Abstract
Type 2 diabetes mellitus (T2DM) and COVID-19 are independently associated with an elevated risk of dementia, and COVID-19 survivors with T2DM may represent a clinically vulnerable population in whom metabolic, inflammatory, and vascular factors coexist. However, whether antidiabetic treatment choice is associated with incident dementia risk in this population remains unknown. In this target trial emulation using the TriNetX US Collaborative Network, we compared COVID-19 survivors aged ≥50 years with T2DM initiating glucagon-like peptide-1 receptor agonist (GLP-1RA) therapy with those initiating dipeptidyl peptidase-4 inhibitor (DPP-4i) therapy between January 1, 2020 and December 31, 2024. We applied an active-comparator, new-user design with a 90-day survivor landmark and 1:1 propensity score matching. The primary outcome was incident all-cause dementia (a composite of vascular dementia, Alzheimer's disease, and other dementias). Secondary outcomes included dementia subtypes, mild cognitive impairment, glycemic outcomes, and all-cause mortality. After matching, 25,263 patients per group were followed for up to 6 years (median, 919 days [GLP-1RA] vs 955 days [DPP-4i]). GLP-1RA initiation was associated with a lower hazard of incident dementia (hazard ratio [HR], 0.64; 95% confidence interval [CI], 0.58-0.72), vascular dementia (HR, 0.67; 95% CI, 0.52-0.85), and Alzheimer's disease (HR, 0.74; 95% CI, 0.59-0.93), as well as more favorable glycemic control and lower all-cause mortality (HR, 0.63; 95% CI, 0.59-0.68). The estimate for mild cognitive impairment was borderline (HR, 0.87; 95% CI, 0.75-1.00) and not consistently supported across sensitivity analyses. The association with incident dementia remained evident after the 6-month landmark analysis (HR, 0.66; 95% CI, 0.58-0.75) and was consistent across sensitivity and subgroup analyses. Among COVID-19 survivors with T2DM, GLP-1RA initiation was associated with a lower hazard of incident dementia than DPP-4i initiation. These hypothesis-generating results warrant confirmation before they can be used to inform clinical decision-making.