To clarify whether glutamic acid decarboxylase65 antibodies (GAD65 Ab) and insulin autoantibodies (IAA) are good predictive markers for insulin-dependency in NIDDM, we studied GAD65 Ab and IAA in NIDDM patients treated with diet alone or in combination with oral hypoglycemic agents. GAD65 Ab were found in 12 or 29 (5.2%, P = 0.079 vs. control) NIDDM patients and IAA in 8 of 229 (3.5%). The frequency of GAD65 Ab and IAA positivity in NIDDM did not differ significantly from those of healthy controls (2/150, 1.3%, 2/150, 1.3%, respectively), but the frequency of patients who were positive for either GAD65 Ab or IAA, or both, was significantly higher than that of normal controls (17/229, 7.4% and 4/150, 2.7%, respectively, P < 0.05). In addition, the prevalences of GAD65 Ab and of IAA in those patients whose disease durations, since the diagnosis of diabetes, were less than one year were significantly higher than those of controls (4/30, 13.3%, P < 0.05, 4/30, 13.3%, P < 0.05, respectively). We found no differences between GAD65 Ab positive- and negative-patients in either BMI or serum C-peptide levels. Over a one to five year follow-up period (mean 2.0 yrs), serum C-peptide levels gradually decreased necessitating insulin treatment in three of the patients positive for GAD65 Ab and/or IAA (3/17, 17.6%; two were positive for both GAD65 Ab and IAA and one was positive for GAD65 Ab only). In contrast, only five patients negative for the two antibodies developed insulin requirement (5/212, 2.4%, P < 0.01). These results suggest that GAD65 Ab and IAA are good markers for predicting the development of insulin dependency in NIDDM patients and that the predictive value for insulin-dependency in NIDDM is enhanced by measuring both antibodies.