(SLA-060) Are diabetes treatments and education associated with cognitive performance? A UK Biobank study on type 2 diabetes

Background: Optimal control of type 2 diabetes (T2DM) in older adults may mitigate cognitive decline, a precursor to dementia. Research on antidiabetics and cognitive performance has shown mixed findings. Education levels appear to be a cognitive reserve marker, but their impact on this population remains uncertain.

Objectives: To investigate the association between repeated measures of antidiabetic use and cognitive function in T2DM and the moderating role of cognitive reserve.

Methods: A retrospective cohort study on UK Biobank participants enrolled between 2006-2010 with linked primary care data. Participants with T2DM, no prior dementia, baseline cognitive assessments – fluid intelligence (FI; scored 0-13) – were included. Participants were followed up to 2015; some had further FI tests during follow-up. Three other domains, symbol digit substitution, numeric memory and trail-making, were assessed in cohort subset. Antidiabetic exposures were determined over 2-year period before each cognitive test and categorised into 5 groups: metformin, insulin, thiazolidinedione (TZD), sulfonylurea (SU), and other antidiabetics. The use of metformin, TZD and SU was classified into 3 duration levels: none, <=1 year or >1 year. Education years were used as a marker of cognitive reserve and presented in 3 levels: 7, 10-15, and 19-20 years. A linear mixed-effect regression model with patient-level random intercept was fitted to FI and symbol digit substitution scores, while a linear regression model was fitted to numeric memory and trail-making. All models included interactions between antidiabetics and education. The model accounted for concomitant antipsychotics and anticholinergics, comorbidities, age, sex, ethnicity, deprivation quintile, body mass index, and smoking and drinking status.

Results: This study included 4556 participants, with 5840 FI scores in 2006-2015. SU was negatively associated with high FI scores (i.e. better cognition) (<=1 year: coefficient: -0.66 [95%CI: -1.16, -0.16]; >1 year: -0.36 [-0.7, -0.02]). Metformin >1 year was linked to high FI scores, (0.37 [0.09, 0.64]).
More years of education (19-20 vs 7 years: 1.75 [1.54, 1.96]), white ethnicity (2.07 [1.88, 2.25]), former smokers (0.14 [0.02, 0.26]) and current drinkers (0.47 [0.24, 0.7]) significantly improved cognition. Education was not a moderator.
Conversely, co-existing stroke (-0.35 [-0.69, -0.01]) and heart disease (-0.18 [-0.34, -0.01]), older age (-0.01 [-0.02, -0.01]) and less deprived areas (Q5 vs Q1: -0.63 [-0.82, -0.45]) negatively impacted FI scores.
Greater education years were significantly related to better cognition in symbol digit substitution (N=1081), numeric memory (N=825) and trail-making (N=786), but no association with antidiabetics.

Conclusion: SU was negatively associated with cognitive function in T2DM, whereas metformin seemed to be protective. This study highlights education as a key marker of cognitive reserve in protecting cognition.