Design of the Coronary ARtery DIsease Genome-Wide Replication And Meta-Analysis (CARDIoGRAM) Study

Abstract

Background

Recent genome-wide association studies (GWAS) of myocardial infarction (MI) and other forms of coronary artery disease (CAD) have led to the discovery of at least 13 genetic loci. In addition to the effect size, power to detect associations is largely driven by sample size. Therefore, to maximize the chance of finding novel susceptibility loci for CAD and MI, the Coronary ARtery DIsease Genome-wide Replication And Meta-analysis (CARDIoGRAM) consortium was formed.

Methods and Results

CARDIoGRAM combines data from all published and several unpublished GWAS in individuals with European ancestry; includes >22 000 cases with CAD, MI, or both and >60 000 controls; and unifies samples from the Atherosclerotic Disease VAscular functioN and genetiC Epidemiology study, CADomics, Cohorts for Heart and Aging Research in Genomic Epidemiology, deCODE, the German Myocardial Infarction Family Studies I, II, and III, Ludwigshafen Risk and Cardiovascular Heath Study/AtheroRemo, MedStar, Myocardial Infarction Genetics Consortium, Ottawa Heart Genomics Study, PennCath, and the Wellcome Trust Case Control Consortium. Genotyping was carried out on Affymetrix or Illumina platforms followed by imputation of genotypes in most studies. On average, 2.2 million single nucleotide polymorphisms were generated per study. The results from each study are combined using meta-analysis. As proof of principle, we meta-analyzed risk variants at 9p21 and found that rs1333049 confers a 29% increase in risk for MI per copy (P=2×10⁻²⁰).

Conclusion

CARDIoGRAM is poised to contribute to our understanding of the role of common genetic variation on risk for CAD and MI.

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 2 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. With respect to CAD and MI, this effort was established by highly informative collections of patients with premature disease. Up to the present, these studies have individually identified at least 13 chromosomal loci with genome-wide significance for association with MI and other forms of CAD.^1–10

However, only a small proportion of the heritability of CAD has been explained. One major reason is likely to be the complex nature of atherosclerosis, with multiple genetic factors contributing only small effects to disease manifestation. In fact, for a typical genome-wide association study (GWAS) with ≈1000 cases and controls, the power to detect any effects at stringent significance levels is low.¹¹ To increase the power, we formed a consortium to pool data across all published and multiple unpublished GWAS for CAD and MI. Here, we aim to provide a detailed description of the structure and functioning of our consortium.

Methods

General Organization of the Consortium

Our consortium—the Coronary ARtery DIsease Genome-wide Replication And Meta-analysis (CARDIoGRAM) consortium—is based on a core of previously successful collaborations among single participating studies.^4–8 These case-control or prospective cohort studies both have detailed phenotyping for CAD, MI, or both as previously described.⁸ Control subjects have been derived from population-based studies in most investigations. For all of the participating studies, genome-wide scans were performed in the years 2006 to 2009. Statistical methods have been standardized across the studies, and an analysis platform has been created to allow summarized analyses on CAD, MI, and related phenotypes.

The organizational structure comprises a steering committee of either the principal or another key investigator of the participating studies and representatives from the statistical groups. The analysis team comprises 1 responsible genetic epidemiologist or statistical geneticist from each study. Standard operating procedures were generated to harmonize the data analyses. A centralized database for aggregated data is provided by the Cardiogenics consortium (http://www.cardiogenics.eu/web/) to allow centralized and decentralized access and analysis by all of the statistical groups (Figure 1). Transparency, with disclosure of any other collaborations with the potential to create conflict (including follow-up experiments), has been encouraged in order to sustain a high level of trust within the consortium. A consensus has been established by the steering committee to discourage intellectual property claims on aggregate findings before publication of results.

Figure 1.

Workflow of CARDIoGRAM. Analyses are performed in every study separately by the statistical group and then submitted to the central database. The analysis group checks the data quality within each study and queries individual studies on summary data. Once the initial quality control has been performed and data summaries are consistent across individual studies, these quality-controlled data are updated in the database and used for meta-analysis.

Statistical Analysis Methods

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.

