To determine whether the predictive value of islet cell antibodies (ICA) and insulin autoantibodies (IAA) is increased by measurement of glutamic acid decarboxylase antibodies (GADAb) in first-degree relatives of patients with insulin-dependent diabetes mellitus (IDDM), we measured GADAb in those developing IDDM and in relatives found to be ICA- or IAA-positive in our family screening study. First-degree relatives (n = 2904) were followed for 2.4 (median, range 0.04-5.8) years. Of the subjects developing IDDM, 11/14 (78%) had ICA > or = 20JDF units, 1/14 (7%) had IAA > or = 100 nU/ml and 6/14 (43%) had GADAb (> or = 460 nU/ml, measured by precipitation of enzymatic activity). Of the four subjects with ICA < 20 and IAA < 100 nU/ml who developed IDDM, one had elevated GADAb. Significant inhibition of GAD enzymatic activity by serum immunoglobulins, a potential cause of false-negative results in our immunoprecipitation assay, was not detected in seven subjects who developed IDDM in the absence of GADAb. Sixty-nine of the 2904 subjects with ICA > or = 20 or IAA > or = 100 were followed for 3.1 (median range 0.1-5.4) years. Survival analysis showed that diabetes-free survival in this group was not influenced significantly by GADAb positivity. In conclusion, GADAb in the absence of ICA and IAA are uncommon in first-degree relatives who progress to IDDM and the presence of GADAb does not increase the risk for IDDM in ICA- or IAA-positive relatives.