Objective The purpose of this study was to develop a high-throughput micro-plate radiobinding assay (RBA) of glutamic acid decarboxylase antibody (GAD-Ab) and to evaluate its clinical application. Methods 35labeled GAD65 antigen was incubated with sera for 24 h on a 96-well plate, and then transferred to the Millipore plate coated with protein A, which was washed with 4℃ PBS buffer, and then counted by a liquid scintillation counter. The GAD-Ab results were expressed by WHO standard unit (U/ml). A total of 224 healthy controls, 162 patients with type 1 diabetes mellitus(T1DM) and 210 patients with newly diagnosed type 2 diabetes (T2DM) were recruited. A total of 119 TI DM and healthy cases with gradually changing GAD-Ab levels were selected to compare the consistency of micro-plate RBA with conventional radioligand assay (RLA). Blood samples were obtained from the peripheral vein and finger tip in 32 healthy controls, 35 T1DM and 24 T2DM patients, and tested with micro-plate RBA and then compared with the conventional RLA to investigate the reliability of finger tip sampling. Linear correlation,student's t-test, variance analysis and receiver operating characteristic (ROC) curve were performed using SPSS 11.5. Results (1) The optimized conditions of micro-plate RBA included 2 μl serum incubated with3 ×104 counts/min 35S-GAD for 24 h under slow vibration, antigen-antibody compounds washed 10 times by 4℃ PBS buffer, and radioactivity counted with Optiphase Supermix scintillation liquid. (2)The intra-batch CV of the micro-plate RBA was 3.8%- 10.2%, and the inter-batch CV was 5.6%- 11.9%. The linearity analysis showed a good correlation when the GAD-Ab in serum samples ranged from 40.3 to 664 U/ml and the detection limit of measurement was 3.6 U/ml. The results from Diabetes Autoantibody Standardization Program (DASP) 2005 showed that the sensitivity and specificity for GAD-Ab were 78% (39 positive among 50 new-onset T1DM) and 98% (2 positive among 100 healthy controls). The results of GAD-Ab obtained with micro-plate RBA and RLA were closely correlated (r=0.915,P<0.001) with a high concordance level of 97.5% and a Kappa value of 0.95. (3)TI DM and T2DM patients showed higher positive rates for GAD-Ab than the healthy controls(46.9% and 5.2% vs 0.89% ,X2=123.5 and 10. 1 ,P <0.001 and <0.01, respectively). (4)The consistency of GAD-Ab measurement with RBA using finger tip blood and RLA measurement using venous blood was 96.7% (r =0.946,P <0.001, Kappa value: 0.905). Conclusions The micro-plate RBA of GAD-Ab has high sensitivity, specificity and reproducibility, and can be measured with finger tip blood sampling. It might be a better alternative for clinical practice.