Baseline data from the Horizon 2020–funded RECOGNISED study were analyzed to examine screening strategies for mild cognitive impairment (MCI) in older adults with type 2 diabetes. Eligibility criteria included age ≥65 years, type 2 diabetes duration >5 years, no prior stroke or neurodegenerative disease, and no overt or only mild-to-moderate non-proliferative diabetic retinopathy. MCI and normocognition (NC) were defined using a comprehensive neuropsychological test battery and Clinical Dementia Rating. The study was published in Diabetologia. 

A total of 313 participants were included (128 NC; 185 MCI). Compared with NC, those with MCI were older (p=0.006), had fewer years of education (p<0.001), lower retinal sensitivity (p=0.01), and reduced gaze fixation capacity (p≤0.001). Differences were also observed in pupillary area ratio (p=0.002) and photopic b-wave amplitude (p=0.03). 

The optimal multivariable logistic regression model combined retinal sensitivity, gaze fixation, photopic b-wave amplitude, pupillary size change, years of education, Diabetes-Specific Dementia Risk Score (DSDRS), and Montreal Cognitive Assessment (MoCA), yielding an AUC of 0.84 with 79.9% sensitivity and 79.0% specificity. Visuo-construction impairment was most prominent and independently associated with retinal sensitivity and gaze fixation.

The cross-sectional design precludes causal inference, and inclusion was limited to individuals without advanced retinopathy. Social determinants beyond education were not assessed.

These findings suggest that combining retinal neurodysfunction markers with simple clinical variables may support identification of MCI in older adults with diabetes. Longitudinal validation is required to determine predictive value and broader applicability.