Diabetes mellitus and the risk of dementia

Methods.

Results.

At baseline, 692 (10.9%) of the 6,370 study participants had diabetes mellitus, of whom 390 (56.4%) did not use antidiabetes medication or were newly diagnosed, 232 (33.5%) used oral medication, and 70 (10.1%) were treated with insulin. Among those using insulin therapy, there were no insulin-dependent diabetes mellitus patients. Diabetes prevalence increased with age. Adjusting for age there was no difference in diabetes rates between men and women (table 1). Except for education and atrial fibrillation all possible confounding variables differed significantly between persons with and without diabetes mellitus.

Duration of follow-up ranged from 4 days to 6 years; quartiles were 1.5, 1.9, and 2.5 years, and average follow-up was 2.1 years. During the follow-up period, 126 (2.0%) participants became demented. Of those 89 (70.6%) were diagnosed with AD (13 with coexistent cerebrovascular disease), 18 (14.3%) were classified into vascular dementia, and 19 (15.1%) had another type of dementia. Duration of follow-up did not differ between dementia subtypes. Dementia patients were significantly older than those who were nondemented (table 2). There was no difference in dementia risk by gender. The age- and sex-adjusted diabetes prevalence was significantly higher in persons who developed dementia as compared to nondemented.

Diabetes mellitus almost doubled the risk of dementia (age- and sex-adjusted relative risk (RR) 1.9, 95% CI 1.3 to 2.8, table 3). A comparable result was obtained when analyses were restricted to those in whom the diagnosis was made by reviewing medical files (adjusted RR 1.7, 95% CI 0.8 to 3.7). Relative risks were similar for men (RR 1.8, 95% CI 0.8 to 4.1) and women (RR 1.9, 95% CI 1.2 to 3.0). We observed no clear modification in diabetes related dementia risk by age. Relative risks were 2.0 for those younger than 75 years, 2.2 in those 75 to 84 years, 1.4 in persons aged 85 and older.

The risk of incident dementia was clearly modified by baseline treatment (see table 3), with the lowest relative risk in newly discovered or untreated diabetic patients (RR 1.3, 95% CI 0.7 to 2.3) and the highest RR among patients treated with insulin (RR 4.3, 95% CI 1.7 to 10.5). In 298 of the 390 (76%) diabetic patients without medication at baseline, diabetes treatment at follow-up was known. Of these, 227 (76%) did not use antidiabetes medication, 67 (22%) used oral medication, and 4 (1%) insulin. The latter 71 diabetes patients were not at increased risk of dementia (RR 0.8, 95% CI 0.1 to 5.6).

Additional adjustment for possible confounders or intermediates, even when all were added to the regression model, gave a maximum decrease of the RR of 0.2 while remaining significant. Diabetes increased the risk of both AD and vascular dementia (table 4). If analysis was limited to patients who underwent neuroimaging, the relative risk for AD was 2.3 (95% CI 0.96 to 5.5). Additional adjustments for confounders again did not result in substantial changes of the estimates.

In this population, the fraction of incident dementia attributable to diabetes mellitus (population attributable risk) was 8.8%; 8.1% in men and 9.2% in women.

Discussion.

In this large prospective study of dementia, persons with diabetes mellitus were at increased risk of dementia, including AD, particularly if they were treated with insulin. These results confirm our previous cross-sectional findings of an association of type 2 diabetes with dementia and AD.9

Major advantages of the Rotterdam Study to investigate risk factors of dementia are the size of the study, the longitudinal and population-based design, and the fact that exposure and most of disease status were defined by active in-person examination. Presence of diabetes was assessed before onset of dementia, which prevented ascertainment bias. Selection bias was minimized by the population-based design and because follow-up of the population at risk of dementia was complete. However, not all participants could be rescreened in person and this may have led to selective underdiagnosis of dementia in this part of the population. Still, we found a similar elevated risk among diabetes patients in whom the dementia diagnosis was made by reviewing their medical files.

A limitation of the current study is that brain imaging could not be obtained of all dementia patients and we may have underestimated cerebrovascular pathology in demented persons. Although increased relative risks of AD by diabetes were not different in patients for whom brain imaging was available as compared to those without MRI, it remains possible that unrecognized strokes are involved in the association we found.