To summarize the evidence across studies, we used aggregated data from association analyses for CAD and MI as well as for important subgroups as outlined a priori in the study protocol. The planned statistical analysis across all studies ending in the meta-analysis has 6 steps, as summarized in Table 1. In brief, after quality control, SNP-wise association tests are computed using log-additive genetic models with adjustment for age and sex in every study. 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 (inverse variance weighting) or random effects (DerSimonian-Laird) models are calculated, and outlying studies excluded.

Table 1.

Algorithm Applied for Analysis

Step	Description
1. Perform analyses in every study separately.	According to an a priori standard operating procedure, analyses are performed in every study separately. Here, quality-control indicators and tests for association with CAD are computed. Specifically, a log-additive model frequency test that takes into account the uncertainty of possibly imputed genotypes is performed with adjustment for age (at first CAD/MI onset for cases or recruitment for controls) and sex.
2. Upload to a central database.	The full set of summary data, including aggregated genotypes, quality-control parameters, and results from the association analyses are uploaded to a central database.
3. Perform quality control of data.	Quality control of the data is performed centrally according to previously agreed criteria, including a check of consistency of the given alleles across all studies, quality of the imputation, deviation from Hardy-Weinberg equilibrium in the controls, the MAF, and the call rate. In every study, the variance inflation factor λ is computed from genotyped SNPs and used for adjustment.
4. Apply meta-analysis procedure.	Separately for every SNP from every study that passes the quality criteria, the following procedure is applied for the meta-analysis: Fixed effects models are calculated together with Q- and I statistics for testing homogeneity. If there is no heterogeneity (P>0.01 for Q), fixed effects models are reported. If there is evidence for heterogeneity (P<0.01 for Q), an outlier test is performed comparing each study with the average of the others. If outliers are found with P<0.01/(number of studies providing data for the SNP), the most extreme study is excluded, and the procedure is repeated from step (1). If no outliers are found but heterogeneity was apparent, random effects models are calculated.
5. Compute the overall λ on the results of the meta-analysis.	Finally, in addition to the adjustment for each individual study inflation factor λ, the overall λ is computed on the results of the meta-analysis. Primary statistical evidence is based on individual study λ adjustment, but results with an additional adjustment for the overall λ based on the meta-analysis also are reported.
6. Conduct the main analyses in at least 2 sites.	To ensure high quality of the statistical analysis, the main analyses are conducted in at least 2 sites in parallel and independently. Importantly, all sites have access to all data sets to provide additional quality checks of the principal findings.

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, we selected 3 SNPs that were reported as lead SNPs in the first publications by McPherson et al² (rs2383206), Helgadottir et al³ (rs10757278), and Samani et al⁴ (rs1333049). In addition, we chose the SNP rs10811661 in the same chromosomal region that was associated with type 2 diabetes but not with CAD.^12,13

Given the potential etiologic heterogeneity of CAD, we predefined several subgroup analyses. Specifically, we chose to compare association results between female and male sex (female cases with female controls versus male cases with male controls) and younger onset and older onset cases (older cases with all controls versus younger cases with all controls). In addition, we analyzed age at first MI as a phenotype in cases only. We illustrate the results of these predefined subgroup analyses for the SNP rs1333049 with a known association to CAD and MI.

Replication Strategy and Levels of Evidence

In addition to providing a powerful meta-analysis, CARDIoGRAM has integrated a replication stage into the analysis plan, assembling a substantial resource of more than 60 000 samples, which are available for wet-laboratory or in silico replication. Replication has been predefined as being successful if the 1-sided P value remains <0.05 after correction for the testing of the number of genetic loci taken forward.

Because of the size of the primary CARDIoGRAM meta-analysis combined with the replication sample size, it is unlikely that a similar-scale experiment will be performed in the near future to independently verify the findings that emerge from our study. Mindful of this and to account for the caveats involved in combining data sets for meta-analysis, the CARDIoGRAM consortium has decided a priori to categorize the principal findings into different levels of evidence that depend on the strength of the association observed in the meta-analysis and replication samples. Establishing such criteria in advance avoids bias, helps with interpretation of findings, and may help to guide future work. The criteria we used are as follows:

• Level 1: Regions with SNPs that display (1) genome-wide significance with P<5×10⁻⁸ in a joint analysis of the GWAS and wet-laboratory replication stages and (2) display independently significant association in the replication stage at a threshold of 0.05 divided by the number of loci tested in replication.

