We evaluated the prevalence of glutamic acid decarboxylase autoantibody (GADA) in nonobese patients with type 2 diabetes mellitus in Korea and investigated the characteristics of GADA-positive and GADA-negative patients. Two years later, we assessed the progression of beta-cell function in these patients. Of the 647 nonobese patients with type 2 diabetes mellitus enrolled in the study, 10.1% was positive for GADA. Glutamic acid decarboxylase antibody-positive patients had lower fasting and stimulated C-peptide levels compared with GADA-negative patients (1.70 +/- 0.72 vs 1.24 +/- 0.59 microg/L, P < .001; 2.59 +/- 1.51 vs 1.99 +/- 0.82 microg/L, P < .001). Patients treated with insulin had lower fasting and stimulated C-peptide levels than those not treated (1.13 +/- 0.52 vs 1.66 +/- 0.73 microg/L, P = .002; 1.85 +/- 0.69 vs 2.49 +/- 0.91 microg/L, P = .004) and had higher titers of GADA (30.5 +/- 7.3 vs 6.0 +/- 4.8 U/mL, P < .001). In terms of progression of beta-cell function, fasting and stimulated C-peptide levels were significantly lower in GADA-positive patients after 2 years (from 1.24 +/- 0.59 to 0.95 +/- 0.54 microg/L, P = .004; from 1.99 +/- 0.82 to 1.61 +/- 0.77 microg/L, P = .007), whereas no such difference was observed in the GADA-negative patients. We demonstrate that a significant proportion of Korean patients may be positive for GADA; this is consistent with studies of white subjects, although disagrees with previous reports on Korean subjects. By assessing the presence of GADA in Korean type 2 diabetic patients, we are able to predict their course of beta-cell function and identify in advance those who are likely to require insulin treatment.