Introduction

Cardiovascular disease (CVD) is the leading cause of death and disability-adjusted life-years world wide, with increasing incidence and prevalence in low- and middle-income countries [1]. By 2020, more than 80 % of global CVD will be in these countries, with the largest burden occurring in the two largest countries, China and India, as they rapidly urbanise [1]. Nonmodifiable risk factors include increasing age, male sex, and heredity. Modifiable risk factors include smoking, hypertension, dyslipidemia, obesity, physical inactivity, and diabetes [2–4]. Biomarkers (e.g., C-reactive protein) have been combined with traditional risk factors to predict CVD events, and molecular markers hold further promise [5]. In 2007, genome-wide association studies on CVD identified a series of associated single-nucleotide polymorphisms (SNPs) in an intergenic region of chromosome 9p21, near the CDKN2A and CDKN2B genes [6, 7]. Recent success in identifying genes involved in complex diseases such as coronary artery disease (CAD) and myocardial infarction (MI) has been largely brought about by two major developments. First, modern array technology now enables simultaneous measurement of hundreds of thousands of single nucleotide polymorphisms (SNPs) across the human genome. Second, collaborations have been formed, bringing together large collections of well-phenotyped individuals.

However, only a small proportion of population of CAD has been explained in East Asia. In fact, for a typical association study with hundreds of cases and controls, the power to detect any effects at stringent significance levels is low. To increase the power, we formed a consortium to pool data across all published and multiple unpublished articles of East Asia population for CAD and MI. Here, we aim to provide a detailed description of the structure and functioning of our consortium.

Methods

Study populations

We systematically searched PubMed, Blackwell, Biosis Previews, Cochrane Library, China National Knowledge Infrastructure (CNKI), Chinese Biomedical (CBM) and Wan Fang (Chinese) databases for relevant articles published in different languages up to February 2012. We used search term “9p21 AND (cardiovascular disease OR coronary artery disease OR acute myocardial infarction OR myocardial infarction OR structural heart disease OR coronary atherosclerosis) AND (China OR Chinese OR Japanese OR Korean OR Mongol OR Asian population)” in this study. Other eligible studies were identified by a search of references cited in published articles.

The inclusion criteria of this meta-analysis were showed as follows: (a) including patients with a diagnosis of contained primary outcomes of coronary heart disease (CHD), myocardial infarction (MI), or coronary artery disease (CAD) and without other relate diseases; (b) describing the association of SNPs in 9p21 polymorphisms with CAD risk; (c) reported race and numbers of affected and unaffected participants, and race was East Asia population; (d) describing the genotyping method; (e) provided frequency genotype and the odds ratio (OR) with confidence intervals (CIs) or data sufficient to compute it. If more than 1 outcome was reported, the best-described phenotype was chosen. Several included articles reported consortium results with multiple independent populations. These populations were listed as separate data sets; all data were extracted independently by two reviewers according to the inclusion criteria. Discrepancies were documented and resolved by discussion with a third reviewer. The following information was extracted from all the eligible studies: patients with a diagnosis of disease not related with CAD, other race, SNPs not in the 9p21.

We studied subjects collected in eight different studies across China, Japan and Korea. All subjects were of East Asian origin. The large majority of cases had CAD as defined by China study (n = 3,413). The remaining cases had evidence of CAD based on either a revascularization procedure or anginal symptoms with a positive stress test. Details of each study are given in Table 1 and the online-only Data Supplement.

Table 1 Description of Probands (Cases/Controls) in Participating Studies
Full size table

Not all researchers use the same 9p21 SNPs, and most articles reported results for multiple SNPs (uniquely identified by their rs number). As a proof-of-principle analysis, a number of SNPs in the 9p21 region with known association to CAD and MI were analyzed upfront. Specifically, 3 SNPs were reported as lead SNPs in the first publications by McPherson et al. (rs2383206), Helgadottir et al. (rs10757278), and Samani et al. (rs1333049). In addition, the SNP rs10757274 and rs2383207 in the same chromosomal region were also associated with CAD. We report here in three common SNPs (rs1333049, rs10757278, and rs2383206) that were included in more than six articles [8–19]. These SNPs are in high linkage disequilibrium (D′ > 0.9; r 2 > 0.85) [19]. The remaining four articles used two additional SNPs, rs2383207 and rs10757278 [8–19]. These were also in high linkage disequilibrium. Information about age at diagnosis and race/ethnicity was also collected.