• Level 2: SNPs that display genome-wide significance with P<5×10⁻⁸ in a joint analysis of the GWAS and wet-laboratory replication stages but do not display independent significant association in the replication stage.

• Level 3: SNPs with levels of association evidence with P>5×10⁻⁸ but <10⁻⁶ in a joint analysis of the GWAS and wet-laboratory replication stages.

These criteria will be applied to the main comparison and to all of the subgroup analyses.

Results

A description of the participants in each study is given in Table 2. Online-only Data Supplement Table 1 summarizes the genotyping platforms and imputation methods used for the individual studies.

Table 2.

Description of Probands (Cases/Controls) in Participating Studies

Study	Full Study Name	No.	MI, %	Female Sex, %	Age, y	BMI	CAD Definition	Control Definition	Ref
ADVANCE	Atherosclerotic Disease, VAscular functioN, and genetiC Epidemiology	278/312	50.4	57.9/59.0	45.8 (6.2)/45.3 (5.7)	31.2 (7.1)/26.5 (5.7)	Clinical nonfatal CAD (men aged ≥45 years, women aged ≤55 years), including AMI (enzymes), typical angina with ≥1 artery with >50% stenosis, positive noninvasive test, or PCI or CABG	No history of clinical CAD, CVA, or PAD	14
CADomics	Coronary Artery Disease and Omics	2078/2952	58.3	21.9/50.5	60.8 (10.1)/55.3 (10.8)	29.4 (5.1)/27.0 (4.7)	CAD: >50% stenosis in 1 major coronary artery and/or MI based on ECG and enzymes	Population sample with no history of MI	−
CHARGE	Cohorts for Heart and Aging Research in	2287/22 024	48.0	33.4/59.6	60.0 (7.9)/63.1 (8.0)	28.1 (7.4)/27.5 (8.0)	CHD: definite or probable MI, PTCA or CABG, or ECG MI	None of the conditions that define CAD	15
decode		6640/27 611	54.7	36.3/61.9	74.8 (11.8)/53.7 (21.5)	27.7 (4.7)/27.0 (5.4)*	MI: MONICA criteria (aged <75 years) or discharge diagnosis of MI; CAD: PCI or participation in CVD genetics program with self-report of CABG or PCI or discharge diagnosis of angina pectoris, MI, or chronic ischemic heart disease	Population sample	3
GERMIFS I	German Myocardial Infarction Family Studies I	884/1604	100.0	49.4/50.8	50.2 (7.8)/62.6 (10.0)†	27.4 (3.6)/27.7(4.5)	MI (<65 years) with >1 first-degree sibling with severe CAD (PTCA, MI, CABG)	Population sample	4
GERMIFS II	German Myocardial Infarction Family Studies II	1222/1287	100.0	33.1/48.3	51.4 (7.5)/51.2 (11.9)†	29.0 (3.8)/27.4 (4.6)	MI (aged <60 years); 59.4% with family history of CAD	Population sample	6
GERMIFS III (KORA)	German Myocardial Infarction Family Studies III (Cooperative Research in the Region of Augsburg)	1157/1748	100.0	20.1/48.9	58.6 (8.7)/55.9 (10.7)†	27.0 (3.6)/27.1 (4.5)	MI (aged <60 years); MONICA criteria	Population sample	−
LURIC/ AtheroRemo 1	Ludwigshafen Risk and Cardiovascular Heath Study	652/213	71.9	20.3/46.0	61.0 (11.8)/58.3 (12.1)	27.7 (4.4)/27.4 (4.2)	Symptoms of angina pectoris, NSTEMI, STEMI, or >50% coronary stenosis	No coronary lesions or minor stenoses (<20%)	16
LURIC/ AtheroRemo 2	Ludwigshafen Risk and Cardiovascular Heath Study	486/296	79.0	23.4/48.6	63.7 (9.4)/56.4 (12.7)	27.1 (3.8)/26.8 (4.0)	Symptoms of angina pectoris, NSTEMI, STEMI, or >50% coronary stenosis	No coronary lesions or minor stenoses (<20%)	16
MedStar		874/447‡	48.1	33.0/54.6	48.9 (6.4)/59.7 (8.9)	31.7 (6.8)/31.3 (7.9)	Angiography (≥1 coronary vessel with >50% stenosis); aged <65 years	Angiography normal, aged >45 years	5
MIGen	Myocardial Infarction Genetics Consortium	1274/1407	100.0	37.2/39.9	42.4 (6.6)/43.0 (7.8)†	27.6 (5.2)/25.8 (4.4)	MI (men aged <50 years/women aged <60 years)	Hospital based, community based, or nested case-control	5
OHGS	Ottawa Heart Genomics Study	1542/1455	61.6	24.1/48.0	48.7 (7.3)/75.0 (5.0)	28.5 (4.9)/26.0 (4.0)	Angiographic (>50% stenosis)	Asymptomatic	17
PennCATH		933/468	50.3	23.7/51.9	52.7 (7.6)/61.7 (9.6)	29.8 (5.6)/28.9 (6.4)	Angiography (≥1 coronary vessel with >50% stenosis); aged <65 years	Angiography normal, men aged >40 years/women aged >45 years	5, 18
WTCCC‖	Wellcome Trust Case Control Consortium	1926/2938	71.5	20.7/50.0	49.8 (7.7)‖	27.6 (4.2)‖	Validated MI, CABG, PTCA, or angina with positive noninvasive testing; aged <66 years	Unselected	1

Age and BMI data are presented as mean (SD). AMI indicates acute myocardial infarction; BMI, body mass index; CABG, coronary artery bypass graft; CHD, coronary heart disease; CVA, cerebrovascular accident; CVD, cardiovascular disease; MONICA, Multinational MONItoring of trends and determinants in CArdiovascular disease; NSTEMI, non–ST-segment elevated myocardial infarction; PTCA, percutaneous transluminal coronary angioplasty; STEMI, ST-segment elevation myocardial infarction; PAD, peripheral arterial disease; PCI, percutaneous coronary intervention.

^*Information on BMI in the deCODE study is available for 81.7% of the cases and 66.8% of the controls.

^†For cases, age at diagnosis; for controls, age at recruitment.

^‡Cases, angiographic CAD (>50% stenosis in at least 1 vessel); controls, angiography normal or <10% stenosis in all vessels.

^‖WTCCC controls comprise an equal number of subjects from the 1958 Birth Cohort and from the National Blood Service Donors. The latter were recruited in equal 10-year age bands from age 11 to 70 years. Additional phenotypes are not available for these controls.

Collectively, our consortium provides >10 times the number of cases and controls than the largest published individual CAD GWAS. Consequently, our meta-analysis will have increased power to detect small genetic effects. For instance, the power is >80% to detect an odds ratio (OR) of only 1.1 at the level of genome-wide significance, provided that the mean minor allele frequency (MAF) is at least 15% (see Figure 2). For the replication stage, the estimated power is shown in Figure 3.

Figure 2.

Estimated power of the meta-analysis. Power for different MAFs and ORs at a P value of 5×10⁻⁸ (n=22 000 cases and 60 000 controls).

Figure 3.

Estimated power for the replication. Power for different MAFs and ORs. Critical P value for a 1-sided test, 0.05×2=0.1 adjusted for the testing of 30 regions (n=15 000 cases and 15 000 controls).

The results for our proof-of-principle analysis are shown in Table 3 and Figure 4. In agreement with prior studies, the results show a strong association between the SNPs rs1333049, rs2383206, and rs10757278 and CAD but no evidence for association for rs10811661. The SNP rs1333049 was also part of the previous meta-analysis by Schunkert et al.¹⁹ In that report, the overall OR was 1.29 with a 95% CI of 1.22 to 1.37, virtually identical to our current result.

Table 3.

Proof-of-Principle Analysis on Chromosome 9p21

SNP	OR	95% CI	P*	Model
rs1333049	1.29	1.22, 1.36	2.06×10−20	RE
rs2383206	1.28	1.22, 1.35	1.64×10−20	RE
rs10757278	1.28	1.21, 1.35	5.79×10−19	RE
rs10811661	1.02	0.98, 1.05	0.3693	FE

The model is either fixed effects or random effects (RE).

