Abstract To compare the clinical usefulness of commercial radioimmunoassay (RIA) kits based on recombinant and pig brain GAD, we measured glutamic acid decarboxylase autoantibody (GADAb) titers in 125 non-obese (body mass index<24) Japanese diabetics without insulin therapy using two commercial RIA kits based on recombinant human (rh) GAD65 (GADAb Cosmic) and purified pig brain native GAD (RIP Anti-GAD Hoechst). The frequencies of GADAb positivity using these two RIA kits (normal ranges; <1.3 and <4.0 U/ml, respectively) were about 4.8 (6/125) and 3.2% (4/125), respectively. The six patients found to be positive with RIA using GADAb Cosmic demonstrated significantly higher prevalence of NIDDM in their parents ( P=0.04), lower beta-cell function estimated by intravenous glucagon loading tests ( P=0.03) and higher prevalence of progression to insulin therapy ( P=0.0001). Five of these six patients slowly progressed to insulin-requiring status within 34±11 months of follow-up evaluation, and one of these five patients progressed to a completely insulin-dependent status within 30 months from the onset of diabetes. Of these six patients, two demonstrated chronic pancreatitis, three had chronic thyroiditis, and five showed HLA DR4. Interestingly, two of the six patients demonstrated very low GADAb titers (2.3 and 2.9 U/ml), while RIP Anti-GAD Hoechst showed no positivity with the same sera. Based on the binding study after pre-incubation of unlabeled GADs, these low titrated GADAb were elucidated to be true specific reactions to rh GAD65 alone. Moreover, one of the two patients with chronic thyroiditis and HLA DR4 slowly progressed to insulin-requiring status over a period of 45 months. These findings suggest that the measurement of GADAb using a commercial assay kit with rh GAD65 may be more useful to detect non-insulin-dependent type I diabetics among non-obese patients than using a commercial kit with purified pig brain native GAD, especially among those with low GADAb titers.