Diabetes and cognitive impairment: a community-based study of elderly subjects

Age and Ageing,  Sept, 1995  by Simon C.M. Croxson,  Carol Jagger

• E-mail
• Print
• Link

Introduction

Studies have shown that elderly diabetic subjects have impaired cognitive function [1-5], but these studies were generally not population-based, excluded subjects with dementia [2-5], and generally used a large battery of tests to show the deficit [1,3-5]. Interestingly, worse cognitive function in these tests was associated with worse glycaemic control [1-5].

Dornan's general-practice-based study [6] examined 98 known diabetic subjects and found that 15 had an abbreviated mental test less than 10/10 [6, personal communication], compared with only 1 of 98 age/sex matched subjects not known to be diabetic. Although this difference is statistically significant, most people would use a cut-off of below 8 rather than below 10 [7].

Other workers have either found no decreased cognitive function in elderly diabetic subjects [8], or found that the decreased cognitive function in diabetic subjects is totally explained by depression in the diabetic subjects [9]. Other workers have found that diabetes is associated with depression [10].

Studies looking at dementia sufferers found a low prevalence of known diabetes in Alzheimer-type dementia (SDAT), from 0 to 6%, but a higher incidence of 12 to 30% in multi-infarct dementia sufferers (MID) [11-14]. These studies vary in subject nationality, sample size, whether or not mixed MID/ SDAT subjects were included as MID subjects, the diagnostic methods to classify the dementias, and the care in accepting the diagnosis of diabetes, which may be erroneous [15]. These studies all show a raised prevalence of diabetes in MID, and a decreased prevalence in SDAT. Studies of the type of dementia in populations suggest that 50-65% is SDAT, 12-24% is mixed SDAT/MID, and 8-29% is MID [12, 16].

Studies of subjects with dementia often do not fully know their glucose tolerance status, and diabetic detection and complication surveys rarely examine cognitive function. It may be that diabetic subjects are less likely to develop SDAT, but more likely to develop M I D. The question is whether diabetic subjects have a different prevalence of cognitive impairment from their normal glucose-tolerant peers. The elderly population of Melton Mowbray has been screened for diabetes [17] and for cognitive impairment [18] and data from these studies were pooled to examine this question.

Melton Mowbray is an ideal place for epidemiological research since its population has a similar age, sex and social class structure to the UK, and because all residents attend one general practice which, in conjunction with the Leicester University Department of Epidemiology and Public Health maintains an accurate computerized population register.

Method

The Melton Mowbray diabetes screening survey [17] screened all residents aged 75, 80 or 85 years of the Melton Mowbray area if alive and willing by means of a modified oral glucose tolerance test (MOGTT) according to WHO criteria. All residents aged 75 and over participated in a health survey administered by trained local people [18], which included mental status assessment by the Folstein mini mental state examination (MMSE) [19].AI1 subjects aged 75, 80 or 85 who were approached for the diabetes survey had their results from the health survey examined. The diabetes survey was performed from December 1987 to August 1988, and the health survey from January 1988 to June 1988.

Results

MMSE scores were available for 21 of the 26 known diabetic subjects, 218 of the 230 subjects agreeing to a MOGTT, and 80 of the 106 subjects who refused to have a MOGTT; missing MMSE scores were due to death, emigration or refusal to answer the questionnaire. The MMSE scores are shown (Table I) grouped into low (<22), middle (22-23), and high (>23) following previous work using the MMSE in Melton [20]. Analysis was performed using a 23/24 cut-off since this in has been found to give a reasonable sensitivity and specificity (over 80%) for detecting all grades of dementia in Melton [20], and has been found to be the optimum cut-off by other investigators [21].

[TABULAR DATA I OMITTED]

Within each age group, there were no sex differences in distribution of MMSE score less than 24 when considering either subjects with normal glucose tolerance, or all subjects for whom MMSE scores were available (Fisher's exact p >0.1); thus further analysts was performed with the sexes combined.

Analysis of data contained within several contingency tables can be problematic [22]. Data from each age group were reduced to 2 by 2 contingency tables for cognitive function (MMSE cut-off 23/24) by glucose tolerance [using various dichotomous classification of glucose tolerance (see Table II)]. This gave larger cell sizes to allow a Mantel Haenszel test for a series of age stratified 2 by 2 tables to be applied [23].

[TABULAR DATA II OMITTED]

Discussion

The data from the Melton surveys (Table II) suggest that in comparison with subjects with normal glucose tolerance, subjects with known diabetes are more likely to have a low MMSE score (odds ratio 3.30, 95% CI 1.29-8.48), arid subjects with newly diagnosed diabetes are less likely to have a low MMSE (upper 95% CL of odds ratio 0.003) than subjects with normal glucose tolerance; overall, these effects cancel out so that all diabetic subjects combined had similar cognitive function to normal glucose tolerance subjects.