Table 3. Two-way analyses of complement factor H 402 and rs11200638.

	PAR%		(95% CI)		M-H test: p-value
Model for rs11200638	CFH 402	rs11200638	CFH 402	rs11200638	LRT p-value	AIC value
Full	3.4 (0, 9.7)	58.3 (50.5, 64.1)	0.07	8.30E-08	0.03	221.8
Recessive	4.6 (0, 10.7)	44 (40.5, 54.0)	0.23	6.20E-09	0.12	221.5
Multiplicative	1.7 (0, 7.8)	79.8 (73.0, 88.1)	*	*	0.02	225.7
Dominant	2.2 (0, 13.7)	58.6 (43.9, 78.9)	0.91	5.80E-04	0.1	246.9

Four genotypic models for rs11200638 are considered: Let r0, r1, and r2 be the marginal relative risks for genotypes GG, GA, and AA. Then, recessive model implies r0=r1; multiplicative model implies r1=r0r2; dominant model implies r2=r1; full model does not have any restriction on relative risks except that r0, r1, r2>0. The 95% confidence intervals (CI) of population attributable risk (PAR) were obtained via a bootstrap re-sampling method with 999 replicates. Mantel-Hanzel (M-H) tests are conducted for one SNP association adjusted for the other SNP; likelihood ratio tests (LRT) for joint single nucleotide polymorphism (SNP) association under a two-way multiplicative model: the relative risk (or OR) for any genotype pair (A, B) relative to the baseline pair (A0, B0) is the product of relative risk (or OR) of A relative to A0 and that of B relative to B0. AIC denotes the Akaike's information criterion to access goodness-of-fit for the rs11200638 model.