The aim of this workshop was to assess the ability of individual autoantibody (ab) assays and their use in combination to discriminate between type 1 diabetic and control sera. Coded aliquots of sera were measured in a total of 119 assays by 49 participating laboratories in 17 countries. The sera were from 51 patients with new onset type 1 diabetes and 101 healthy control subjects with no family history of diabetes. In the final analysis, data on diabetic sera were restricted to 43 subjects younger than age 30 years. The laboratories were asked to report results for these sera using their currently available anti-islet autoantibody assays. In addition, they were asked to combine information from their assays to classify sera as having high, moderate, or low probability of originating from a patient with type 1 diabetes. Actual strategies for combining assays were determined by each laboratory. There were no significant differences in sensitivity among 19 radioimmunoassays (RIAs) for IA-2 autoantibodies (cytoplasmic islet cell antibody [ICA] 512) using different constructs that included the intracellular portion of the molecule (mean sensitivity 73%). However, an enzyme-linked immunosorbent assay (ELISA) using the extracellular portion of the IA-2 molecule did not discriminate between diabetic and control sera. Among GAD autoantibody assays that achieved sensitivity >70%, 26 were RIAs and one was an ELISA. When the sera were ranked according to their autoantibody levels, the concordance for insulin autoantibodies (IAAs) in different laboratories was markedly less than for IA-2ab and GADab. Using a combination of autoantibody assays, several laboratories achieved excellent discrimination between diabetic and control sera (sensitivity up to 80% with false-positive rate of 0%). A variety of strategies for combining information from different assays were successful (e.g., those including and excluding ICA), and no one strategy emerged as clearly superior. In conclusion, IA-2/ICA512 autoantibodies are a marker of type 1 diabetes and can be measured consistently by most assays. Several different strategies for combining assays achieved high sensitivity with a low false-positive rate.