Glycated hemoglobin and body mass index as mediators of GLP‐1RAs and Alzheimer's disease and related dementias in patients with type 2 diabetes

Huilin Tang, William T. Donahoo, Steven T. DeKosky, Yao An Lee, Mikael Svensson, Jiang Bian, Jingchuan GuoAlzheimer 2025 · DOI

TL;DR

This study investigated whether the protective effect of GLP-1 receptor agonists (GLP-1RAs) against Alzheimer's disease and related dementias (ADRD) in type 2 diabetes patients is mediated by reductions in HbA1c or body mass index (BMI). The findings indicate that GLP-1RAs significantly reduce ADRD risk, but this effect is largely independent of their impact on HbA1c and BMI, suggesting alternative neuroprotective mechanisms.

What Problem Does This Paper Address?

Alzheimer's disease and related dementias (ADRD) are a growing public health concern, particularly in older adults with comorbidities like type 2 diabetes (T2D). Glucagon-like peptide-1 receptor agonists (GLP-1RAs), commonly used to treat T2D, have shown promise in reducing the risk of all-cause dementia. However, the precise mechanisms behind this neuroprotective effect remain unclear. Previous research suggests that GLP-1RAs can lower HbA1c and BMI, both of which are implicated in dementia risk. This paper aims to fill a critical knowledge gap by using causal mediation analysis to determine the extent to which reductions in HbA1c and BMI mediate the association between GLP-1RA use and ADRD risk. Understanding these pathways is crucial for optimizing treatment strategies and potentially developing new preventative measures for ADRD.

Methods

The researchers employed an active comparator new user study design using electronic health record data from the OneFlorida+ Clinical Research Consortium. This design helps mitigate confounding by indication and time-related biases. Causal mediation analysis was the core statistical approach used. This method allows for the decomposition of the total effect of an exposure (GLP-1RA use) on an outcome (ADRD risk) into direct and indirect effects, where the indirect effect is mediated through specific variables (HbA1c and BMI reduction).

Specifically, they used a multivariate Cox proportional hazards model to build the outcome model and a multivariate linear regression model for the mediator model. These models were used to estimate hazard ratios (HRs) for the natural direct effect (effect of GLP-1RAs on ADRD risk independent of HbA1c/BMI) and the natural indirect effect (effect mediated by HbA1c/BMI reduction). The percentage mediated was also calculated. Sensitivity analyses, including E-value calculations and analyses with different definitions of mediators and subgroups, were conducted to assess the robustness of the findings.

Datasets

The study utilized electronic health record data from the OneFlorida+ Clinical Research Consortium. This repository contains de-identified patient data including demographics, diagnoses, medications, procedures, vital signs, and laboratory test results, covering the period from January 1, 2012, to June 30, 2023. No new benchmarks were introduced; the study relied on existing clinical data.

Key Insights and Ideas

  • GLP-1RAs offer broad neuroprotection: The most striking finding is that the protective effect of GLP-1RAs against ADRD is largely independent of their well-established benefits in improving glycemic control (HbA1c) and reducing body weight (BMI). This suggests that GLP-1RAs may have direct neuroprotective actions that are not solely reliant on metabolic improvements.
  • Beyond metabolic control: This challenges the assumption that the primary benefit of GLP-1RAs for brain health is solely through better diabetes management. It opens the door to exploring other biological pathways through which these drugs might exert their effects.
  • Complexity of BMI and ADRD: The study highlights the complex and sometimes contradictory relationship between BMI and ADRD risk, reinforcing the need for nuanced investigation rather than simple assumptions.
  • Causal mediation analysis is powerful: The application of causal mediation analysis provided a more precise understanding of the relationships between GLP-1RAs, metabolic factors, and ADRD risk, moving beyond simple associations.

Key Results

  • GLP-1RA users experienced significant reductions in HbA1c levels (adjusted mean difference of -0.16%) and BMI (adjusted mean difference of -0.23 kg/m²) compared to users of other glucose-lowering drugs (GLDs).
  • GLP-1RA users had a 26% lower risk of ADRD compared to other GLD users (adjusted HR = 0.74).
  • When HbA1c reduction was considered as a mediator, the natural direct effect on ADRD risk was HR 0.73, while the natural indirect effect was HR 1.02, indicating minimal mediation.
  • Similarly, when BMI reduction was the mediator, the natural direct effect was HR 0.75, and the natural indirect effect was HR 1.02, again showing minimal mediation.
  • The proportion mediated by HbA1c reduction was -5.21%, and by BMI reduction was -5.03%, with confidence intervals that included zero, further supporting the lack of significant mediation.

Limitations and Future Work

The authors acknowledge several limitations:

  • Missing Data: A substantial proportion of missing data for HbA1c (approx. 40%) and BMI (approx. 18%) is a major limitation, although multiple imputation and complete case analyses were used to mitigate this.
  • Observational Design: Despite using advanced statistical methods, the observational nature of the study inherently limits definitive causal inference. Residual confounding, such as unmeasured diabetes severity, remains a possibility.
  • Short Follow-up Period: The follow-up period might be too short to fully capture the long-term effects of GLP-1RAs on ADRD development.
  • ADRD Identification: Reliance on administrative data (CCW algorithms) for ADRD identification may lead to misclassification, especially for early or mild cases. Mild cognitive impairment (MCI) is also often underdiagnosed.
  • Generalizability: The study was conducted using data from OneFlorida+, which may not fully represent the broader U.S. population or international settings.
  • Prescription vs. Dispensing: The study used provider prescriptions, not actual dispensing or adherence data, which could affect causal inference.

Future work should focus on:

  • Mechanistic Studies: Investigating alternative neuroprotective pathways, such as direct effects on brain insulin signaling, reduction of neuroinflammation, modulation of amyloid metabolism, and improvements in mitochondrial function and vascular outcomes.
  • Long-term Cognitive Effects: Conducting studies with extended observation periods to better characterize temporal relationships.
  • Biomarker and Neuroimaging: Utilizing neuroimaging and biomarker analyses to elucidate mechanisms.
  • Other Mediators: Exploring other potential mediators, including cardiovascular and renal outcomes, inflammation markers, and AD pathology biomarkers.
  • Clinical Decision-Making: While the findings support the use of GLP-1RAs for ADRD risk reduction, the decision to prescribe should still be primarily based on approved indications for diabetes and cardiovascular risk.

Summary updated

Abstract

Glucagon-like peptide-1 receptor agonists (GLP-1RAs) were associated with reductions in glycated hemoglobin (HbA1c) and body mass index (BMI) compared to other second-line glucose-lowering drugs(GLDs). GLP-1RA users were associated with a 26% lower risk of Alzheimer's disease and related dementias (ADRD) than other GLD users. The protective effect of GLP-1RAs against ADRD in adults with type 2 diabetes (T2D) is largely independent of their effects on HbA1c and BMI.
BibTeX
@inproceedings{tang2025glycated,
  author = {Huilin Tang and William T. Donahoo and Steven T. DeKosky and Yao An Lee and Mikael Svensson and Jiang Bian and Jingchuan Guo},
  title = {{Glycated hemoglobin and body mass index as mediators of GLP‐1RAs and Alzheimer's disease and related dementias in patients with type 2 diabetes}},
  booktitle = {Alzheimer},
  year = {2025},
  doi = {10.1002/alz.70161},
}
Figure