To investigate the association between sodium-glucose cotransporter 2 (SGLT2) inhibitors and the likelihood of incident all-cause dementia in older patients with psychiatric disorders.
Approach:
Study Design: A retrospective cohort study using US Department of Veterans Affairs electronic health record data from January 1, 2016, to June 1, 2024.
Population: Included patients aged 65 years or older with at least two diagnostic codes for major depressive disorder, bipolar disorder, or schizophrenia spectrum disorder, excluding those with prior dementia diagnosis or SGLT2 inhibitor use.
Outcomes: Primary outcome was incident all-cause dementia; secondary outcomes included time to psychiatric emergency department utilization and hospitalization.
Analysis: Used marginal structural models with inverse probability weighting to account for treatment and censoring.
Key Findings:
In the intention-to-treat analysis, SGLT2 inhibitor initiation was associated with 39% lower odds of all-cause dementia.
SGLT2 inhibitor initiation was associated with 20% lower odds of psychiatric emergency department visits.
In the per-protocol analysis, sustained SGLT2 inhibitor use was associated with a 46% lower odds of all-cause dementia and psychiatric hospitalization.
Estimated 5-year dementia risk was 1.6% for SGLT2 initiators compared to 2.9% for noninitiators.
Interpretation:
Limitations:
Potential residual confounding related to psychiatric disorder severity and baseline cognitive function.
Misclassification due to reliance on diagnosis and prescription codes.
Generalizability may be limited due to the predominantly male cohort from the Veterans Affairs health care system.
This week's research makes one thing clear: who someone is before they get sick — their relationships, their partner's health, the back of their eye — is doing a lot of work medicine is only beginning to account for.
Plasma proteomic models of more than 40 cell types were associated with incident Alzheimer's disease, amyotrophic lateral sclerosis, cancer, and mortality across three large cohorts.
