Normal but increasing hemoglobin A1c levels predict progression from islet autoimmunity to overt type 1 diabetes: Diabetes Autoimmunity Study in the Young (DAISY)

Objective: The aim of this study was to assess the utility of hemoglobin A1c (HbA1c) measurements in the early detection of clinical type 1 diabetes during prospective follow-up of children with islet autoimmunity. Methods: The Diabetes Autoimmunity Study in the Young (DAISY) has followed for development of islet autoimmunity and diabetes general population newborns carrying human leukocyte antigen (HLA) genotypes conferring risk for type 1 diabetes and young siblings or offspring of people with type 1 diabetes. Testing for autoantibodies was performed at least once in 2234 and twice or more in 1887 children. Among the latter, 100 children tested positive on at least two consecutive visits. To date, 92 children have developed persistent islet autoantibodies to glutamic acid decarboxylase 65 (GAD65), insulin, or insulinoma associated antigen-2 (IA-2) and had at least two subsequent clinic visits. These children had study visits with point-of-care testing for HbA1c and random glucose every 3-6 months and those with random plasma glucose above 11.1 mmol/L or HbA1c above 6.3% were evaluated by a pediatric endocrinologist for clinical diabetes. Results: During a mean follow-up of 3.4 yr from onset of autoimmunity, 28 children developed type 1 diabetes, at mean age of 6.5 yr. Mean prediagnostic HbA1c was 5.1% [standard deviation (SD) = 0.4%]. In a Cox regression model accounting for changes in values in individuals over time, increase in HbA1c predicted increased risk of progression to type 1 diabetes, hazard ratio = 4.8 (95% confidence interval 3.0-7.7) for each SD of 0.4%, independent of random glucose and number of autoantibodies. Increase in random plasma glucose levels only marginally predicted risk of progression (hazard ratio = 1.4, 95% confidence interval 1.02-1.8, per SD of 1.1 mmol/L). Conclusions: Normal but increasing Hb1Ac may be useful in early detection of type 1 diabetes, whereas random plasma glucose levels were less predictive.