Type 1 Diabetes and Incident Dementia: An Analysis in the All of Us Cohort

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: T1DM: HR 2.82, T2DM: HR 2.08 compared to no diabetes

• Primary Outcome: Incident dementia

• Nnt Or Sensitivity: T1DM classification algorithm: sensitivity 0.59-0.76, specificity 0.79-0.90 depending on cutoff

• Confidence Interval: T1DM: 95% CI 2.28-3.48, T2DM: 95% CI 2.87-2.31

Clinical Application

Risk assessment feasible using existing clinical data, but requires longer follow-up for meaningful prognostication Relevant to Australian diabetes care given similar healthcare systems. Could inform RACGP diabetes guidelines and specialist referral patterns. PBS-subsidised diabetes medications may facilitate better glycaemic control to potentially reduce dementia risk Adults aged 50+ with type 1 or type 2 diabetes in primary care and endocrinology settings

Abstract

BACKGROUND AND OBJECTIVES: Although diabetes mellitus (DM) is a well-established determinant of dementia risk, most studies have evaluated type 2 DM (T2DM) or any DM. The influence of type 1 DM (T1DM) on dementia risk remains unclear. We evaluated associations of T1DM and T2DM, separately, with incident dementia using linked electronic health records (EHRs). METHODS: This prospective cohort study used previously collected survey and EHR data from the All of Us (AoU) cohort, a convenience sample of US adults. Eligible participants were 50 years or older and completed baseline surveys. Enrollment began in 2017, with data available through October 2023, along with records before enrollment. Mean follow-up from baseline was 2.4 years. We developed an algorithm to distinguish DM type based on count of T1DM encounters. This algorithm was validated against 2 reference measures: self-reported diabetes type and C-peptide values. Using AoU data, we classified participants as having no DM, T1DM, or T2DM. We ascertained incident dementia using ICD-9, ICD-10, and Systematized Nomenclature of Medicine codes in participants' EHRs. We estimated hazard ratios (HRs) and 95% CIs for the association of diabetes type with incident dementia using Cox proportional hazards models. RESULTS: Among 283,772 participants (mean [SD] age 64.62 [8.96] years; 56.7% women), 60.3% identified as non-Hispanic White and 13.3% as Hispanic/Latino. Optimal DM classification algorithm cutoffs varied by reference standard: self-reported diabetes: ≥1 T1DM EHR encounter (sensitivity: 0.59; specificity: 0.90) and C-peptide: ≥3 T1DM EHR encounters (sensitivity: 0.76; specificity: 0.79). Defining T1DM as having ≥1 T1DM encounter, 5,442 participants had T1DM. Compared with those without DM, participants with T1DM had higher dementia incidence (sociodemographic-adjusted HR 2.82; 95% CI 2.28-3.48) and those with T2DM also had elevated risk (sociodemographic-adjusted HR 2.08; 95% CI 2.87-2.31). Results were similar across sex, race, and ethnicity strata. DISCUSSION: In AoU, DM was associated with elevated dementia risk, with the highest risk among those with T1DM. These findings highlight the need to better understand mechanisms tying T1DM to dementia.

References

1. 1.Pederson, A. M., Buto, P., Zimmerman, S. C., Velez, M., Sims, K. D., Murchland, A. R., Wang, J., Brennan, A. T., Glymour, M. M., & Weuve, J. (2026). Type 1 diabetes and incident dementia: An analysis in the All of Us cohort. Neurology. https://doi.org/10.1093/aje/kwaf118