Prediction of IDDM in the General Population: Strategies Based on Combinations of Autoantibody Markers

Abstract

Strategies for assessing risk of progression to IDDM, based on single and combined autoantibody measurement, were evaluated in 2,855 schoolchildren (median age 11.4 years) and 256 children with newly diagnosed IDDM (median age 10.2 years), recruited to a population-based study in the Oxford region. In 256 children with IDDM, levels of antibodies ≥97.5th centile of the schoolchild population were found in 225 (88%) for islet cell antibodies (ICAs), in 190 (74%) for antibodies to GAD, in 193 (75%) for antibodies to protein tyro-sine phosphatase IA-2 (IA-2), and in 177 (69%) for autoantibodies to insulin (IAAs). Estimates of risk of progression to IDDM within 10 years, derived by comparing the distribution of antibody markers in the two populations (schoolchildren and children with IDDM), were 6.7% (ICAs), 6.6% (GAD antibodies), 5.6% (IA-2 antibodies), and 4.8% (IAAs) for schoolchildren with levels above the 97.5th centile, increasing to 20, 23, 24, and 11%, respectively, for antibody levels >99.5th centile. Most children with IDDM had multiple antibody markers, and 89% of those diagnosed over age 10 years had ≥2 antibodies above the 97.5th centile, as compared against 0.7% of schoolchildren, in whom this combination gave a 27% 10-year estimated risk of IDDM. Risk increased but sensitivity fell as combined antibody thresholds were raised, or the number of antibodies above the threshold was increased. Strategies based on detection of ≥2 antibodies with primary testing for GAD and IA-2 antibodies and second line testing for ICAs and/or IAAs were evaluated. Detection of at least two markers selected from GAD antibodies ≥97.5th centile and/or IA-2 antibodies ≥99.5th centile and/or ICAs ≥97.5th centile identified 0.25% of schoolchildren and 83% of children with newly diagnosed IDDM, with an estimated risk of 71% (95% CI 57–91). Although confirmation from prospective studies is still needed, this analysis suggests that antibody combinations can predict diabetes in the general population.