Combination of the eight selected SNPs in
IL1B by tagging SNPs in the HapMap CHB+JPT population may represent different
IL1B haplotypes. In addition, MDR analysis indicated that the best interaction model for predicting the development of GO is the eight-locus model composed by all eight SNPs analyzed in the present study (testing accuracy, 63.9%; cross-validation consistency, 100/100;
P < 0.0001). Therefore, we determined the haplotypes in the eight SNPs that had frequencies of >5% and identified the 13 haplotypes shown in
Table 4). Ht3-GCGCACTT, Ht5-ACACACTT, and Ht6-GTGCCCTC were found only in patients with GD, whereas Ht9-ACATCTTC, Ht10-ACATCTCC, Ht11-GCGCCCCT, Ht12-ACACCTTC, and HT13-GCGTCCCT were found only in the healthy individuals. Haplotype-specific analysis showed that the Ht1-ACACCTCC and Ht2-GCGCCCTT haplotypes may increase the risk of GD (
P = 1.241 × 10
−44 and 7.388 × 10
−8; OR, 21.599 and 3.917; 95% CI, 12.221–38.175 and 2.304–6.658, respectively) compared with the risk in healthy individuals. Ht4-GCGCCTCC may reduce the risk of GO among the patients with GD (
P = 0.025; OR, 0.502; 95% CI, 0.273–0.925).
Table 5 showed that 141 patients with GD bore the diplotype A-A/A-A, and it appeared more frequently in patients with GO than did A-A/G-G or G-G/G-G (
P = 0.008; OR, 1.650; 95% CI, 1.141–2.384). In 419 patients with GD, the non-Ht4/non-Ht4 diplotype was less frequently found in the patients who also had GO compared with at least one Ht4 haplotype (diplotypes Ht4/Ht4 and Ht4/non-Ht4,
P = 0.007; OR, 0.414; 95% CI, 0.214–0.797). These findings confirm that the results from genotype and haplotype analysis. In addition, the LD plots of
IL1B in the healthy individuals and GD patients with or without GO showed an apparent variation in these polymorphisms (
Fig. 2). These observations suggest that Ht1-ACACCTCC and Ht2-GCGCCCTT haplotypes put the individual at risk for the development of GD, whereas Ht4-GCGCCTCC may play a protective role against the development of GO.