OBJECTIVE - To determine the best autoantibody-based testing strategy for recruiting relatives for future intervention trials and to establish the role of islet cell antibodies (ICAs) within this strategy. RESEARCH DESIGN AND METHODS - ICAs, insulin autoantibodies (IAAs), GAD antibodies, and 1A-2 antibodies were determined in serum samples at study entry of 3,655 nondiabetic first-degree relatives of patients with type 1 diabetes who were follow ed for a median of 5.5 years. The cumulative risk of diabetes associated with single and combined antibody marker levels of greater than or equal to 97.5th percentile in schoolchildren was calculated by using life-table analysis. RESULTS - Of the 26 relatives who developed insulin-requiring diabetes during follow-up, 16 were aged <20 years and 7 were aged 20-39 years at study entry Of the 23 cases aged <40 years, 83% had 1A-2 and/or GAD antibodies, and 87% had 1AA and/or GAD antibodies greater than or equal to 97.5th percentile compared with 61% who had ICAs of greater than or equal to 5 Juvenile Diabetes foundation units JDF U. A two-step strategy with parallel testing for IA-2/GAD antibodies followed by WA testing identified 50% of cases aged <20 years and was associated with a 71% risk within 10 years. In subjects aged 20-39 years, this strategy conferred a 51% risk, whereas using ICAs as the second test gave 86% sensitivity and a 74% risk. Primary screening for 1;1A-2 and/or GAD antibodies followed by testing for IAA and/or ICA antibodies achieved the highest sensitivity in both age-groups and conferred a 63% risk. In contrast, ICAs of greater than or equal to 20 JDF U (the inclusion criteria for the European Nicotinamide Diabetes Intervention Trial) gave 48% sensitivity and 35% risk. CONCLUSIONS - ICA testing can be replaced as a primary screening measure by 1A-2/GAD or 1AA/GAD antibody testing. The sensitivity of ICAs (used alone or in combination with 1AAs) gives them a useful role in second-line testing. Combination testing could reduce the size of screening populations needed for recruitment in future intervention trials by similar to 50% compared with testing based on ICAs alone.