Although the shortest (class I) minisatellite (i.e., variable number of tandem repeats [VNTR]) alleles in the 5' region of the insulin gene are positively associated with IDDM in Caucasians, the majority of Japanese are homozygous for class I alleles. Here, we determined the exact length, in number of repeat units (RUs), of class I alleles in Japanese subjects. The distribution of class I alleles in Japanese was trimodal, with peaks located at 32/33, 41, and 44 RUs. The shortest component (i.e., 1S [25-38 RUs]) alleles were significantly increased in the IDDM group compared with the control group (54 vs. 46%; P = 0.040). The 1S/1S genotype was significantly increased in the IDDM patients (34 vs. 20%; P = 0.005; relative risk 2.1). Furthermore, the transmission disequilibrium test of Japanese families with 1S/1M or 1S/1L heterozygous parents confirmed the association of 1S alleles; 17 alleles of 1S and 6 alleles of 1M (39-41 RUs) or 1L (42-44 RUs) were transmitted to affected offspring (P = 0.022). In addition, we found tight linkage of 1S with allele 9 of the tyrosine hydroxylase gene microsatellite and allele (-) of the IGF-II gene Apa I polymorphism, but neither 9 nor (-) alleles were significantly associated with IDDM. The present study suggests that a class I subset may have a role in IDDM susceptibility in Japan. It was revealed that the difference between 1S alleles and 1M or 1L alleles is almost consistently characterized by a sequence variation generated by deletion of two copies of an ACAGGGGTCC CGGGG repeat element, implying that sequence variation of class I alleles may influence disease susceptibility.