The reporting of this systematic review complies with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement
Eligible studies were comparative studies of human subjects, provided genotyping was done at the 9p21-3 locus in a population with known coronary artery disease (previous/recent MI, or known epicardial coronary stenosis at enrollment). Applicable study designs included observational studies (case–control, cohort and cross sectional) where an association between the 9p21-3 allele and poor outcome or prognostic marker was reported. Only studies written in English were included due to feasibility.
We searched Ovid MEDLINE from 1948 until August 2012 and Ovid EMBASE, Web of Science and SCOPUS, from inception to August 2012. Subject headings (MeSH, EMTREE) were used: Chromosomes, Pair 9, Coronary artery disease, alleles and atherosclerosis. Keywords (9p21*) were used in Web of Science and Scopus. The detailed search strategy is attached in Additional file
A team of two trained reviewers independently screened all articles identified in the literature search. Discrepancies between the reviewers were resolved through discussions and consensus.
Markers of atherosclerotic severity included number of diseased vessels, Gensini Score and Duke CAD Prognostic Index (DPI). Markers of atherosclerotic severity and coronary disease progression are defined elsewhere
. We also assessed change in minimum lumen diameter (∆MLD) and number of new lesions at follow-up. Outcomes of interest included all-cause mortality, recurrent MI, need for coronary revascularization, triple vessel disease, Gensini score, DPI, ∆MLD, and number of new lesions. In studies where all-cause and cause-specific mortalities were separately tested, we analyzed all-cause mortality only.
Recurrent MI was defined any acute coronary syndrome associated with troponin elevation and/or ST segment elevation on electrocardiography (ECG). Need for coronary re-vascularization included surgical and percutaneous procedures performed either at target or non-target coronary vessels.
We extracted details on sample size, mean age, race, the identification (rs number) of the particular SNP genotyped, and outcomes of interest. SNPs previously reported in GWAS studies or in strong linkage disequilibrium with them were considered in the analysis.
In keeping with our goal to determine locus-outcome association we did not limit our analysis to a single SNP but instead tested for all available SNPs published in reports chosen for the meta-analysis. In studies reporting > 1 SNP-outcome association, we chose the SNP not elsewhere tested in other data sets. This allows us to capture all known markers in the locus and test as many markers as possible.
We used the Newcastle-Ottawa Quality Assessment to assess the risk of bias of the included studies
. The following items were used: selection of patients, comparability, assessment of exposure and/or outcome, length of follow-up, lost to follow-up. We were unable to assess potential publication bias due to limited number of studies included for each outcome
Genotypes were classified as either homozygous low risk (LR) heterozygous intermediate risk (IR) or homozygous high risk (HR). Study results were variedly reported using recessive [LR vs. (IR + HR)], dominant [(LR + IR) vs. HR)] and additive models [LR vs. IR vs. HR]. For the purpose of this manuscript we included additive models. For dichotomized outcomes, we extracted or calculated relative risk (RR) and its 95% confidence intervals (CI). We then pooled RR across the studies using the DerSimonian and Laird random effects methods with the heterogeneity from the Mantel–Haenszel method
. For continuous outcomes, we pooled weighted mean difference (WMD) using the same DerSimonian and Laird random effects methods.
We assessed the optimal information size (OIS), similar to power calculation in clinical trials, to evaluate the minimum sample size required in the literature to reach reliable conclusions
We assessed the consistency of the outcomes by testing heterogeneity using the I
statistic, where I
> 50% suggests a high level of heterogeneity
. All statistical analyses were conducted using STATA version 12 (StataCorp, College Station, TX).