Incident dementia risk among patients with type 2 diabetes receiving metformin versus alternative antihyperglycaemic agents: a retrospective cohort study using UK primary healthcare records.
Problem
4.2 million individuals have type 2 diabetes in the UK and the prevalence is increasing due to physical inactivity, obesity and population ageing. Diabetes is a key risk factor in later life for dementia, but it is unclear whether this association is modifiable with antihyperglycaemic therapies like metformin.
Approach
This historical cohort study used UK electronic health records from the Clinical Practice Research Datalink. We followed individuals diagnosed with diabetes aged ≥40 years-old with no history of dementia or cognitive impairment from first prescription of metformin or alternative oral antihyperglycaemic drug (index date) between 1990 and 2019. We identified Read-coded diagnoses of incident all-cause dementia (primary outcome) and mild cognitive impairment (MCI) (secondary outcome). Our single-failure survival analysis used Cox regression with age as the underlying time scale and adjusted for all available confounders. We assessed for interaction with age and gender using the Likelihood Ratio Test (LRT), and conducted sensitivity analyses restricting to individuals initiating therapy post 2004 and 2012 and excluding early dementia diagnoses up to 2 years post index date.
Findings
We identified 211,396 eligible individuals (57.2% male; median age 63, IQR 54-71 years) with a median follow-up of 5.4 years. 179,333 (84.8%) were metformin users, who were younger and more likely to be overweight, but had less recorded cardiovascular disease. Metformin users experienced a lower risk of incident dementia and MCI, adjusted HR (aHR) 0.86 (95% CI 0.79 – 0.94, n 146,266) and aHR 0.92 (95% CI 0.86 – 0.99, n 146,883) respectively. We found evidence of effect modification by age-group for dementia (LRT p 0.03) and MCI (LRT p 0.0014): metformin users aged 40-79 years experienced a meaningful relative risk reduction of dementia with HR 0.77 (95% CI 0.68 – 0.85, n 136,770) compared with older individuals with an estimated null effect HR 0.95 (95% CI 0.87 – 1.05, n 26,126). Estimates from sensitivity analyses were consistent with the main analysis. We identified potential exposure misclassification because metformin use was so ubiquitous: 95.3% of all participants were prescribed metformin at least once.
Consequences
We confirmed that metformin use is associated with a lower risk of incident dementia and MCI in adults with diabetes compared with alternative antihyperglycaemic therapy in a large, demographically-representative UK population. To our knowledge, this is the largest observational study to date, but we acknowledge limitations caused by missing data, exposure misclassification and confounding by indication. Our finding of effect modification by age is consistent with current literature. This protective association has the potential to contribute to global dementia prevention strategies and merits further investigation with robust observational studies to strengthen causal inference. Possible strategies include propensity scoring, investigation for dose-response relationships for metformin and assessment of differences in patient characteristics driving possible confounding by indication.