Genotyping

Different genotyping platforms have been used across the studies. An analysis restricted only to SNPs genotyped on all platforms would have been severely limited. For instance, the estimated overlap between the Affymetrix Genome Wide Human SNP Array 6.0 and the Illumina Human-1 mol/L chip is only about 250,000 SNPs. To allow for combined analyses across different platforms, missing SNPs were imputed by each study.

Haplotype analysis

Haplotype association analysis was performed with the Haploview software and SAS software (version 9.1.3, SAS Institute Inc) implementing the Stochastic-EM algorithm. Using this approach, we were able to simultaneously estimate haplotype frequencies for each haplotype pair consistent with the observed data. When ambiguous haplotypes were encountered, multiple, probability-weighted haplotypes were created.

Statistical analysis methods

For each data set, the observed genotype frequencies in controls were compared with expected frequencies based on Hardy–Weinberg equilibrium (χ 2 test with 2 degrees of freedom). All P values are 2-sided at the P = 0.05 level. After upload of the summary data and centralized quality control, a meta-analysis across all studies is performed for every SNP separately. Here, depending on the heterogeneity between studies, fixed or random effects models are calculated, and outlying studies excluded. Summary ORs and corresponding 95 % CIs were derived (by reanalysis when possible) and summarized using random-effects modeling weighted by each data set’s total variance (STATA) [20]. Subgroup differences were compared using the Q test for heterogeneity for each covariate separately. Publication bias was examined by performing a cumulative effects analysis [20, 21]. Wider ranges of these summary ORs indicate potential for publication bias.

Results

Literature search and meta-analysis databases

A total of 35 articles were identified in the initial search with the medical subject heading terms. After screening the abstracts or full contents, 23 articles were excluded (twelve were not related to CAD, five were not about polymorphism of SNPs in 9p21, four reported other races, two were republished studies). Finally, twelve studies (9,813 cases and 10,710 controls) were included in this meta-analysis. Eight studies of them were performed in China [8–13, 15, 18, 19], two in Japan [14, 16], and two in Korea (Table 1; Fig. 1). [15, 17] All genotype frequencies of studies were conformed to the HWE in control group.

Fig. 1
figure 1

Paper identification and exclusion

Full size image

Quantitative synthesis

The results of all the East Asia population for our proof-of-principle analysis are shown in Figs. 2, 3, 4 for rs1333049 (8 data sets), using a fixed-effects model, the summary OR was 1.29 (95 % CI, 1.23–1.36, P = 0.001). Heterogeneity was checked separately using Q statistic (Q = 10.14; I 2 = 31 %; P = 0.181). For rs2383206 (6 data sets), using a fixed-effects model, the summary OR was 1.24 (95 % CI, 1.18–1.31, P = 0.001). Heterogeneity was low (Q = 8.24; I 2 = 39.3 %; P = 0.144). For rs10757278 (6 data sets), using a random-effects model, the summary OR was 1.34 (95 % CI, 1.21–1.50, P = 0.001). Heterogeneity was checked separately using Q statistic (Q = 13.20; I 2 = 62.1 %; P = 0.022).

Fig. 2
figure 2

Forest plots for SNP rs1333049 (risk allele = C); the fixed effects (FE) model was calculated. Heterogeneity between the studies is indicated by I2

Full size image
Fig. 3
figure 3

Forest plots for SNP rs2383206 (risk allele = G); the fixed effects (FE) model was calculated. Heterogeneity between the studies is indicated by I2

Full size image
Fig. 4
figure 4

Forest plots for SNP rs10757278 (risk allele = G). Random effects (RE) models were calculated. Heterogeneity between the studies is indicated by I2

Full size image

In subgroup analyses, the G allele was also significantly associated with CAD in every race (Table 2). But there were little differences in ORs found between China population (5 data sets) and other races (Japan 2 data sets, Korea 1 data sets) (ORs of rs1333049, 1.27, 1.38, and 1.19).

Table 2 Main results of rs1333049 in the total and subgroup analysis
Full size table

Sensitivity analysis

Sensitivity analyses were performed by excluding the studies that were not in HWE, and the pooled OR were not materially altered in recessive genetic model (data not shown), indicating that the model results were statistically robust. Sensitivity analyses could not be done because of the small sample size in the subgroup analyses.

Publication bias

Begg rank correlation method and Egger weighted regression method were performed to assess the publication bias in recessive genetic model. No publication bias evidence was found (Table 2).

Discussion

Based on previous studies, evidence for the association between heart disease and chromosome 9p21 SNP markers had strong credibility. The 9p21 SNP markers have been identified though genome wide association studies and appear independent of traditional risk factors or family history [20, 22, 23]. The associated SNPs were rs1333049 (18 researches, OR = 1.23), rs10757274 (17 researches, OR = 1.24), rs2383207 (6 researches, OR = 1.28), rs2891168 (4 researches, OR = 1.29), rs10757278 (2 researches, OR = 1.27) [2]. In Asian race, rs1333049 (Japan, China), rs10757274 (China), rs10757278 (China) showed an association with the MI [2].

Most data sets are limited to white populations, usually of European descent, so the results for other racial groups might differ. We aim to investigate the association of this locus with CAD in 12 case–control studies of East Asians and undertook a meta-analysis for effect size, heterogeneity, publication bias, and strength of evidence. The SNP rs1333049 was also part of the previous meta-analysis by Schunkert et al. [24, 25]. In that report, the overall OR was 1.29 with a 95 % CI of 1.22–1.37, virtually identical to our current result. In our research, for rs1333049 (8 data sets), we received the similar result that the summary OR was 1.29 (95 % CI, 1.23–1.36, P = 0.001). The OR associated with rs2383206 was 1.24 (95 % CI, 1.18–1.31, P = 0.001). It was lower than rs2383206 analyzed in Europe population (OR = 1.28 (95 % CI, 1.22–1.35, P = 1.64 × 10−20), by Michael Preuss et al.). The OR associated with rs10757278 was 1.34 (95 % CI, 1.21–1.50, P = 0.001). It was higher than rs10757278 analyzed in Europe population (OR = 1.28 (95 % CI, 1.21–1.35, P = 1.64 × 10−20), by Michael Preuss et al.).

HapMap data suggests that rs10757274 and rs2383206 were located within 20 kb of each other on chromosome 9p21 and were in strong linkage disequilibrium (r 2 = 0.89) in North American population [5], and the five SNPs (rs2383206, rs2383207, rs10757274, rs10757278, and rs1333049) are in 1 block, and the D’ between each of the 5 SNPs was 1.0 in Chinese Han origin [13]. In our study, we identified one haploblock in the 9p21 region (rs10757274, rs2383206, rs10757278, and rs1333049) in Chinese population, and it’s similar to previous North American population study. On the basis of the stringent analysis of these SNPs and replication in our research, we are confident that 9p21 region is a strong candidate locus for CAD susceptibility in the Chinese population. Figure 5 shows the results of haplotype analysis for the SNPs examined. And then we calculated possible haplotype frequency of 8 loci in 9p21, using haploview and SAS software. Using the genotypes of 1,185 persons, we defined the haploblock structure of SNPs within the region of 9p21 in the Chinese population. By defining a solid spine of LD as D′ > 0.90, we identified one haploblock in the 9p21 region.

Fig. 5
figure 5

Pair-wise LD among eight SNPs in 9p21 in the China population. The numbers inside the squares are D′ × 100

Full size image

The region of chromosome 9p21 contained the coding sequences of gene for 2 cyclin-dependent kinase inhibitors, CDKN2A and CDKN2B, which played an important role in the regulation of the cell cycle and would be implicated, through their role in transforming growth factor (TGF)-β-induced growth inhibition, in the pathogenesis of atherosclerosis [13, 26–28]. Given the strong implication of this genomic region in CHD, the exact molecular mechanism of its involvement remains to be confirmed. McPherson et al. [7] resequenced the coding regions of the 2 genes most proximal to the risk locus and found no explanation of the CHD risk associated with this locus. However, the same region has recently been associated with increased susceptibility to type 2 diabetes [29, 30]. These results suggest that the region of 9p21 plays a role in multiple complex diseases. Further studies should focus on the identification of the underlying mechanism at this locus of CHD. In addition, the interactions between gender, smoking, and overweight with these SNPs need to be replicated in other populations.