Autoantibodies to glutamate decarboxylase (GADA) identify individuals at increased risk of type 1 diabetes, but many people currently found GADA positive are unlikely to develop clinical disease. More specific GADA assays are therefore needed. Recent international workshops have shown that reactivity of sera from healthy donors varies according to assay type, and indicated that use of N-terminally truncated GAD65 radiolabels in GADA radiobinding assays is associated with higher specificity. To determine whether a radiobinding assay using radiolabeled GAD65(96-585) identified individuals at higher diabetes risk, samples from recent-onset patients and GADA positive first-degree relatives participating in the Bart’s-Oxford type 1 diabetes family study were re-assayed with full-length or N-terminally truncated GAD using the NIDDK harmonized protocol. The sensitivity in patients was the same with both labels, but fewer relatives re-tested positive with truncated GAD. Among relatives who progressed to diabetes, similar proportions were GADA positive when tested with either label, but because of their higher specificity the cumulative risk of diabetes was higher in those with autoantibodies to GAD65(96-585). Autoantibodies to GAD65(96-585) in relatives are more closely associated with diabetes risk than those to full-length GAD, suggesting assays using N-terminally truncated GAD should be used to select participants for intervention trials.