We report here our prospective study of 15,224 non-diabetic, first-degree relatives of probands with immune-mediated (type 1) diabetes (IMD), of which 135 were found to eventually develop diabetes. We determined islet cell, insulin, GAD65, insulinoma-associated antigen-2 and 2βautoantibodies (ICA, IAA, GAD65A, IA-2A and IA-2βA), on the first available serum samples. The latter three autoantibodies were however assayed on subsets of the relatives with and without ICA, IAA and/or GAD65A, plus most of the relatives who developed diabetes. Of the relatives who progressed to diabetes, 94% had at least one of these autoantibodies on the first screening, while ICA proved to be the most sensitive single marker (sensitivity 74%). Risk of diabetes was however negligible when ICA was found in the absence of the others (5-year risk=5.3%), but increased dramatically whenever two or more autoantibodies were present (5-year risk=28.2% and 66.2%, respectively). The most predictive combination of markers was ICA plus IA-2A and/or IA-2β A. Loss of first phase insulin release to IVGTT also occurred only in those ICA-positive relatives who had one or more of the other autoantibodies. The data suggests that significant β-cell damage is seen only when the underlying autoimmunity has spread to multiple antigenic islet cell determinants. Combinations of the autoantibodies occurred most often in relatives with the highest risk HLA-DR/DQ phenotypes. These data document that only relatives positive for at least two or more of these five autoantibodies are at significant risk of diabetes themselves. Intervention trials for the prevention of type 1 diabetes could be designed based on testing for these autoantibodies alone, without the need for HLA typing and IVGTT testing.