Association between PPARG genetic polymorphisms and ischemic stroke risk in a northern Chinese Han population: a case-control study.

Two common polymorphisms of the peroxisome proliferator-activated receptor gamma (PPARG) gene, rs1801282 and rs3856806, may be important candidate gene loci affecting the susceptibility to ischemic stroke. This case-control study sought to identify the relationship between these two single-nucleotide polymorphisms and ischemic stroke risk in a northern Chinese Han population. A total of 910 ischemic stroke participants were recruited from the First Hospital of China Medical University, Shenyang, China as a case group, of whom 895 completed the study. The 883 healthy controls were recruited from the Health Check Center of the First Hospital of China Medical University, Shenyang, China. All participants or family members provided informed consent. The study protocol was approved by the Ethics Committee of the First Hospital of China Medical University, China on February 20, 2012 (approval No. 2012-38-1). The protocol was registered with the Chinese Clinical Trial Registry (registration number: ChiCTR-COC-17013559). Plasma genomic DNA was extracted from all participants and analyzed for rs1801282 and rs3856806 single nucleotide polymorphisms using a SNaPshot Multiplex sequencing assay. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using unconditional logistic regression to estimate the association between ischemic stroke and a particular genotype. Results demonstrated that the G allele frequency of the PPARG gene rs1801282 locus was significantly higher in the case group than in the control group (P < 0.001). Individuals carrying the G allele had a 1.844 fold increased risk of ischemic stroke (OR = 1.844, 95% CI: 1.286-2.645, P < 0.001). Individuals carrying the rs3856806 T allele had a 1.366 fold increased risk of ischemic stroke (OR = 1.366, 95% CI: 1.077-1.733, P = 0.010). The distribution frequencies of the PPARG gene haplotypes rs1801282-rs3856806 in the control and case groups were determined. The frequency of distribution in the G-T haplotype case group was significantly higher than that in the control group. The risk of ischemic stroke increased to 2.953 times in individuals carrying the G-T haplotype (OR = 2.953, 95% CI: 2.082-4.190, P < 0.001). The rs1801282 G allele and rs3856806 T allele had a multiplicative interaction (OR = 3.404, 95% CI: 1.631-7.102, P < 0.001) and additive interaction (RERI = 41.705, 95% CI: 14.586-68.824, AP = 0.860; 95% CI: 0.779-0.940; S = 8.170, 95% CI: 3.772-17.697) on ischemic stroke risk, showing a synergistic effect. Of all ischemic stroke cases, 86% were attributed to the interaction of the G allele of rs1801282 and the T allele of rs3856806. The effect of the PPARG rs1801282 G allele on ischemic stroke risk was enhanced in the presence of the rs3856806 T allele (OR = 8.001 vs. 1.844). The effect of the rs3856806 T allele on ischemic stroke risk was also enhanced in the presence of the rs1801282 G allele (OR = 2.546 vs. 1.366). Our results confirmed that the G allele of the PPARG gene rs1801282 locus and the T allele of the rs3856806 locus may be independent risk factors for ischemic stroke in the Han population of northern China, with a synergistic effect between the two alleles.

Chinese Library Classification No. R446; R741

Introduction

Worldwide, stroke is the main cause of disability and death (Powers et al., 2015; Salinas et al., 2016; Lee et al., 2017; Gao et al., 2018; Liska et al., 2018). In China, the incidence of stroke exhibits substantial geographic variation: Northern China has the highest incidence of stroke, whereas Southern China has a significantly lower incidence (Peng et al., 2014; Wang et al., 2017a; Sun et al., 2018). Ischemic stroke is a multifactorial and multigenic disease. Identifying genetic variants associated with ischemic stroke risk could shed light on the pathogenesis of stroke and lead to new approaches for the management and prevention of this complicated disease. Although several genome-wide association studies have assessed various polymorphisms that may contribute to ischemic stroke (Deng et al., 2010; Liu et al., 2011; Ranta and Lichtman, 2015; Wang et al., 2016b, 2017b), the genetic variants correlated with the predisposition to ischemic stroke have not been unequivocally determined in Chinese individuals.

