Prediction of type 1 diabetes is largely based on islet cell antibodies, but may be improved by combined analysis with other markers. We conducted a screening of first-degree relatives of type 1 diabetic patients in Germany using islet cell antibodies (ICA, indirect immunofluorescence on human pancreas) and insulin autoantibodies (IAA, radioimmunoassay) as screening markers. Of 1460 relatives tested, 2.3% (n = 33) were identified to be ICA+ (> or = 10 JDFu) and 1.9% (n = 27) to be IAA+ (> or = 50 nU/ml) in at least two subsequent serum samples. Of 44 antibody-positive relatives, 17 (39%) progressed to clinical insulin-dependent diabetes mellitus (IDDM) within 5 years. Life-table analysis showed a 58% risk of IDDM for ICA+ and 46% risk for IAA+ individuals. ICA combined with IAA gave a risk of 67% (p < 0.02 compared with ICA-, n.s. compared with IAA-). Of all relatives who progressed to clinical IDDM, only one was negative for ICA, but 6 were negative for IAA, resulting in a sensitivity of 94% for ICA and 65% for IAA. All antibody-positive relatives were characterized for HLA DR and DQ markers by genotyping. Relatives with 2 non DR3/ non-DR4 (DRx/x) alleles had no risk of IDDM, although they were consistently positive for one or more antibody specificities. We conclude that IAA screening is less sensitive than screening with ICA and relatives who lack ICA rarely progress to clinical disease. HLA analysis may be useful among antibody-positive relatives to define subgroups with a low risk of progression to exclude those from future intervention trials.