SummaryThe value of a test for islet cell cytoplasmic antibodies together with a test for GAD6s antibodies to predict the subsequent development of diabetes over a period of 11.5 years was assessed in an open childhood population comprising 2,805 individuals. A single serum sample was obtained from each individual between 1975 and 1977 and screened for islet cell cytoplasmic antibodies for which eight individuals were positive (0.29 %). During the average follow-up period of 11.5 years, four of eight islet cell antibody positive and three islet cell antibody negative individuals developed clinical diabetes. Sera from all individuals, who were islet cell antibody positive and/ or developed diabetes (total of 11) and from 100 randomly selected control subjects were analysed for GAD65 antibodies. Six of eight islet cell antibody positive individuals were GAD65 antibody positive including all four who subsequently developed IDDM. Furthermore, one of the three islet cell antibody negative individuals who developed IDDM was GAD65 antibody positive both in 1976 and in 1989. Thus, a positive test for GAD6s antibodies alone correctly predicted diabetes in five of seven children, who developed the disease. Only one of the children, who developed diabetes was positive for insulin autoantibodies and this individual was also positive for islet cell cytoplasmic antibodies and GAD65 antibodies. One of the 100 control subjects was positive for GAD6s antibodies (1%). The results suggest that a single GAD65 antibody test may have a higher sensitivity for predicting IDDM than a test for islet cell cytoplasmic antibodies, but that a combined positive test for both antibodies increases the specificity for predicting IDDM over a period of 11.5 years. [Diabetologia (1994) 37: 917-924] Key words GAD65 antibodies, islet cell cytoplasmic antibodies, predictive value, IDDM in childhood.The destruction of pancreatic beta cells, which precedes the clinical onset of IDDM, is mediated by autoimmune mechanisms [1]. The presence of islet specific autoantibodies in the pre-diabetic period [2-4] is likely to reflect the ongoing autoimmune process, one that eventually leads to critical beta-cell depletion and insulin dependency. A major goal of diabetes research is to develop immune interventions that block or otherwise interfere with the destruction of beta cells and the development of clinical diabetes. Concomitant with this goal is the necessity for methods for early accurate identification of susceptible individuals [5]. Antibody assays that detect early signs of humoral autoimmunity associated with beta-cell destruction are obvious candidates for this purpose. Furthermore, susceptibility to develop diabetes is associated with certain MHC-class II haplotypes [6] and analysis of both HLA haplotypes and autoantibodies provide a test of high predictive value in family members of IDDM patients [3,7]. Since most new cases of [DDM involve individuals without a first degree relative with the disease [8, 9], predictive meth-