Abstract
AIMS/HYPOTHESIS: This study aimed to examine the association between diabetes and hyperglycaemia-assessed by HbA(1c)-and change in cognitive function in persons with and without diabetes. METHODS: This was a prospective cohort study of 8,442 non-diabetic and 516 diabetic participants in the Atherosclerosis Risk in Communities (ARIC) study. We examined the association of baseline categories of HbA(1c) with 6 year change in three measures of cognition: the digit symbol substitution test (DSST); the delayed word recall test (DWRT); and the word fluency test (WFT). Our primary outcomes were the quintiles with the greatest annual cognitive decline for each test. Logistic regression models were adjusted for demographic (age, sex, race, field centre, education, income), lifestyle (smoking, drinking) and metabolic (adiposity, blood pressure, cholesterol) factors. RESULTS: The mean age was 56 years. Women accounted for 56% of the study population and 21% of the study population were black. The mean HbA(1c) was 5.7% overall: 8.5% in persons with and 5.5% in persons without diabetes. In adjusted logistic regression models, diagnosed diabetes was associated with cognitive decline on the DSST (OR 1.42, 95% CI 1.14-1.75, p = 0.002), but HbA(1c) was not a significant independent predictor of cognitive decline when stratifying by diabetes diagnosis (diabetes, p trend = 0.320; no diabetes, p trend = 0.566). Trends were not significant for the DWRT or WFT in either the presence or the absence of diabetes. CONCLUSIONS/INTERPRETATION: Hyperglycaemia, as measured by HbA(1c), did not add predictive power beyond diabetes status for 6 year cognitive decline in this middle-aged population. Additional work is needed to identify the non-glycaemic factors by which diabetes may contribute to cognitive decline.