Objective Insulin autoantibody (IAA) is known to exist in sera of type 1 diabetes mellitus (T1DM) patients and pre-T1DM individuals. The aim of this study was to establish a novel microtiter plate radioimmunoassay (RIA) for IAA and evaluate its clinical value. Methods Diluted 125Ⅰ-insulin was mixed with 5 ul serum samples in a 96-well microtiter plate and then incubated for 72 h on an orbital plate shaker (4℃). The immunocomplexes were transferred to another protein a coated Millipore plate, and then the plate was washed with Tri-Buffered Saline Tween-20 (TBT) buffer. Counts per minute (CPM) was measured with liquid scintillation and luminescence counter. The positive cut-off point of IAA index was defined as ≥0.06 based on the 99-percentile of the distribution in 317 healthy individuals. The specificity and sensitivity of the assay were calculated from the samples provided by the fourth Diabetes Autoantibodies Standardization Program (DASP 2005). The IAA levels were determined in 71 T1 DM and 551 newly diagnosed type 2 diabetes (T2DM) patients, and 317 healthy controls. The t test, non-parametric test, x2 test and linear correlation analysis were performed on the data using SPSS 11.5 software. The concordance rate was estimated with Kappa value. Results (1) The optimized testing condition was described as 2×104 CPM of 125Ⅰ-insulin, 5 ul serum sample and slowly horizontal shaking for 72 h. (2) The intra-assay CV was 4.8%-8.9% and inter-assay CV was 6.4%-10.5%. Based on DASP 2005 samples, the specificity and sensitivity of the assay were 97% (97/100) and 50% (25/50), respectively. Ninety-six serum samples with different IAA levels were selected and tested to compare between our new method and a domestic IAA RIA kit. The results showed that the IAA indices from the two methods were positively correlated (r= 0.678, P<0.001). The concordance rate was 72.9 %(Kappa value=0.402). There were 25 samples with discordant results, which were positive for IAA titer using the corresponding microtiter plate RIA but negative using the novel RIA kit. (3) In TIDM group the positive rate of IAA was 19.7% (16/71), higher than the healthy controls (0.9%, x2=54.36, P<0.001). The subgroup of T1DM children (with 0-9 years) showed the highest IAA positive rate (55.6% ,x2=4.85, P<0.05). In T2DM group the frequency of IAA was 1.5% (8/551), which had no significant difference comparing with that of healthy controls (x2= 0.95, P >0.05). Conclusions Our proposed microtiter plate RIA method for IAA is highly sensitive and specific, likely to be feasible for clinical application. The frequency of IAA is high in children with T1DM.