Table 10. .
Two-Locus Diplotypes for IL12B and IL23R
| Discovery Sample SetGlobala P=6.16×10-5 |
Replication 1 Sample SetGlobala P=9.72×10-4 |
Replication 2 Sample SetGlobala P=.005 |
Combined AnalysisGlobalb P=7.88×10-8 |
|||||||||||
| No. (%) of |
No. (%) of |
No. (%) of |
||||||||||||
| Two-Locus Diplotypec |
Cases | Controls | OR | Pd | Cases | Controls | OR | Pd | Cases | Controls | OR | Pd | ORcommon (95% CI)e | Pcombf |
| A-G/A-G–C-G/C-G | 175 (.375) | 110 (.239) | 1.91 | 9.62×10−6 | 174 (.350) | 121 (.243) | 1.67 | 2.98×10−4 | 168 (.349) | 116 (.274) | 1.43 | 1.50×10−2 | 1.66 (1.41–1.95) | 1.33×10−8 |
| A-G/A-G–C-G/X | 56 (.120) | 69 (.150) | .77 | .211 | 68 (.137) | 62 (.125) | 1.11 | .638 | 59 (.123) | 48 (.113) | 1.10 | .681 | … | … |
| A-G/A-G–X/X | 7 (.015) | 7 (.015) | .98 | 1 | 9 (.018) | 7 (.014) | 1.29 | .802 | 4 (.008) | 7 (.017) | .50 | .364 | … | … |
| A-G/X–C-G/C-G | 139 (.298) | 145 (.315) | .92 | .569 | 159 (.320) | 160 (.322) | .99 | 1 | 154 (.320) | 138 (.325) | .98 | .887 | .96 (.82–1.13) | .968 |
| A-G/X–C-G/X | 51 (.109) | 61 (.133) | .80 | .314 | 48 (.097) | 80 (.161) | .56 | .003 | 57 (.119) | 56 (.132) | .88 | .547 | .73 (.58–.91) | .019 |
| A-G/X–X/X | 1 (.002) | 9 (.020) | .11 | .011 | 5 (.010) | 10 (.020) | .49 | .298 | 3 (.006) | 12 (.028) | .22 | .016 | .27 (.11–.53) | .003 |
| X/X–C-G/C-G | 25 (.054) | 33 (.072) | .73 | .279 | 23 (.046) | 34 (.068) | .66 | .172 | 28 (.058) | 25 (.059) | .99 | 1 | .78 (.57–1.06) | .415 |
| X/X–C-G/X | 10 (.021) | 25 (.054) | .38 | .009 | 10 (.020) | 20 (.040) | .49 | .094 | 8 (.017) | 20 (.047) | .34 | .011 | .40 (.25–.61) | 7.42×10−4 |
| X/X–X/X | 3 (.006) | 1 (.002) | 2.97 | .624 | 1 (.002) | 3 (.006) | .33 | .624 | 0 | 2 (.005) | .00 | .219 | … | … |
Because of small counts in some of the cells, as well as the distribution of these counts, global P values were obtained by performing a permutation procedure on the data and generating a log-likelihood ratio homogeneity statistic for each permutation. The convergence of the statistic to its limiting distribution was measured by plotting the first two central moments as a function of the number of replicates. When the number of replicates reached a number in which the error of the P value estimates from a subsequent modeling procedure was negligible (<1%), a gamma distribution was fit using maximum-likelihood estimates for the parameters. Integration of the resulting gamma distribution yielded the global P values.
Calculated using Fisher’s combined test.
Allele 1 rs3212227-allele 1 rs6887695/allele 2 rs3212227-allele 2 rs6887695–allele 1 rs7530511-allele 1 rs11209026/allele 2 rs7530511-allele 2 rs11209026. For this analysis, the three nonrisk haplotypes for each gene were combined and were termed “X.”
Calculated using Fisher’s exact test.
Calculated using a Mantel-Haenszel common OR.
Calculated for diplotypes with the same effect (risk or protection) in all three sample sets, with use of Fisher’s combined test.