In 2009, Tregouet et al. identified the SLC22A3-LPAL2-LPA gene cluster as a risk cluster and haplotypes CTTG and CCTC formed by rs2048327, rs3127599, rs7767084 and rs10755578 as risk haplotypes for CAD in six White populations . From then on, several GWHS have focused on this hot spot. In a study consisted of 3657 patients with MI and 1211 control individuals, Koch et al. observed significant association between haplotypes formed by the same four SNPs in the SLC22A3-LPAL2-LPA region and MI (P = 0.0005), and found 3 risk haplotypes (CTTG, CCTC, and TTTC) . Later, Sawabe M etal analyzed rs2048327 (C/T) and rs10755578 (C/G) in 1,150 Japanese autopsy cases, and ascertained that haplotypes TC and TG worked as risk factors for both coronary sclerosis and CAD . In addition, Shaw et al. found that genetic variants at the SLC22A3-LPAL2-LPA locus were associated with decreased early-outgrowth colony-forming units, thereby increased the risk of MI , which may support the findings in population studies mentioned above. However, Qi et al. did not confirm the association of haplotypes at the SLC22A3-LPAL2-LPA locus with nonfatal MI risk in Hispanics .
For gene association studies, repeating previous findings across different populations is essential for exploring the full scape of their pathogenic nature. To date, there is no study focusing on the association between CAD and the SLC22A3-LPAL2-LPA gene cluster in Chinese people. Our study for the first time attempted to explore such association in Chinese Hans. We evaluated the association between four SNPs in this gene cluster and CAD by examming all kinds of associations (allelic, genotypic and haplotype). Nevertheless, we did not identify any significant evidence to link this gene cluster to CAD risk in this Chinese Han sample. The genotypic and allelic association between individual SNP and CAD drawn from our data were consistent with results from previous GWHS . Whereas, there are differences exist between our study and previous studies. The most common haplotype we found was TCTG instead of TCTC, which was reported in European populations [1, 10, 11]. Moreover, we did not confirm the association of haplotypes CTTG, CCTC, and TTTC with CAD reported in European populations [1, 10]. In addition, we did not find any association between CAD and two haplotypes TC and TG composed by rs2048327 and rs10755578, which was inconsistent with the results from a Japanese study .
There are many reasons for heterogeneity in genetic association studies. Ethnic differences in genetic structure may produce different LD, thereby differences in the significance of the association test, which also exist in other genetic association studies . Besides, differences in environmental, dietary or behavioral factors may also partially explain the heterogeneity in the genetic associations across ethnicities [15, 16]. Furthermore, different disease definitions under different criteria may also be partly responsible for the variation between studies.
To limit the potential influence of factors mentioned above, we carefully designed and implemented this study. First, we used an adequate sample with enough statistical power, to detect the genetic association, therefore, the discrepancies between our study and others in different populations may be more likely due to the ethnic differences in genetic structure. Second, we performed multivariate logistic model to adjust several possible covariates, such as age, gender, smoking status and BMI. Moreover, we identified the case subjects in a strict accordance with a generally accepted definition of CAD and excluded patients taking niacin which could decrease the plasma level of Lp(a) and/or patients with diabetes since diabetes status was reported to attenuate the relation between Lp(a) and cardiovascular risk .
Despite our study was well organized, several limitations still exist in this exploratory study. First, as a complex disease, many factors may contribute to CAD, such as environmental and polygenic backgrounds, dietary and behavioral factors, hence, the genetic parameter estimates (odd ratios, risk allelic or genotype frequencies) may be biased. In addition, uncontrolled confounding factors may lead to spurious associations. Although many important confounding factors were controlled or adjusted in our analysis, some potential confounders, such as lipid level, were unavailable for a large number of subjects and thus not controlled.