Metformin vs sulfonylurea use and risk of dementia in US veterans aged ≥65 years with diabetes

Abstract

Objective: To determine whether metformin is associated with a lower incidence of dementia than sulfonylureas.

Methods: This was a retrospective cohort study of US veterans ≥65 years of age with type 2 diabetes who were new users of metformin or a sulfonylurea and had no dementia. Follow-up began after 2 years of therapy. To account for confounding by indication, we developed a propensity score (PS) and used inverse probability of treatment weighting (IPTW) methods. Cox proportional hazards models estimated the hazard ratio (HR) of incident dementia.

Results: We identified 17,200 new users of metformin and 11,440 new users of sulfonylureas. Mean age was 73.5 years and mean HbA1c was 6.8%. Over an average follow-up of 5 years, 4,906 cases of dementia were diagnosed. Due to effect modification by age, all analyses were conducted using a piecewise model for age. Crude hazard ratio [HR] for any dementia in metformin vs sulfonylurea users was 0.67 (95% confidence interval [CI] 0.61-0.73) and 0.78 (95% CI 0.72-0.83) for those <75 years of age and ≥75 years of age, respectively. After PS IPTW adjustment, results remained significant in veterans <75 years of age (HR 0.89; 95% CI 0.79-0.99), but not for those ≥75 years of age (HR 0.96; 95% CI 0.87-1.05). A lower risk of dementia was also seen in the subset of younger veterans who had HbA1C values ≥7% (HR 0.76; 95% CI 0.63-0.91), had good renal function (HR 0.86; 95% CI 0.76-0.97), and were white (HR 0.87; 95% CI 0.77-0.99).

Conclusions: After accounting for confounding by indication, metformin was associated with a lower risk of subsequent dementia than sulfonylurea use in veterans <75 years of age. Further work is needed to identify which patients may benefit from metformin for the prevention of dementia.

Figure 1. Study scheme: Beginning with all US veterans who were ≥65 years of age and new users of single therapy of metformin or sulfonylureas in 2001–2012 (n = 432,922)

^aHad valid data for race, body mass index, renal function, and hemoglobin A1c. Those with diabetes complications were excluded. ^bDid not cross over to the other study drug and were not lost to follow-up. VA = Veterans Administration.

Figure 2. Standardized difference percentage of covariates before and after inverse probability of treatment weighting

^aEstimated glomerular filtration rate (eGFR) by measured laboratory value or ICD-9 code. ^bTIA, stroke, cardiovascular disease. ^cMacular degeneration, glaucoma, impaired vision. BMI = body mass index; HbA1c = hemoglobin A1c; ICD-9 = International Classification of Diseases–9.