A total of 85 islet cell antibody (ICA)+ or insulin autoantibody (IAA)+ relatives of patients with type 1 diabetes have been followed as part of the Seattle Family Study for a mean of 2.8 years. Of the subjects followed, 10 developed diabetes during this time period. The presence of GAD antibodies was strongly associated with the development of diabetes. In contrast, the presence of IAAs did not influence the risk of diabetes among ICA+ GAD+ subjects. When either the initial absolute acute insulin response to glucose (AIRg) or the AIR percentile, which accounts for the individual's insulin sensitivity, was below the 10th percentile of normal subjects, the risk of diabetes approached 50% at 5 years. However, impaired beta-cell function did not influence the risk of diabetes among those who were GAD+. There were 13 subjects with low AIRg and 13 subjects with two or more antibodies who had not progressed to diabetes during the course of the study. Other measurements of beta-cell function or demographic characteristics were not different in this group of nonprogressors compared with those with low AIRg who did not progress to diabetes. We conclude that ICA+ relatives with GAD antibodies or low AIRg have a high risk for development of diabetes, but among ICA+ GAD+ relatives, the addition of IAA or a single determination of AIRg does not enhance the prediction of diabetes. We suggest that prediction of diabetes risk depends on both the type and the number of antibodies present. In addition, there are a group of ICA+ relatives with low AIRg and/or multiple antibodies who have not progressed to diabetes over the course of the study.