Peroxisome proliferator-activated receptor gamma (PPARG) is a protective factor against the pathogenesis of ischemic stroke and is involved in inhibiting the pathogenesis and progression of hypertension, type 2 diabetes, hyperlipidemia, metabolic syndromes, obesity, and some inflammatory diseases (Ivanova et al., 2015; Liu and Wang, 2015; Liu et al., 2015, 2018; Yuan et al., 2015; Lee et al., 2017; Lv et al., 2017; Kim et al., 2018). A number of single-nucleotide polymorphisms (SNPs) in the PPARG gene have gained increasing attention, and influence the expression or activity of PPARG (Janani and Ranjitha Kumari, 2015). Among these SNPs, two common polymorphisms of the PPARG gene, rs3856806 (His477His, also known as C1431T or C161T) and rs1801282 (Pro12Ala), have been extensively researched in terms of their associations with atherosclerosis. There are reports of positive associations between increased risk and polymorphisms of the PPARγ gene; however, some reports have suggested a lower risk of atherosclerosis in carriers (Wu et al., 2013; Wang et al., 2015; Wei et al., 2016). Furthermore, single-locus independent main effects could explain only a part of the observed heritability – a phenomenon often referred to as the ‘‘missing heritability’’ problem (Maher, 2008). Therefore, studying combined SNP-SNP effects can improve the interpretation of genetic information (Prabhu and Pe’er, 2012). Therefore, SNP-SNP interactions and their departure from independent effects, which is identified by epistasis in genetics and epidemiological effect modification, could improve knowledge of the hereditary component of ischemic stroke.

Therefore, we aimed to assess the association between PPARG genetic polymorphisms at the rs1801282 and rs3856806 loci and ischemic stroke risk, and the interactions between the two loci in multiplicative and additive models.

Participants and Methods

Design

A case-control study.

Subjects

Our sample included 895 ischemic stroke patients and 883 age- and sex-matched healthy controls. Participant data were collected from December 2013 to December 2015, from the First Hospital of China Medical University, Shenyang, China. The criteria for the inclusion and exclusion of patients and controls are in the study flow chart ().

Study flow chart showing the inclusion and exclusion criteria of patients and controls.

All patients were between 40 and 80 years of age. If the patients are first diagnosed with an acute ischemic stroke based on neurological examination and radiography, the patient was determined to be eligible. None of the controls showed evidence of stroke or other neurological diseases. Patients with transient ischemic attack, cerebral embolism, cerebral trauma, cerebral vascular malformations, coagulopathy, autoimmune diseases, tumors, or chronic infectious diseases were excluded. Cases of kidney or liver disease, blood disease, occlusive arterial disease or venous thrombosis of the extremities were also excluded. All participants were of the Han nationality and lived in Liaoning province in northern China. Body mass index was classified according to the Asian obesity classification; body mass index > 22.9 kg/m2 was considered overweight (Jih et al., 2014). Hyperlipidemia was defined as a total plasma cholesterol level of 5.72 mM, a plasma triglyceride level of 1.7 mM (Wang et al., 2011), or if lipid-lowering drugs were currently in use. This study was approved by The Institutional Ethical Committee of the First Hospital of China Medical University on February 20, 2012 (approval No. 2012-38-1; ). This study was conducted according to the Code of Ethics of the World Medical Association (Declaration of Helsinki). All participants signed written informed consent. The protocol was registered with the Chinese Clinical Trial Registry (registration number: ChiCTR-COC-17013559).

SNP selection

Well-studied or functional SNPs were selected in view of previous studies that recorded associations between SNPs in PPARG and atherosclerotic diseases (Wu et al., 2013; Wang et al., 2015; Wei et al., 2016). SNP ID numbers and detailed sequence information for rs1801282 and rs3856806 were acquired from the public dbSNP database (http://www.ncbi.nlm.nih.gov/SNP/). The minor allele frequency of the above-mentioned two SNPs was greater than 0.05 in the Chinese Han population.

DNA extraction and genotyping

Genomic DNA was extracted and the SNaPshot reaction was performed for genotyping as previously described (Wang et al., 2016, 2017) using the SNaPshot Multiplex Kit (Applied Biosystems Co., Ltd., Foster City, CA, USA). shows the sequences of the PCR primers. Experimental results were analyzed using an ABI 3130XL DNA sequence detector and GeneMapper 4.0 (Applied Biosystems Co., Ltd.).

