Several publications have shown that certain alleles at the HLA-DRB1, -DQA1, and -DQB1 loci are associated with insulin-dependent diabetes mellitus (IDDM). Many of these studies have claimed that HLA-DQalpha1Arg52 and DQbeta1Asp57 showed the strongest association with IDDM, but these results could not be confirmed in different populations. We have recently found that DRbeta1Lys71+ provided major susceptibility to IDDM and that DQbeta1Asp57- had an additive effect to DRbeta1Lys71+ [Zamani et al., 1994a: Eur J Hum Genet 2:177-184]. This was confirmed with haplotype analysis in multiplex IDDM families [Zamani et al., 1996a: J Med Genet 33:899-905]. Therefore, we have reanalyzed the data from the literature on the association of the human leucocyte antigen (HLA) DRB1, DQB1, and DQA1 with IDDM in different ethnic groups to determine whether different amino acids in the antigen binding cleft of HLA class II molecules play a preponderant role in the development of IDDM. The results showed that the DRbeta1Lys71+ allele provided the highest relative risk for IDDM in the Belgian, Danish, Greek Taiwanese, and Chinese population while this was not the case in Norwegians, Sardinians, and Algerians. Indeed, in the Sardinian and Algerian population the DRB1*0401 allele encoding Lys71+ is very rare. Nevertheless, the few positive cases were always in the patient group. We also measured the clinical relevance of the testing for DRbeta1Lys71, DQbeta1Asp57, and DQalpha1Arg52 by calculating a prevalence-corrected positive predictive value (PcPPV), a prevalence corrected negative predictive value (PcNPV), the sensitivity and specificity of these tests. The results indicated that the sensitivity of the test for DRbeta1Lys71+ was lower than for DQalpha1Ag52+ and DQbeta1Asp57-, while testing for DRbeta1Lys71+ was more specific than testing for DQbeta1Asp57- and DQalpha1Arg52+ and that the DRbeta1Lys71+ allele had a higher PcPPV than DQalpha1Arg52+ and DQbeta1Asp57- in all studied populations. These results also showed that testing for DRbeta1LyS71+/+ can be useful in IDDM risk assessment particularly in populations with a high prevalence (P) of IDDM such as the Danish (P[IDDM] = 0.65%). PcPPV for DRbeta1Lys71+/+ was 0.2313 in the Danish, indicating a 23.13% risk for an individual who is homozygous for the genotype DRbeta1Lys71+/+ to develop IDDM. Some mechanisms which might explain the role of these HLA class II alleles in susceptibility to IDDM are discussed.