The Giessen-Bad Oeynhausen family study: improved prediction of type I diabetes in a low incidence population of relatives using combinations of islet autoantibodies in a dual step model

Abstract

To determine the value of a combined antibody screening for prediction of type I diabetes in a low incidence cohort, we prospectively studied 882 first-degree relatives (485 parents, 382 siblings and 15 offsprings) for up to 11 years who were not preselected for islet cell antibody (ICA) status. During the observation period, 16 individuals developed diabetes. The first serum sample obtained at study entry was analyzed for ICA and antibodies to insulin (IAA), glutamic acid decarboxylase (GADA) and anti-IA-2ic. A multivariate analysis, according to the Cox proportional hazard model considering the joint effects of all baseline variables, selected the four antibodies and the specific family history as significant risk confounding factors (p < 0.05). Further analysis by Kaplan-Meier Life-table methods confirmed a significantly increasing risk of diabetes with the number of autoantibodies present (p < 0.001). In accordance with the Cox model, relatives with more than one affected family member (a multiplex pedigree) and siblings and offsprings vs. parents were at increased risk of IDDM (p < 0.05). In addition to technical problems, a screening strategy based on initial ICA testing has the potential of missing ICA negative subjects among future cases of type I diabetes (19% were ICA negative in the present study) and we therefore set out to evaluate an alternative approach using a dual step strategy with a combination of GADA and anti-IA-2ic for initial screening followed by retesting of positive individuals for ICA and IAA. The combination of GADA and anti-IA-2ic for primary screening (step 1) proved to be more sensitive, identifying 94% of future cases of type I diabetes compared to 81% using ICA as initial test and this antibody combination identified 93% of those individuals with ICA of 20 JDF or more. Retesting of positive individuals for ICA and IAA (step 2) significantly improved the positive predictive value confering a risk of diabetes for siblings and offsprings with more than 2 antibodies within 5 years of 67% (95%CI: 39-90). We conclude that the prognosis of contracting IDDM in relatives is strongly related to the number of autoantibodies present, but the family history should be additionally considered for individual risk assessment. The proposed screening strategy could overcome the inherent problems of the ICA and IAA assays for large-scale screening. In the present study it allows 5-year risk estimates of up to 67% identifying 94% of future cases of type I diabetes.