Statistical analysis

Unless otherwise noted, all statistical analyses were performed using SPSS 20.0 software (IBM Corporation, Armonk, NY, USA). All tests were two-tailed, and significance was defined as P < 0.05. Pearson’s chi-square test was used to compare the distribution of demographic variables and examine differences between risk factors and genotypes for alleles and haplotypes between patients and controls. A chi-square goodness-of-fit test was used for testing Hardy-Weinberg equilibrium for each genotype. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using unconditional logistic regression to estimate the association between ischemic stroke and a particular genotype. The Quanto power calculator was used to calculate power in our cohort. Assuming a genotypic relative risk for a recessive model of 2, a minor allele frequency of 0.15, a 1.88% population prevalence of ischemic stroke, and a Type I error probability of 0.05 in the sample size of 895 patient samples and 883 healthy controls, we would be unable to reject the null hypothesis that OR = 1 with a power of 99.99%. Based on the observed frequencies of the two SNPs, the SHEsis analysis platform was used to calculate linkage disequilibrium indices (D’ and r2) and infer haplotype frequencies (Shi and He, 2005; Li et al., 2009).

The attributable proportion due to interaction (AP) and relative excess risk due to interaction (RERI) were used to test additive SNP-SNP interactions. If RERI and AP equal 0, there is no biologic interaction (Leng et al., 2016). Gene interactions between the two SNPs were evaluated under two models of statistical epistasis, the additive and multiplicative models, using a customized Microsoft Excel spreadsheet as described by Andersson et al. (2005). Confounding factors, including sex, age, body mass index, diabetes mellitus, hypertension, history of alcohol use, history of smoking, history of ischemic stroke, and hyperlipidemia, were controlled in additive interaction analysis.

Results

Number of subject analysis

To determine the potential associations of PPARG SNPs with the risk of ischemic stroke, 910 ischemic stroke patients and 883 controls were recruited, but only 895 patients completed the study (): 9 patients were excluded for cardiogenic cerebral embolisms, 2 patients were excluded for cerebral arteritis, and 4 patients were excluded for unknown risk factors for ischemic stroke. All controls completed the study. This study collected final clinical data and peripheral blood specimens from 895 patients with ischemic stroke and 883 age- and sex-matched controls.

Clinical characteristics of study subjects

The summary of the general characteristics of the patient and control groups are shown in . There were no significant differences in sex (P = 0.091) or age (P = 0.179) between controls and patients. However, conventional risk factors for ischemic stroke, such as diabetes mellitus, body mass index, hyperlipidemia, hypertension, history of alcohol use, and history of smoking, were remarkably more prevalent in the patient group than in the control group.

Genotype analysis

All allele distributions were in Hardy-Weinberg equilibrium (rs1801282: P = 0.128; rs3856806: P = 0.959). The allele frequencies and genotypes of the two SNPs among the 895 patients with ischemic stroke and 883 control participants are shown in .

Compared with individuals with the homozygous wild-type CC genotype for rs1801282, individuals with the heterozygous CG genotype (14.4%) had an increased risk of ischemic stroke (adjusted OR = 1.748, 95% CI: 1.209–3.736, P = 0.003). In the dominant model, compared with the CC genotype of rs1801282, the CG + GG genotype was correlated with a significantly increased risk of ischemic stroke (adjusted OR = 1.844, 95% CI: 1.286–2.645, P < 0.001). The CT (36.0%) and TT (6.5%) rs3856806 genotypes were represented with increased frequency in the patients (adjusted OR = 1.290, 95% CI: 1.011–1.647, P = 0.041 and adjusted OR = 2.317, 95% CI: 1.247–4.214, P = 0.006, respectively). The CT + TT rs3856806 genotype was significantly associated with ischemic stroke (P = 0.010).

Haplotype analysis

The two SNPs were in linkage disequilibrium in our study population (i.e., D’ > 0.8 or r2 > 0.4). Of all possible haplotypes, only 3 had a frequency of > 0.03 among both patients and healthy controls and were included in the haplotype analysis. The haplotype analysis indicated that the frequency of the G-T haplotype of rs1801282-rs3856806 was significantly higher in patients than in healthy controls (OR = 2.953, 95% CI: 2.082–4.190, P < 0.001; ).

SNP-SNP interaction analysis

The dominant model was assessed in a further additive interaction analysis. The analysis below determined a significant interaction (RERI = 41.705, 95% CI: 14.586–68.824; AP = 0.860, 95% CI: 0.779, 0.940; S = 8.170, 95% CI: 3.772–17.697, P < 0.001) between rs1801282 and rs3856806 by controlling for age, sex, body mass index, diabetes mellitus, hypertension, history of alcohol use, history of smoking, history of ischemic stroke, and hyperlipidemia. Among all cases, those carrying the risk allele for both loci accounted for 86%.

The effect of the G risk allele at rs1801282 was modified by the T risk allele at rs3856806, as shown in , along with the results of each locus stratified by the other risk allele for ischemic stroke. The effect of the T risk allele at rs3856806 was stronger when rs1801282 had the G risk allele (adjusted OR = 8.001, 95% CI: 2.559–25.016, vs. adjusted OR = 1.188, 0.928–1.522), indicating synergistic epistasis (). A similar trend was detected for the effect of rs1801282 correlating with rs3856806 (adjusted OR = 2.546, 95% CI: 1.639–3.954, in the absence of rs3856806 vs. OR = 0.525, 95% CI: 0.245–1.123, when carrying both).

Discussion

This study evaluated the association between two genetic polymorphisms of PPARG with ischemic stroke risk in a northern Chinese Han population of 895 cases and 883 controls. To our knowledge, this is the first study to suggest that two PPARG genetic variants, rs1801282 and rs3856806, play a considerable role in the etiology of ischemic stroke. Our study also revealed a synergistic SNP-SNP interaction between the two risk alleles. Our findings suggest that the PPARG rs1801282 CG genotype and rs3856806 CT/TT genotypes are associated with an increased risk of ischemic stroke.

PPARG is involved in controlling immune responses and preventing atherosclerosis. Additionally, activation of PPARG effectively attenuates neurodegeneration and inflammatory processes in the brain (Culman et al., 2007; Wang et al., 2016a; Liu et al., 2017; Wu et al., 2018; Xia et al., 2018; Li et al., 2019). Furthermore, neuronal PPARG deficiency increased oxidative stress and brain damage during middle cerebral artery occlusion in mice (Zhao et al., 2009). In combination with the recent observation that activation of retinoid X receptors (RXR)/PPARG is the basis for neuroprotection by bexarotene in ischemic stroke (Certo et al., 2015), these results support the notion that PPARG plays an important protective role against ischemic injury.

The common polymorphism rs1801282 (Pro12Ala) in the PPARG gene is associated with greater insulin sensitivity (Stumvoll et al., 2001; Moffett et al., 2005; Youssef et al., 2014). Because the PPARG-rs1801282 polymorphism likely reduces transcriptional activity (Stumvoll et al., 2001), it may reduce insulin signaling and suppress the function of pancreas islet B cells (Stefan et al., 2001), ultimately resulting in diabetes mellitus (Altshuler et al., 2000; Ye et al., 2014; Kasim et al., 2016; Priya et al., 2016). In recent years, the role of PPARG-rs1801282 expression has been identified to decrease lipoprotein lipase activity (Xie et al., 2014; Zhang et al., 2016) and influence triglyceride removal, leading to atherosclerosis and dyslipidemia. In the present study, using participants with the PPARG-rs1801282 CC genotype as the reference group, the CG + GG genotype had a 1.844-fold higher risk of ischemic stroke after logistic regression. Potentially, the effect of rs1801282 on ischemic stroke risk is, to some degree, mediated by an effect on glucose and lipid metabolism. However, previous findings regarding the association between PPARG-rs1801282 polymorphisms and ischemic stroke risk are inconsistent (Lee et al., 2006; Volcik et al., 2008). Genetic heterogeneity due to differences in ethnicity could contribute to these discrepancies. Further reasons for the varying results may be sample size, study design, or patient selection criteria. The precise mechanisms underlying these observations need further elucidation.

We demonstrated a significantly increased risk of ischemic stroke with high-risk genotypes, such as the PPARG-rs3856806 T allele. However, this polymorphism in exon 6 of the PPARG gene had a protective effect on atherosclerotic lipid profiles in previous studies (Wan et al., 2010; Yilmaz-Aydogan et al., 2011). PPARG-rs3856806 and reduced coronary artery disease risk were associated in patients in an Australian Caucasian cohort (Wang et al., 1999) and in Chinese patients (Wu et al., 2013). These studies placed emphasis on cardiogenic stroke. Distinctive mechanisms may be crucial in the development of acute ischemic coronary and cerebrovascular events (Cheng et al., 2012). This phenomenon might be explained at least in part by differences in sample selection and study design, as well as sample size. Although the functional significance of rs3856806 (His477His) is limited because it is a silent mutation, evidence determining the role of this polymorphism in the expression of PPARG by changing the efficiency of protein translation is scarce. Moreover, this polymorphism possibly participates in modification of PPARG activation, resulting in altered transcription levels of target genes in metabolic pathways (Meirhaeghe et al., 2000, 2005; Wu et al., 2013).

SNP-SNP interaction or statistical epistasis, as defined by Fisher in 1918, is hypothesized to explain part of the hereditary nature of the disease (Marchini et al., 2005; Evans et al., 2006). Furthermore, finding significant statistical links (epistatic or otherwise) between SNPs could provide a strong indication of the different molecular level interactions between cases and controls. Interestingly, through SNP-SNP interaction studies, we have identified a combination of two SNPs that contribute to a strong risk of ischemic stroke. SNP-SNP interactions were statistically significant under both models, and the additive model provided a better fit than the multiplicative model with respect to biological synergy. This suggests that these two loci are in linkage disequilibrium. We confirmed that the G-T haplotype of rs1801282-rs3856806 is a potential susceptibility marker for ischemic stroke in our Chinese Han population in northern China.

Some limitations should be considered when interpreting these results. First, as the participants of our analysis were limited to a Han Chinese population, whether these results can be extrapolated to other populations is uncertain. Nevertheless, potential confounding factors were minimized by matching age, ethnicity, and sex. Second, the current study lacked functional assessments through in vivo and in vitro experiments to identity whether the genetic variants suggested in our study modulate ischemic stroke risk through their influences on the expression or functions of their host genes. The most important limitation of this study was the sample size and the absence of large-scale investigations. Thus, some of these results could be attributed to chance.

In conclusion, two polymorphisms in PPARG, rs1801282 C>G and rs3856806 C>T, increase the susceptibility to ischemic stroke in a northern Chinese Han population. We further identified two SNPs that may play a combined role in ischemic stroke. Further study of the association between these two SNPs and ischemic stroke is necessary in larger studies. Moreover, the functional mechanisms by which these two SNPs act should be clearly elucidated to better understand their role in the etiology of ischemic stroke. Given that the disability caused by ischemic stroke increases socioeconomic burden, excellent strategies for prevention and for the early detection of high-risk populations is essential to address this global problem. Not only must individuals at risk be directed into disease surveillance networks, but we must also increase the likelihood that reperfusion therapy can be administered within the therapeutic window to provide benefit when a stroke occurs.

Additional files:

Ethics document (Chinese)

Additional file 2: Open peer review reports

Table 1

Primer sequences

SNP	Primer sequence	Product size (bp)
rs1801282	Forward: 5′ CAG CCA ATT CAA GCC CAG	296
	TCC T 3′
	Reverse: 5′ CCG TAT CTG GAA GGA ACT
	TTA CCT TGT 3′
rs3856806	Forward: 5′ GCC AAG CTG CTC CAG AAA	161
	ATG A 3′
	Reverse: 5′ TGG CAG TGG CTC AGG ACT
	CTC T 3′

SNP: Single-nucleotide polymorphism.

Table 2

Characteristics and risk factors for stroke

Variable	Cases [n (%)]	Controls [n (%)]	P value
Age (≤ 60/> 60, years)	319 (35.6)/	288 (32.6)/	0.179
	576 (64.4)	595 (67.4)
Sex (male/female)	499 (55.8)/	457 (51.8)/	0.091
	396 (44.2)	426 (48.2)
BMI (≤ 22.9/> 22.9, kg/m2)	454 (50.7)/	274 (65.0)/	< 0.001
	441 (49.3)	148 (35.0)
Diabetes mellitus	234 (26.1)	64 (7.2)	< 0.001
Hypertension	552 (61.7)	178 (20.2)	< 0.001
History of smoking	310 (34.6)	137 (15.5)	< 0.001
History of alcohol use	154 (17.2)	98 (11.1)	< 0.001
History of ischemic stroke	63 (7.0)	46 (5.2)	0.108
Hyperlipidemia	324 (36.2)	177 (20.0)	< 0.001

Age, BMI, sex, hypertension, diabetes mellitus, history of alcohol use, history of smoking, history of ischemic stroke, and hyperlipidemia were assessed using Pearson’s chi-square test. There was no significant difference in age, sex and history of ischemic stroke between the patients and controls. BMI, hypertension, diabetes mellitus, history of alcohol use, history of smoking, and history of ischemic stroke and hyperlipidemia were associated with ischemic stroke. BMI: Body mass index.

