There is increasing interest in the use of glutamic acid decarboxylase antibodies (GADAbs) for identification of subjects at increased risk of developing insulin-dependent diabetes mellitus (IDDM). However, considerable variation exists between laboratories in the reported frequency of GADAb in various clinical situations, and disease sensitivity and specificity have not yet been compared between assays. An international workshop was held in which 101 coded freeze-dried sera, including 39 from subjects with newly diagnosed IDDM, 32 from healthy control subjects, 4 from nondiabetic subjects with Graves' disease, and 4 from islet cell antibody–positive subjects, were analyzed in 52 assays (radiobinding assay [RBA], 26; enzyme-linked immunosorbent assay [ELISA], 19; and enzymatic immunoprecipitation assay [EIP], 7). The mean sensitivity for RBAs (76.2%) was higher than for ELISAs (36.5%) and EIPs (49.9%) (P < 0.01). The mean specificity was similar for each assay format (RBA, 89.4%; ELISA. 89.4%; and EIP, 92.3%). The lower sensitivities of the ELISA and EIP were predominantly due to the inability of these assays to detect low levels of GADAb in IDDM. To convert results to standard units, standard curves were constructed using duplicate dilutions of the anti–glutamic acid decarboxylase monoclonal antibody MICA 3 and serum from a patient with stiff-man syndrome (SMS). Curves could be derived in 28 assays using the MICA 3 serum and in 29 using the SMS serum. The mean coefficients of variation between assays for disease and control samples were 45% when results were converted to MICA units, 77% for SMS units, and 76% for SD scores. The relatively lower scatter observed using the MICA 3 serum as a standard indicates that this serum may be useful as an international reference. Because assays for GADAb with a high level of sensitivity and specificity are now available, prospective studies are now required to establish the role of GADAbin clinical applications such as preclinical screening for IDDM.