^*Two-sided P value with adjustment for study-wise λ.

Figure 4.

Forest plots for SNPs on chromosome 9p21. Random effects (RE) models were calculated for SNPs rs1333049 (risk allele =C), rs2383206 (risk allele=G), and rs10757278 (risk allele=G); the fixed effects (FE) model was calculated for SNP rs10811661 (risk allele=C). Heterogeneity between the studies is indicated by I², and for every study, it is indicated whether the respective SNP was genotyped (G) or imputed (I) or a mixture of genotyped and imputed (NA).

Results for the subgroup analyses of rs1333049 are shown in Figure 5. Except for the analysis of older cases, results are more homogenous across study groups so that fixed effects models are selected. Overall, the association effect is strongest among younger cases. Moreover, there was an association of the number of the risk-conferring alleles with earlier age of first MI (β̂ = 0.37; SE=0.12; P=0.0015).

Figure 5.

Results for subgroup association analysis for SNP rs1333049. Shown are ORs with 95% CIs comparing female cases versus female controls, male cases versus male controls, old cases (≥50 years) versus all controls, and young cases (<50 years) versus all controls. Numbers below the x axis denote P values, fixed effects (FE) or random effects (RE) models, and number of cases/controls.

Discussion

The GWAS approach has proven useful in the identification of genetic variants affecting the risk of complex diseases. Specifically, several GWAS investigations have identified genes that reproducibly demonstrate association with CAD and MI.^5–6,8 Given that up to 2.5 million comparisons are carried out in parallel, a limitation of the approach is the clear discrimination between true and false associations. Consequently, stringent thresholds for genotyping quality and statistical significance need to be passed for reliable demonstration of a true-positive association. Large sample sizes are required to detect modest, but biologically important associations, and replication studies are required to minimize any remaining doubt about the reproducibility and relevance of such findings.

Almost all variants that have so far been associated with CAD or MI demonstrate risk ratios between 1.1 and 1.3 per risk allele. Given the small effect sizes, only 1 or few novel chromosomal loci were identified by each of the published studies. The newly formed CARDIoGRAM consortium will enhance the statistical power to detect true association by increasing the sample size by a factor of 10 for cases and 20 for controls. Indeed, all of the predefined subgroups are larger than the sample sizes of currently published GWAS. These larger samples are likely to substantially enhance the detection of true associations for CAD risk. Furthermore, we have prepared for a substantial replication phase and defined hierarchical levels of evidence a priori to help attach an appropriate level of confidence to our various findings as they emerge.

An unresolved problem in the interpretation of an individual GWAS is the potential for heterogeneity of risk allele effects across different individual populations. To meet this challenge, CARDIoGRAM has prespecified methods to uncover outliers and potential false-positive associations. Compared with disease states with intermediate or quantitative phenotypes, we anticipate generally higher levels of heterogeneity in the clinical definition of CAD and MI as a result of greater variation in clinical expression (eg, MI versus angiographic CAD) and in ascertainment (eg, cutoff for age at onset at young age in some clinical samples versus predominantly advanced age of onset in population-based samples). Relevant, but to a lesser degree, may be local or population-based differences in the genomic structure. We believe that the sample size of CARDIoGRAM will allow us to address this heterogeneity by performing stratified analyses that investigate clinically important questions, including age of onset of disease and sex.

The CARDIoGRAM consortium wishes to facilitate the in silico replication of findings from other investigators, with independent samples studying either candidate genes or genome-wide data. The consortium is limited in its ability to provide insights into the genetic risk of CAD in nonwhite populations. However, CARDIoGRAM will provide a substantial number of validated loci in white individuals that can be tested for association in nonwhite populations. Reciprocally, we will offer a large white race study sample for testing the relevance of novel loci discovered in GWAS of nonwhite populations. It will be of particular interest, therefore, to compare our findings for individuals of European ancestry with those observations made for other studies recruiting members of other racial/ethnic groups.

