In 1996, we designed a randomized multicenter study to assess the effects of small doses of insulin on beta cell failure in slowly progressive type 1 diabetes (the Tokyo Study). We report here the preliminary results of this study. Glutamic acid decarboxylase 65 antibody (GADA)-positive patients were randomly divided into 2 groups: one group received insulin (Ins group), the other a sulfonylurea (SU group). Fifty-four patients (24 Ins group, 30 SU group) were analyzed at the end of a 4-year period. All patients underwent a 75 g oral-glucose test (O-GTT) every 6-12 months. The insulin-dependent stage was defined based on an integrated value of serum C-peptide levels on O-GTT ( summation operator CPR; sum of CPR at 0, 30, 60, 90, and 120 min) below 4.0 ng/mL. The summation operator CPR in the SU group decreased progressively from 22.0 +/- 10.6 to 11.3 +/- 7.5 ng/mL over the 48-month period (p < 0.001 vs. baseline). The summation operator CPR in the Ins group was unchanged. Among the SU group, 30% of subjects (9/30) progressed to IDDM, while 8.3% of Ins group subjects (2/24) progressed to IDDM (p = 0.087). With regard to the subjects who had a preserved C-peptide response ( summation operator CPR >/= 10 ng/mL), the proportion of SU group subjects who progressed to IDDM was significantly higher than that of the Ins group (7/28, 25% vs. 0/21, 0%, p = 0.015). Among subjects with a high GADA titer (>/=0 U/mL), 9/14 (64.3%) of the SU group, but only 2/16 (12.5%) of the Ins group, developed IDDM (p = 0.0068). As to those with a high GADA titer and a preserved C-peptide response, SU group subjects progressed to IDDM (7/12, 58.3%) more frequently than Ins group subjects (0/14, 0%) (p = 0.0012). In summary, our results suggest that small doses of insulin effectively prevent beta cell failure in slowly progressive type 1 diabetes. We recommend avoiding SU treatment and instead administering insulin to NIDDM patients with high GADA titer.