Most studies evaluating immune markers for prediction of Type 1 (insulin-dependent) diabetes mellitus have focused on first degree relatives, although only 10% of newly-diagnosed patients have an affected first degree relative. The Swedish Childhood Diabetes Register identifies 99% of all diabetic children at diagnosis. In this population-based study, islet cell antibodies and insulin autoantibodies in 0-14-year-old Swedish consecutively-diagnosed patients and control subjects were analysed to define their sensitivity and specificity. Over 16 months (1986-1987), 515 Swedish children developed diabetes. Plasma samples were obtained from 494 (96%) patients, and 420 matched control children. Among patients, the frequency of islet cell antibodies was 84% (415 of 494), insulin autoantibodies 43% (145 of 334); 40% (135 of 334) were positive for both and 88% (294 of 334) were positive for one or both. Among control children, 3% (14 of 420) had islet cell antibodies, 1% (4 of 390) insulin autoantibodies, and 4% (16 of 390) had either autoantibody marker. The predictive value of finding a patient with the disease was only 7% since 4% of the control children were antibody-positive and the cumulative incidence rate up to 15 years of age is 0.38%. None of the autoantibody-positive (n = 21) or negative control children developed diabetes during 3 to 5 years of follow-up. Longitudinal investigations of islet cell or insulin-autoantibody-positive healthy children are necessary to accurately determine the conversion rate from marker positivity to disease onset.