Previous cross-sectional studies have shown positive associations of diabetes with vascular dementia, but an inverse relation with AD.4,5,7,8 However, these studies were based on selected patients and controls. The presence of diabetes was assessed from medical records and not actually screened for. Furthermore, in these studies, patients with indicators of vascular or cerebrovascular disease were mostly excluded from the AD group, thereby obstructing the possibility to study these factors, and thus diabetes mellitus, in relation to AD. Only two prior longitudinal studies on the risk for dementia by diabetes have been published.11,12 In these studies, the relative risk of AD by diabetes is of the same magnitude as in our study: RR 2.2 (95% CI 0.97 to 4.9) in the prospective Hisayama study, which had a longer follow-up period but included fewer participants (n = 828).11 In the Rochester study, which was retrospective and register-based, RR were 2.3 (95% CI 1.6 to 3.3) for men and 1.4 (95% CI 0.9 to 2.0) for women.12

The relation between diabetes mellitus and dementia could either be explained through vascular disease or by nonvascular effects of diabetes. Diabetes mellitus is notorious for micro- and macrovascular complications.28 It is a well-known risk factor for strokes,1,2 which, if they accumulate or strike vital brain segments, may cause dementia.25 An increased risk of vascular dementia by diabetes is to be expected.

Our finding with respect to AD is less evident. One explanation for this finding would be that we misdiagnosed patients with dementia due to vascular causes as AD, although this would require a considerable number of wrongly subtyped patients to get the substantial RR we found. It may also be that the syndrome, defined by the currently accepted AD criteria,25 indeed has a vascular connection. The notion of vascular involvement in the pathogenesis of AD is growing.25,29,30 Our finding of a high RR for AD with cerebrovascular disease points toward a second possible explanation: silent cerebral infarcts may have uncovered imminent AD. Even in those we diagnosed as having AD only, silent strokes may have contributed to the dementia. These small infarctions, which tend to occur deep in the brain and cause no neurologic symptoms, are hard to diagnose without pathology. In conflict with our findings, a recent retrospective postmortem study found no difference between 49 patients with diabetes and 52 age- and sex-matched controls with respect to Alzheimer-type brain pathology.31 In another recent study, however, an association was found between features of the insulin resistance syndrome and AD, which may support our findings.10

There are several arguments to consider the possibility that other than vascular factors may contribute to dementia in diabetic patients. In our study the limited effect of adjustment for vascular risk indicators supports a nonvascular pathway. Previous studies have reported nonvascular functional and structural changes in the CNS of diabetic patients.32,33 Also, diabetic patients without clinical cerebrovascular disease34,35 were found to perform significantly poorer on cognitive tests than healthy controls. Recently, advanced glycation end-products (AGE), which are the result of nonenzymatic glycosylation of proteins,36 were found in plaques and tangles of Alzheimer brains, even in early stages of disease.13 Excessive AGE formation by chronic hyperglycemia may play a role in long-term diabetic complications.15 In brains of AD patients the receptor for AGE appears overexpressed,14 and this receptor is activated by β-amyloid, which is involved in Alzheimer pathology.37 Activation of this receptor leads to increased oxidative stress that may result in cellular damage.29

In both this and our previous study we found the strongest association between diabetes and AD for patients who at baseline were on insulin treatment.9 This finding may reflect greater risk of dementia in more severe diabetes. Patients on insulin treatment may have more severe diabetes, or a longer history of diabetes, and are thus longer exposed to diabetes related risk factors. Leibson et al., however, found no effect of diabetes duration on dementia risk.11 Direct or indirect effects of insulin could contribute to the risk of dementia. Hypoglycemic episodes frequently complicate insulin treatment,38 and contrary to hyperglycemia, hypoglycemia can cause irreversible brain damage.32,33

The almost doubled risk of dementia in diabetic patients implies that diabetes may play a role in the pathogenesis of dementia in a considerable proportion of all dementia patients. Further studies to unravel the underlying mechanisms are required.