Predicting type 1 diabetes mellitus (T1DM) is a prerequisite for disease prevention. Prediction is currently performed on three levels, which include the genetic susceptibility for disease, the identification of preclinical T1DM by way of circulating islet autoantibodies, and the use of metabolic tests to stage preclinical disease into late or early prediabetes. Combinations of genetic markers such as HLA genotype, INS genotype, and if and how much family history of T1DM is present can stratify disease risk more than 1000-fold, and can be used for selection of first-degree relatives of patients with T1DM for primary intervention trials. Measurement of autoantibodies in genetically at-risk subjects identifies future cases of T1DM. Further stratification of diabetes risk in autoantibody-positive subjects can be made on the basis of autoantibody characteristics that correspond to the magnitude of the autoantibody response.