Metformin, Other Antidiabetic Drugs, and Risk of Alzheimer's Disease

A Population-Based Case-Control Study

Patrick Imfeld, MSc; Michael Bodmer, MD; Susan S. Jick, DSc; Christoph R. Meier, PhD


J Am Geriatr Soc. 2012;60(5):916-921. 

In This Article


The findings of this large case–control study do not provide evidence that use of metformin is associated with lower risk of developing AD. The findings even suggest that long-term use of metformin may be associated with a slightly higher risk of developing AD than nonuse of this drug, whereas such a finding was not seen for use of other antidiabetic drugs such as sulfonylureas, thiazolidinediones, or insulin. This finding supports evidence from an animal study that found that metformin increased the generation of Aβ protein, which is pivotal in the genesis of AD,[11] although the findings regarding the effect of metformin have to be interpreted with caution, because this greater risk was not confirmed in a subgroup analysis of users of metformin only and because there was no consistent trend toward a greater risk with increasing number of prescriptions.

In the current study, short-term users of insulin had a substantially lower risk of developing AD than nonusers of this drug, whereas no risk alteration was observed in long-term users. A possible explanation for this could be that individuals with DM who show signs of cognitive impairment but have not yet been diagnosed with dementia are less likely to be started on insulin therapy than individuals with DM whose cognitive abilities are not impaired and who can adhere to treatment.

The current findings are largely consistent with those of a recent study that explored the risk of developing AD in a cohort of 1,248 individuals without dementia in association with DM and glycemic control. Individuals with diagnosed DM at baseline did not have a greater risk of developing AD during follow-up, whereas those with borderline DM were at a marginally greater risk of developing AD. A subgroup of individuals with undiagnosed DM at baseline but high blood glucose levels (≥11 mmol/L) during follow-up had a greater risk of developing AD.[19] In contrast with the findings of the current study, participants with DM who were treated with insulin had a greater risk of developing AD than those without DM in the Rotterdam Study,[3] although the authors of that study stated that they could not exclude the possibility of having misclassified participants with VaD as having AD. Because DM has been clearly linked to a greater risk of developing VaD,[7,20] this misclassification may have distorted the relative risk estimates for the association between DM and AD in the Rotterdam Study.[3]

The finding of a slightly greater risk of AD in association with metformin use in this large observational study is consistent with observations from a recent in vitro study, in which metformin was found to increase the biogenesis of Aβ protein.[11] By contrast, in other in vitro studies, metformin modified important steps in the biogenesis of neuritic plaques and neurofibrillary tangles or improved impaired neuronal insulin signaling,[9,10] raising speculations about the potential to reduce the risk of developing AD, but all of these observations were made in cortical neurons of mice, and the results may not be applicable to humans.

The current study also examined the role of thiazolidinediones in the risk of developing AD and found that individuals with DM treated with these drugs had no different risk from nonusers of these drugs. In animal models of AD, thiazolidinediones have been shown to ameliorate disease-related pathology and to improve learning and memory deficits.[21] Based on these observations, the efficacy of various thiazolidinediones (mainly rosiglitazone) in improving cognitive deficits in individuals with AD has been tested in clinical trials, albeit with inconsistent findings. Although one study reported greater cognitive improvement after 6 months of rosiglitazone treatment in participants with mild AD than in placebo-treated controls,[22] another study found such an association only in individuals without the apolipoprotein E (ApoE) ε4 allele.[23] A recent Phase III trial in which participants were stratified according to ApoE ε4 status, extended-release rosiglitazone did not improve cognition in those with mild to moderate AD with or without the ApoE ε4 allele.[24]

A limitation of the current study that needs consideration is that the diagnosis of AD and of other dementia types is not straightforward, and the recording of the diagnosis in a primary care record is by definition delayed (i.e., it does not occur until after an individual has had symptoms for a certain period of time before the recording date). Thus, as with many other slowly developing degenerative diseases, the disease onset and therefore the index date in an observational study is not a precise point in time. This may affect some risk estimates, particularly if early symptoms of the diseases of interest may affect the likelihood of beginning or stopping a given drug therapy before the index date, potentially leading to spuriously low or high risk estimates for current short-term use, as may have occurred in short-term users of insulin in the present study. Long-term use of each study drug was examined to account for the unknown date of disease onset, and it was found that long-term use was not associated with risk of AD. Furthermore, some degree of outcome misclassification is likely to occur because not all dementia diagnoses can be assigned to a certain subtype with certainty. Nevertheless, it is a strength of the current study that cases were selected through use of a questionnaire and by defining a sophisticated algorithm to classify cases in the absence of any knowledge of the exposure of interest. The fact that the GPs confirmed up to 80% of all potential AD cases using accepted diagnostic criteria for an AD diagnosis corroborated the validity of this algorithm. This point is of great importance because DM is clearly associated with VaD, and significant misclassification could have spuriously increased the risk in this study.

Diagnostic bias might have played a role in the study because individuals with DM may be more likely of receiving an AD diagnosis as the result of closer follow-up by a GP than individuals without DM, although the reverse is also possible because individuals with long-standing severe DM may be less likely to be investigated for AD.

It was not possible to adjust for certain potential confounders such as ApoE ε4 status,[25] level of education,[26] or certain lifestyle factors such as physical activity[27] or dietary habits[28] because these factors are not regularly recorded in the GPRD, but BMI, which is to some degree related to physical activity and dietary habits, was adjusted for.

In summary, the findings of this large observational study do not provide evidence that use of metformin reduces the risk of developing AD but even suggest that long-term use of metformin, as opposed to use of other antidiabetic drugs, is associated with greater risk, although there was not a consistent trend with increasing number of prescriptions, and the result was not confirmed in a subgroup analysis of individuals prescribed metformin only. Long-term use of sulfonylureas, thiazolidinediones, or insulin was not associated with an altered risk of developing AD in individuals with DM.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: