1.Study Population and Baseline Characteristics
There were in total 1981 subjects, including 1078 cases of ischemic stroke patients and 903 cases of control group. The clinical characteristics of the patients and controls were listed in Table 1. The proportion was higher of hypertension, diabetes, hyperlipidemia, and coronary heart disease, smoking in patient group, and also higher fibrinogen level, lower high-density lipoprotein cholesterol (HDL-C) level and lower blood platelet (PLT) level. There were significant differences of the proportion of men in our study between the case group and the control group. LDL-C level between two groups had no significant difference. These risk factors such as sex, hypertension, diabetes, smoking, alcohol consumption, triglyceride level, platelet level and fibrinogen level were associated with the three subtypes— LAA, CES and SVO groups of ischemic stroke, and the distribution of the three groups was also statistical significantly different.
2.Genotype and allele frequency distributions of SNPs in ischemic stroke and controls(Table 2)
At COX-1 (rs5788) andCOX-2 (rs20417), no significant differences were observed in genotype and allele frequencies between the ischemic stroke group including the three subtypes (LAA, CES, SVO) and the control group, even after we using multivariate analysis by adjusting for age, gender and related risk factors.
At COX-1(rs1330344), compared to the GG genotype, the AA genotype carriers had a lower susceptibility of ischemic stroke, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR = 0.657, 95% CI = 0.437-0.988, P = 0.044). And the AA genotype carriers also had a lower susceptibility of CES or SVO (OR = 0.506, 95% CI = 0.262-0.978, P = 0.043; OR = 0.616, 95% CI = 0.382-0.995, P = 0.048). Meanwhile, compared to the G allele, the A allele carriers had a lower susceptibility of ischemic stroke, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR = 0.812, 95% CI = 0.675-0.978, P = 0.029). And the A allele carriers also had a lower susceptibility of LAA or CES (OR = 0.778, 95% CI = 0.618-0.981, P = 0.034; OR = 0.702, 95% CI = 0.518-0.951, P = 0.023).
At COX-1(rs3842788), compared to the G allele carriers genotype (GG+GA), the AA genotype carriers had a higher susceptibility of ischemic stroke, this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR =5.203, 95% CI=1.519-5.159, P =0.016). And the AA genotype carriers also had a higher susceptibility of CES or SVO (OR = 9.821, 95% CI = 2.754-12.832, P = 0.041; OR = 4.603, 95% CI = 1.220-5.201, P = 0.025).
At COX-2(rs689466), compared to the G allele carriers genotype (GG+GA), the AA genotype carriers had a higher risk of LAA, this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR =1.404, 95% CI=1.019-1.934, P =0.038).
3. Genotype and allele frequency distributions of SNPs in hypertension/diabetes/smoking and their controls(Table 3)
We respectively divided all participants into high risk factor group (hypertension /diabetes/smoking groups) and control group.
As for hypertension group, at all the five SNPs, no significant differences were observed in genotype and allele frequencies between the hypertension group and control group. After we using multivariate analysis by adjusting for age, gender and related risk factors, the result remains negative.
As for diabetes group, (1) At COX-1(rs1330344), compared to the GG genotype, the GA genotype carriers had a lower susceptibility of diabetes, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors(OR = 0.711, 95% CI = 0.507-0.996, P = 0.047), and compared to the G allele carriers genotype (GG+GA), the AA genotype carriers had a higher susceptibility of diabetes, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors(OR = 1.319, 95% CI =1.036-1.680, P = 0.025).(2)At COX-2(rs20417), compared to the GG genotype, the GC genotype carriers had a lower susceptibility of diabetes, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors(OR = 0.616, 95% CI = 0.391-0.969, P = 0.036).(3) At COX-2(rs689466), compared to the GG genotype, the AA genotype carriers had a lower susceptibility of diabetes, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors(OR = 0.700, 95% CI = 0.502-0.977, P = 0.036).
As for smoking group, (1) At COX-1(rs5788), compared to the CC genotype, the A allele carriers genotype (CA+AA) carriers had a difference between smoking group and control group, this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR =1.525, 95% CI=1.031-2.255, P =0.034). (2) At COX-2(rs3842788), compared to the GG genotype, the GA genotype and the A allele carriers genotype (GG+GA) had a significant difference between smoking group and control group, and this difference remained significant in multivariate analysis by adjusting for age, gender and traditional risk factors (OR =0.631, 95% CI=0.416-0.957, P =0.030. OR =0.626, 95% CI=0.415-0.943, P =0.025).
4. MDR analysis
Gene-gene interactions were investigated for ischemic stroke and its subtypes by using MDR, and two models of SVO were statistical significant (Permutation test P value <0.05, Table 4). Among all gene-gene interaction models evaluated, a two-locus model (rs1330344 and rs689466) with the maximum cross-validation consistency (CVC: 10/10) and testing accuracy (0.5304) was regarded as the overall best model in our study. In addition, the best three-locus model (including rs1330344, rs3842788 and rs689466) had a maximum CVC (10) and higher testing accuracy (0.5125).
5. Additive effect analysis
We used multiple logistic regression analysis to analyze the correlation between the additive effect of COX-1 (rs1330344, rs3842788), COX-2 (rs689466) and ischemic stroke. Setting: the risk allele for each locus was 1, and the remainder was 0. We compared the carrying 1, 2, 3 risk loci to carry 0 risk loci without taking into account the specific loci or genes, just carrying any polymorphic loci count. Compared with those carrying none of the variants, individuals carrying one variant, carrying two variants and carrying three variants had no significant differences of the risk of ischemic stroke or its subtypes. (Table 5)