Background: Evidence of an increased risk of dementia with insulin use in type 2 diabetes is weakened by confounding by indication and disease severity.
Objectives: To reassess the association between insulin use and dementia, while accounting for confounding through design and analysis.
Methods: We conducted a cohort study using administrative healthcare data from British Columbia, Canada housed by Population Data BC and identified patients with a new diagnosis of type 2 diabetes (1998-2016). To adjust for confounding through design, from a sub-cohort of those who received two non-insulin antihyperglycemic classes, we identified new users of insulin compared to new users of a non-insulin class. We adjusted for confounding using: (1) conventional multivariable adjustment and (2) inverse probability of treatment weighting (IPTW) based on the high-dimensional propensity score algorithm. The hazard ratio [HR] (95% confidence intervals [CI]) of dementia was estimated using cause-specific hazards models with death as a competing risk.
Results: We identified 414,089 patients with type 2 diabetes, of whom 33,093 received two antihyperglycemic classes and initiated a third class (7,863 insulin and 25,230 non-insulin) between ages 40-70 years. A total of 257 dementia events occurred over 4.5-year median follow-up. The HR (95% CI) was 1.68 (1.29-2.20) before adjustment and 1.39 (1.05-1.86) after multivariable adjustment, which was further attenuated to 1.14 (0.81-1.60) after IPTW weighting.
Conclusions: After accounting for confounding through design and analysis, initiation of insulin after two antihyperglycemic classes was not associated with an increased risk of all-cause dementia, providing reassurance to prescribers and patients with type 2 diabetes.
