Abstract
INTRODUCTION: Although poor glycemic control is associated with dementia, it is unknown if variability in glycemic control, even in those with optimal glycosylated hemoglobin A(1c) (HbA(1c)) levels, increases dementia risk. METHODS: Among 171,964 people with type 2 diabetes, we evaluated the hazard of dementia association with long-term HbA(1c) variability using five operationalizations, including standard deviation (SD), adjusting for demographics and comorbidities. RESULTS: The mean baseline age was 61 years (48% women). Greater HbA(1c) SD was associated with greater dementia hazard (adjusted hazard ratio = 1.15 [95% confidence interval: 1.12, 1.17]). In stratified analyses, higher HbA(1c) SD quintiles were associated with greater dementia hazard among those with a mean HbA(1c) < 6% (P = 0.0004) or 6% to 8% (P < 0.0001) but not among those with mean HbA(1c) ≥ 8% (P = 0.42). DISCUSSION: Greater HbA(1c) variability is associated with greater dementia risk, even among those with HbA(1c) concentrations at ideal clinical targets. These findings add to the importance and clinical impact of recommendations to minimize glycemic variability. HIGHLIGHTS: We observed a cohort of 171,964 people with type 2 diabetes (mean age 61 years). This cohort was based in Northern California between 1996 and 2018. We examined the association between glycosylated hemoglobin A(1c) (HbA(1c)) variability and dementia risk. Greater HbA(1c) variability was associated with greater dementia hazard. This was most evident among those with normal-low mean HbA(1c) concentrations.