Islet cell antibodies (ICA) in the sera of nondiabetic relatives of patients with insulin-dependent diabetes (IDD) are predictive of the disease, a finding that permits the design of intervention strategies to prevent it. However, 85% or more of patients with new onset IDD have no affected relative. We therefore screened 9,696 schoolchildren between the ages of 5 and 18 yr (mean age 10.7 yr) in Pasco County, Florida for ICA in three surveys during 1984/5, 1987/8, and 1990/1 and have followed them prospectively. Approximately 4,000 of these children have been followed for nearly 8 yr. ICA titers > or = 10 Juvenile Diabetes Foundation units on replicate tests were detected in 57 of the children (0.59%). 10 children have developed diabetes so far, and all had ICA detected beforehand. The likelihood of developing IDD among the ICA-positive children was compared with 2,959 age-matched nondiabetic first degree relatives of IDD probands who were screened for ICA by our laboratory during the same time period and also followed prospectively. Of 103 (3.5%) ICA-positive relatives, 31 have developed IDD. Life table analysis reveals no statistically significant differences in the probability of developing IDD between the ICA-positive schoolchildren and ICA-positive first degree relatives (P = 0.3). The estimated risk of developing IDD by 7 yr in the ICA-positive schoolchildren was 45% (95% confidence interval 15-74%) compared with 43% (confidence interval 22-63%) in the relatives. We conclude that ICA appear to be as predictive of IDD in low-risk schoolchildren as they are in high-risk relatives. These data suggest that it is feasible to predict IDD by screening a general population of schoolchildren for ICA and that those found to be positive could be considered, in addition to relatives, for intervention protocols to prevent the disease.