Dual-parameter model for prediction of type I diabetes mellitus.

The recent cloning and recombinant expression of novel islet autoantigens [glutamic acid decarboxylase (GAD) 65 and islet-cell autoantibody 512 (ICA512)] has made possible the determination of whether the quantitative expression of autoantibodies to these molecules is correlated with age of diabetes onset and rate of progression to diabetes, similar to insulin autoantibodies (IAAs). We measured autoantibodies reacting with GAD65 (GAD65AA), ICA512 (ICA512AA), and insulin in patients who recently had received a diagnosis of diabetes and in first-degree relatives prospectively identified and then followed because of the expression of high titers of ICA. Levels of IAAs (but not GAD65AA or ICA512AA) correlated inversely with age at diagnosis of diabetes and directly with time to diabetes onset among the ICA-positive relatives. In multiple linear regression models, the level of IAAs remained a significant predictor of the time to diabetes after allowing for first-phase insulin secretion. The unique and dramatic association of IAAs with progression to diabetes suggests that IAAs contribute directly to disease pathogenesis or that levels of IAAs are influenced uniquely by the process, leading--at different rates in different prediabetic individuals--to type I diabetes. In addition, the linear regression model described (involving two variables, first-phase insulin secretion and levels of IAAs) aids in the prediction of time to diabetes among ICA-positive relatives.