Aging
Unmeasured Confounding in Studies on GLP-1 Receptor Agonist Initiation and Incident Dementia Sarah Ackley* Sarah Ackley Jingxuan Wang Kaley Hayes Andrew Zullo Minhyuk Choi Peter Buto Andrew Stokes Maria Glymour
Glucagon-like peptide-1 and glucose-dependent insulinotropic polypeptide receptor agonists (GLP-1 RAs) are new, effective drugs for diabetes and weight management. GLP-1 RAs reduce the incidence of diabetes-related complications associated with dementia (e.g., stroke) and may have direct effects on the brain. Randomized-controlled-trial evidence on GLP-1 RAs and dementia is limited. Observational studies suggest that GLP-1 RAs reduce incident dementia, but omit potential confounders like education and genetic factors. Using current and retrospective electronic health record data from All of Us cohort participants with a diabetes diagnosis (n=37,821), we evaluated whether these known dementia risk factors are associated with GLP-1 RA initiation (n=7,419 initiators). Logistic regression models for initiation of any GLP-1 RA were adjusted for age at liraglutide approval (12/23/14), sex, race/ethnicity, BMI, and genetic ancestry (only for genetic risk). Compared to those with less than a high school degree, GLP-1 RA initiation was associated with completion of high school (OR=1.14, 95% CI (1.03, 1.26)), some college (OR=1.36, 95% CI (1.24, 1.50)), and college or above (OR=1.39, 95% CI (1.26, 1.53)). Compared to non-APOE-ε4 carriers, one and two APOE-ε4 alleles were associated with 0.88, 95% CI (0.82, 0.95), and 0.93, 95% CI (0.76, 1.15), respectively, times the odds of GLP-1 RA use. A one standard deviation increase in non-APOE-ε4 Alzheimer’s disease genetic risk did not appreciably change the odds of GLP-1 RA use (OR=1.02, 95% CI (0.98, 1.05)). In conclusion, education is associated with increased risk of GLP-1 RA use, while APOE-ε4 carriers have decreased risk. These associations suggest that education and APOE-ε4 are plausible confounders that may account for a portion of GLP-1 RA-dementia associations, reducing credibility of observational studies to date.