TABLE 5

Association of HLA class II with GADA, using rs660895 and rs2187668, to model HLA-DR4 and -DR3 (in 1,327 and 1,342 autoimmune diabetic case subjects, respectively)

SNPN (frequency)OR [95% CI]P
GADA positiveGADA negative
rs2187668
 A (DR3)*586 (0.26)76 (0.17)1.90 [1.42–2.55]6.03 × 10−6
rs660895
 G (DR4)*847 (0.38)168 (0.36)1.23 [0.97–1.57]0.090
 DR3569 (0.26)72 (0.16)1.43 [1.02–1.99]2.53 [1.82–3.51]
 DR4833 (0.39)166 (0.38)1.00 (ref.)1.78 [1.36–2.33]
 DRX754 (0.35)202 (0.46)0.56 [0.43–0.74]1.00 (ref.)5.31 × 10−9
 4/4147 (0.14)22 (0.10)2.57 [1.44–4.59]1.37 [0.72–2.59]
 3/X194 (0.18)23 (0.10)2.37 [1.38–4.09]1.26 [0.69–2.32]
 3/374 (0.07)6 (0.03)3.27 [1.25–8.61]1.74 [0.64–4.76]
 3/4227 (0.21)37 (0.17)1.88 [1.18–2.98]1.00 (ref.)
 4/X312 (0.29)85 (0.39)1.00 (ref.)0.53 [0.34–0.85]
 X/X124 (0.12)47 (0.21)0.51 [0.32–0.81]0.27 [0.16–0.47]
  • *Single SNP analyses use all samples genotyped at the test SNP.

  • †Uses the DR3 and DR4 SNPs to generate a non-DR3, non-DR4 allele, DRX, and to analyze the class II locus HLA-DRB1 as a single locus including alleles using two alleles in the model and using the third as reference. Only samples genotyped at both SNPs (rs2187668 and rs660895) were included in this analysis. The evidence that the DR3/4/X genotype coding was a more appropriate model than the combined DR3 and DR4 allelic effects model was unconvincing (P = 0.024). N, number of chromosomes or genotypes.