This DPP/DPPOS randomized follow-up evaluated associations between initial randomization to metformin versus placebo or intensive lifestyle intervention (ILS) and later cognitive outcomes. Adults at high risk for type 2 diabetes were enrolled at 27 U.S. centers. Cognitive impairment syndromes were adjudicated in 1,483 participants (median age 74, IQR 68–80) during 2022–2024 using the National Alzheimer’s Coordinating Center Uniform Dataset version 3. Repeated cognitive testing was performed between 2009 and 2024. Multinomial logistic regression and mixed-effects models examined links between trial arm and both adjudicated syndromes and longitudinal test trajectories.
Participants were originally enrolled in the Diabetes Prevention Program (1996–1999) and followed in the Diabetes Prevention Program Outcomes Study. The follow-up included prospective long-term assessments across 27 U.S. centers. Cognitive adjudication occurred in 2022–2024 and repeated cognitive testing spanned 2009–2024. The adjudicated sample numbered 1,483 with a median age of 74 years (IQR 68–80) at adjudication.
Initial randomization was to metformin, placebo, or an intensive lifestyle intervention (ILS) for a mean randomized treatment period of about 3.2 years. Participants in the originally randomized metformin group received open-label metformin beyond the randomized phase and accumulated substantially greater total metformin exposure. Reported mean total metformin exposure was 15.5 ± 7.7 years/person for the metformin-randomized group. Many participants in the placebo and ILS arms later received out-of-study metformin, typically after diabetes onset, with mean total exposure of 4.5 ± 5.1 years/person in placebo and 3.8 ± 4.8 years/person in ILS.
Adjudication of cognitive impairment syndromes used the National Alzheimer’s Coordinating Center Uniform Dataset version 3 procedures during 2022–2024. In addition to adjudicated syndromes (including dementia), repeated cognitive tests assessed domain-specific performance, with particular focus reported on executive and memory domains. The methods for adjudication and the cognitive measures are those described in the DPPOS protocols and applied consistently across participants in the adjudicated sample.
The investigators used multinomial logistic regression to compare distributions of adjudicated cognitive syndromes across randomized groups and mixed-effects models to examine longitudinal cognitive test trajectories. Analyses adjusted for demographic variables, education, income, and APOE-ε4 genotype where indicated. Effect estimates and confidence intervals are reported for primary comparisons between randomization to metformin versus placebo and versus ILS.
Although the overall frequency distributions of adjudicated cognitive syndromes did not differ significantly by randomized treatment group, randomization to metformin was associated with lower odds of dementia after covariate adjustment. Specifically, the odds of dementia were lower by approximately 60% when comparing metformin versus placebo (OR 0.40; 95% CI 0.17–0.97) and by approximately 62% when comparing metformin versus ILS (OR 0.38; 95% CI 0.16–0.89). In longitudinal cognitive testing, randomization to metformin was associated with better memory performance over time (β = 0.58; 95% CI 0.09–1.1; p = 0.02), with a small effect size reported (Cohen’s d = 0.1).
The authors report that long-term metformin exposure in people with prediabetes or type 2 diabetes was associated with reduced dementia risk and modestly better memory trajectories. They note that estimates are imprecise because the number of dementia cases was limited, and that longer follow-up with more dementia events is needed to confirm these associations. The manuscript is a preprint and has not been peer reviewed; the authors caution that findings should not yet guide clinical practice.
Several authors disclosed relationships with industry and funding sources; specific disclosures are reported in the preprint. The work is part of registered clinical trials: Diabetes Prevention Program (NCT00004992), Diabetes Prevention Program Outcomes Study (NCT00038727), and a DPPOS AD/ADRD registration (NCT05704309). Ethical approvals and participant informed consent were obtained at participating centers according to the Declaration of Helsinki and local institutional review boards.
Within the limits of the adjudicated sample and available dementia events, randomization to long-term metformin was associated with lower odds of adjudicated dementia and better memory test performance over time compared with placebo and ILS. The authors emphasize the need for continued follow-up and additional events to provide more precise estimates and to confirm these preliminary findings.