Table 3

Allele and genotype frequencies of genetic polymorphisms among cases and controls and their main effects on stroke risk

SNP	Cases	Percent	Controls	Percent	OR (95% CI) *a	P value b
rs1801282
CC (reference)	756	84.5	807	91.4	1.00 (reference)
CG	129	14.4	75	8.5	1.748 (1.209–2.581)	0.003
GG	10	1.1	1	0.1	3.736 (0.430–32.436)	0.232
Dominant model CG + GG vs. CC					1.844 (1.286–2.645)	0.001
Recessive model CC + CG vs. GG					0.298 (0.034–2.601)	0.274
rs3856806
CC (reference)	515	57.5	587	66.5	1.00 (reference)	–
CT	322	36.0	274	31.0	1.290 (1.011–1.647)	0.041
TT	58	6.5	22	2.5	2.317 (1.274–4.214)	0.006
Dominant model CT + TT vs. CC					1.366 (1.077–1.733)	0.010
Recessive model CC + CT vs. TT					0.424(0.239–0.751)	0.003

ORs and 95% CIs were calculated based on multivariable logistic regression. Individuals with the heterozygous CG genotype for rs1801282 had a higher risk of ischemic stroke than individuals with the homozygous wild-type CC. The CG + GG genotype was associated with a significantly increased ischemic stroke risk. The CT and TT genotypes of rs3856806 were represented with increased frequency in the ischemic stroke patient group. The CT + TT genotype of rs3856806 was associated with ischemic stroke. *ORs and 95% CIs were calculated by logistic regression. a, bAdjusted OR (95% CI) and P value, adjusted for age, sex, body mass index, hypertension, diabetes mellitus, history of alcohol use, history of smoking, history of ischemic stroke, and hyperlipidemia. SNP: Single nucleotide polymorphism; OR: odds ratio; CI: confidence interval.

Table 4

Haplotype frequencies [n(%)] in cases and controls and their relationship to stroke risk

Haplotype	Cases	Controls	OR (95% CI)	P value
C-C	1329 (74.3)	1415 (80.1)	0.682 (0.580–0.803)	< 0.001
C-T	312 (17.4)	274 (15.5)	1.140 (0.954–1.362)	0.148
G-T	126 (7.1)	44 (2.5)	2.953 (2.082–4.190)	< 0.001

ORs and 95% CIs were calculated using the SHEsis program. The frequency of the G-T haplotype of rs1801282-rs3856806 was significantly higher in patients than in controls. OR: Odds ratio; CI: confidence interval.

Table 5

Stratified odds ratios for rs1801282 and rs3856806 with respect to the risk of ischemic stroke

Stratum	SNPs	OR (95%CI) *a	P value b
Ischemic stroke
Stratified by rs1801282
CG or GG	rs3856806 (CT or TT vs. CC)	8.001 (2.559–25.016)	< 0.001
CC	rs3856806 (CT or TT vs. CC)	1.188 (0.928–1.522)	0.171
Stratified by rs3856806
CT or TT	rs1801282 (CG or GG vs. CC)	2.546 (1.639–3.954)	< 0.001
CC	rs1801282 (CG or GG vs. CC)	0.525 (0.245–1.123)	0.097

ORs and 95% CIs were assessed based on multivariable logistic regression. The effect of the T risk allele at rs3856806 was stronger when rs1801282 had the G risk allele. A similar result was demonstrated for the effect of rs1801282 in relation to rs3856806. *ORs and 95% CIs were calculated by logistic regression. a, bAdjusted OR (95% CI) and P value, adjusted for age, sex, body mass index, hypertension, diabetes mellitus, history of alcohol use, history of smoking, history of ischemic stroke, and hyperlipidemia. SNP: Single nucleotide polymorphism; OR: odds ratio; CI: confidence interval.