Prediction of type I diabetes in first-degree relatives using a combination of insulin, GAD, and ICA512bdc/IA-2 autoantibodies

Islet cell antibodies (ICAs) are predictive of type I diabetes in first-degree relatives, but this immunohistochemical assay has proven difficult to standardize. As an alternative, we assessed the use of radioassays for antibodies against three molecularly characterized islet autoantigens, including ICA512bdc (amino acid residues 256-979 of the IA-2 molecule, incorporating the intracellular domain). We measured insulin autoantibodies (IAAs), GAD autoantibodies (GAAs), and ICA512bdc autoantibodies (ICA512bdcAAs) by radioassay, in addition to ICAs, in 882 first-degree relatives of patients with type I diabetes, 50 of whom later developed diabetes with a median follow-up of 2.0 years (maximum 11.3 years). The cutoff for each radioassay was determined by testing >200 control subjects. When autoantibody frequencies among the relatives were analyzed according to relationship to the proband, the offspring of diabetic fathers had a higher frequency of ICA512bdcAAs (P = 0.008), IAAs (P = 0.0001) and GAAs (P = 0.0001) than the offspring of diabetic mothers. ICA512bdcAAs and IAAs both showed a significant association with HLA-DR4-DQ8 (P = 0.0005). Among relatives developing diabetes, 98% had one or more of IAAs, GAAs, or ICA512bdcAAs, and 80% had two or more of these autoantibodies, compared with none of the control subjects. Using survival analysis to allow for different lengths of follow-up, there was a significant increase in the risk of diabetes with the number of these autoantibodies present, comparing zero, one, two, and three autoantibodies (P < 0.0001, log-rank test), and by Cox regression analysis, this was independent of ICAs and age. For relatives with two or more of these autoantibodies, the risk of diabetes within 3 years was 39% (95% CI, 27-52) and the risk within 5 years was 68% (95% CI, 52-84). Relatives with all three autoantibodies had a risk within 5 years estimated to be 100%. The presence of low first-phase insulin release further increased the risk for relatives with one or two autoantibodies. We conclude that the presence of two or more autoantibodies (out of IAAs, GAAs, and ICA512bdcAAs) is highly predictive of the development of type I diabetes among relatives.