Insulin-dependent diabetes mellitus (IDDM) risk was evaluated in 765 siblings based on prospective observation of islet cell antibodies (ICAs) and insulin autoantibodies (IAAs) as a function of the degree of HLA identity to the proband and HLA-DR alleles. Twenty-eight (3.7%) siblings progressed to IDDM over a median observation period of 5.8 years. ICAs had higher sensitivity than IAAS (100% vs. 33% , P < 0.001), whereas persistent ICA positivity and double ICA/IAA positivity defined the highest actuarial risk (47% and 70%). Diabetes manifested after a mean of 3.2 years from the detection of ICAs in those siblings who were initially ICA negative and, importantly, the risk was equal to that of the siblings constantly positive from the first sample obtained. Although the combination of HLA identity and ICAs at or above 80 Juvenile Diabetes Foundation units carried the highest positive predictive value (77%), the high-risk HLA markers were insufficient to predispose siblings with low ICA levels to IDDM and low-risk HLA markers did not provide complete protection against high ICA levels and from subsequent IDDM. These results emphasize ICAs as the primary tool for risk evaluation in siblings followed by restricted HLA subtyping to reduce the population to be subjected to clinical intervention trials.