Multiple islet autoantibody positivity is currently believed to best predict progression to Type I (insulin-dependent) diabetes mellitus. We compared its predictive value with that of positivity for a particular type of islet autoantibody, directed against the IA-2 antigen. Autoantibodies against islet cell cytoplasm (ICA), insulin (IAA), GAD (GADA) and IA-2 (IA-2A) were measured at initial sampling in 1724 non-diabetic siblings (median age range:16 0-39 years) of Type I diabetic patients with a median follow-up of 50 months. On initial sampling 11% of siblings were positive for one antibody type or more and 2.1% for three of more types. During follow-up, 27 antibody-positive siblings developed diabetes. Using survival analysis, the risk for clinical onset within 5 years was 34% in subjects positive for three or more types compared with 13% in those with one type or more. Progression to diabetes amounted to 12% within 5 years among siblings positive for IAA, 20% for ICA, 19% for GADA but 59% for IA-2A (p<0.001 vs absence of the respective antibody). IA-2A were detected in 1.7% of all siblings and in 56% of the prediabetic subjects on first sampling. Initial positivity for two or three antibody markers was associated with a higher progression rate in IA-2A positive as compared to IA-2A negative siblings (p=0.001). In absence of IA-2A initial positivity for another antibody (IAA, ICA or GADA) conferred a low (<10% within 5 years) risk of diabetes compared to subjects lacking this antibody. In siblings of Type I diabetic patients, IA-2A positivity is a more direct predictor of impending clinical onset than multiple antibody positivity per se. Assessment of IA-2A status allows us to select subjects with homogeneously high risk of diabetes for participation in prevention trials.