Prevention trials of type I diabetes are limited by recruitment of individuals at high risk of the disease. We investigated whether demographic and biological characteristics can identify rapid progressors among first-degree relatives of known patients at intermediate (< 10%) 5-year risk. Diabetes-associated antibodies, random proinsulin : C-peptide (PI/C) ratio and HLA DQ genotype were determined (repeatedly) in 258 islet antibody-positive IA-2Antibody-negative (Abpos/IA-2Aneg) normoglycaemic first-degree relatives. During follow-up (median 81 months), 14 of 258 Abpos/IA-2Aneg relatives developed type I diabetes; 13 (93%) of them had persistent antibodies conferring a 12% [95% confidence interval (CI): 5-19%] 5-year risk of diabetes. In Abpos/IA-2Aneg relatives with persistent antibodies (n = 126), the presence of >/= 1 HLA DQ susceptibility haplotype in the absence of a protective haplotype (P = 0.033) and appearance on follow-up of a high PI/C ratio (P = 0.007) or IA-2A-positivity (P = 0.009) were identified as independent predictors of diabetes. In persistently antibody-positive relatives with HLA DQ risk a recurrently high PI/C ratio or development of IA-2A identified a subgroup (n = 32) comprising 10 of 13 (77%) prediabetic relatives and conferred a 35% (95% CI: 18-53%) 5-year risk. Under age 15 years, 5-year progression (95% CI) was 57% (30-84%) and sensitivity 62%. In the absence of IA-2A, the combination of antibody persistence, HLA DQ risk and elevated PI/C ratio or later development of IA-2A and young age defines a subgroup of relatives with a high risk of type I diabetes (>/= 35% in 5 years). Together with initially IA-2A-positive relatives these individuals qualify for standardized beta cell function tests in view of prevention trials.