Early metformin initiation was not associated with incident dementia: a population‐based cohort study

Abstract

AbstractBackgroundMetformin is the first‐line pharmacotherapy for type 2 diabetes. Although metformin was previously suggested to prevent dementia, most existing epidemiologic studies did not adequately address immortal time bias or confounding due to disease severity. We aim to determine the association of metformin with dementia risk using various bias mitigation strategies.MethodThis population‐based retrospective cohort study utilized data housed at ICES (https://www.ices.on.ca/). Using a validated algorithm for diabetes (Specificity = 99.1%, PPV = 91.4%), residents of Ontario, Canada ≥66 years newly diagnosed with diabetes from January 1, 2008 to December 31, 2017 entered the cohort. We excluded potential type 1 diabetes or advanced diabetes cases by excluding people with insulin as the first received glucose‐lowering medication. To select people at an early disease stage who could either initiate metformin or start lifestyle interventions alone, we excluded people with hemoglobin A1c <6.5% or ≥8.0% or estimated glomerular filtration rate (eGFR) <45 mL/min/1.73m2. Using a landmark analysis to address immortal time bias, the exposure was classified into “metformin monotherapy initiation within 180 days after new diabetes diagnosis” or “no glucose‐lowering medications within 180 days”. To mitigate reverse causality, a 1‐year lag period was applied to the end of the 180‐day landmark period, and therefore follow‐up started 1.5 year after cohort entry. Incident dementia was ascertained using a validated algorithm for Alzheimer’s disease and related dementias (Sensitivity = 79.3%, Specificity = 99.1%). Adjusted hazard ratios (aHR) and confidence intervals (CI) were obtained from propensity‐score weighted Cox proportional hazard models.ResultCompared to no or delayed metformin initiation (N = 22,369; 1,768 events; 117,414 person‐years), metformin initiation within 180 days (N = 12,331; 978 events; 65,761 person‐years) showed no association (aHR [95% CI] = 1.047 [0.956‐1.147]) with dementia risk over mean follow‐up of 5.28 years. The sensitivity analysis with a 1‐year landmark period also showed no association (aHR [95% CI] = 1.040 [0.942‐1.148]) when comparing metformin initiation within 1 year (N = 13,902; 1,047 events; 68,404 person‐years) to no or delayed initiation (N = 19,078; 1,433 events; 91,917 person‐years).ConclusionEarly metformin initiation was not associated with a lower dementia risk. The present study does not support metformin as a drug to prevent dementia.