APOE ε4 and Alzheimer’s disease decreases with increasing age (42,43), and APOE ε4 is also associated with excess mortality (44). Thus, it is likely that those with ε4 allele have either died or been diagnosed with dementia in this elderly population, and the observed borderline diabetes–Alzheimer’s disease association among APOE ε4 allele noncarriers might be due to selective survival (45). Furthermore, carriers of ε4 allele develop sufficient Alzheimer’s pathology over the course of life to bring them to the threshold for expressing dementia, while those without an ε4 allele require further physiologic insults, such as diabetes, to reach this threshold (46). In addition, Alzheimer’s disease subjects with APOE ε4 tend to have normal insulin and glucose levels, but those without APOE ε4 allele are more likely to have higher glucose and insulin levels (47). However, the statistical power was limited for these interactive analyses due to the small number of borderline diabetic subjects with severe systolic hypertension, obesity (BMI ≥30 kg/m²), or APOE ε4 at baseline.

The major strengths of our study are the relatively long term of follow-up on a large-scale community cohort and the comprehensive assessments to reach the diagnosis of dementia or Alzheimer’s disease. However, a few limitations deserve to be mentioned. First, we defined borderline diabetes according to a random blood glucose level, which might result in an attenuation for the given associations (27). Second, we only analyzed baseline borderline diabetes in relation to dementia. Some participants have actually developed borderline diabetes during the follow-up period, which would lead to an underestimation of the association between borderline diabetes and dementia risk. Third, as the defined population of the present study consisted of subjects with a minimum age of 75 years at entry, caution is required when generalizing our findings to younger populations. Finally, we were unable to assess the effect of insulin resistance per se on dementia because insulin levels were not measured in our study.

In summary, the major findings that borderline diabetes is associated with an increased risk of dementia and Alzheimer’s disease supports the view that impaired glucose regulation could lead to dementia, including Alzheimer’s disease. These findings have relevant implications for public health as previous studies (33,48) have shown that early stage diabetes could be improved by lifestyle changes. Our findings also highlight the need to detect borderline diabetes in order to effectively prevent both diabetes and dementia.

Research grants were received from the Gamla Tjänarinnor Foundation (Sweden), the American Alzheimer’s Association (U.S.), the Loo and Hans Ostermans Foundation (Sweden), and the Swedish Research Council for Medical Research.

We thank all members in the Kungsholmen Project Study Group for their collaboration in data collection and management.