A secondary benefit that we foresee for the future use of the CARDIoGRAM data set and for further study of the results of planned meta-analyses will be to explore and strengthen evidence for the existence of a causal association of MI, CAD, or both with biomarkers and other intermediate traits. Many of these biomarkers and traits are known to have robust associations with CAD (eg, C-reactive protein, high-density lipoprotein cholesterol, hyperhomocysteinemia), and the list is continually expanding. Determining whether these associations are causal (ie, the biomarker or trait is involved in the pathogenesis of CAD) versus a consequence of reverse causality (ie, the biomarker is raised or decreased by the presence of CAD) versus a result of pleiotropic, but independent effects of another factor on both the marker and the CAD risk is an important clinical question with particular relevance for identifying therapeutic targets.²⁰ If variants causally affect the level of biomarker or trait, then one can immediately investigate whether those variants also affect CAD risk to a degree that is comparable to the quantitative association of the biomarker or trait with CAD risk.²⁰ Because the effect of the variant on the trait may be modest, the impact on CAD risk is likely to be small, and large sample sizes are required to demonstrate or refute an association of the variant with CAD risk. In this regard, we expect CARDIoGRAM to make a significant contribution.

In summary, we describe the design, structure, and plans of a large consortium formed to investigate the genetic basis of CAD and MI using genome-wide association data. Further, we discuss the likely benefits of the resource that will be created for the cardiovascular genetics community.

Appendix

CARDIoGRAM Consortium (affiliations listed in the online-only Data Supplement): Executive Committee: Heribert Schunkert, Nilesh J. Samani, Sekar Kathiresan, Muredach P. Reilly; Executive Secretary: Jeanette Erdmann; Steering Committee: Eric Boerwinkle, Jeanette Erdmann, Alistair Hall, Christian Hengstenberg, Sekar Kathiresan, Inke R. König, Reijo Laaksonen, Ruth McPherson, Themistocles L. Assimes, Muredach P. Reilly, Nilesh J. Samani, Heribert Schunkert, John R. Thompson, Unnur Thorsteinsdottir, Andreas Ziegler

Statisticians: Inke R. König (chair), John R. Thompson (chair), Devin Absher, Li Chen, L. Adrienne Cupples, Eran Halperin, Mingyao Li, Kiran Musunuru, Michael Preuss, Arne Schillert, Gudmar Thorleifsson, Benjamin F. Voight, George A. Wells

ADVANCE: Devin Absher, Themistocles L. Assimes, Stephen Fortmann, Alan Go, Mark Hlatky, Carlos Iribarren, Joshua Knowles, Richard Myers, Thomas Quertermous, Steven Sidney, Neil Risch, Hua Tang

CADomics: Stefan Blankenberg, Tanja Zeller, Arne Schillert, Philipp Wild, Andreas Ziegler, Renate Schnabel, Christoph Sinning, Karl Lackner, Laurence Tiret, Viviane Nicaud, Francois Cambien, Christoph Bickel, Hans J. Rupprecht, Claire Perret, Carole Proust, Thomas Münzel

CHARGE: Maja Barbalic, Joshua Bis, Eric Boerwinkle, Ida Yii-Der Chen, L. Adrienne Cupples, Abbas Dehghan, Serkalem Demissie-Banjaw, Aaron Folsom, Nicole Glazer, Vilmundur Gudnason, Tamara Harris, Susan Heckbert, Daniel Levy, Thomas Lumley, Kristin Marciante, Alanna Morrison, Christopher J. O’Donnell, Bruce M. Psaty, Kenneth Rice, Jerome I. Rotter, David S. Siscovick, Nicholas Smith, Albert Smith, Kent D. Taylor, Cornelia van Duijn, Kelly Volcik, Jaqueline Whitteman, Vasan Ramachandran, Albert Hofman, Andre Uitterlinden

deCODE: Solveig Gretarsdottir, Jeffrey R. Gulcher, Hilma Holm, Augustine Kong, Kari Stefansson, Gudmundur Thorgeirsson, Karl Andersen, Gudmar Thorleifsson, Unnur Thorsteinsdottir

GERMIFS I and II: Jeanette Erdmann, Marcus Fischer, Anika Grosshennig, Christian Hengstenberg, Inke R. König, Wolfgang Lieb, Patrick Linsel-Nitschke, Michael Preuss, Klaus Stark, Stefan Schreiber, H.-Erich Wichmann, Andreas Ziegler, Heribert Schunkert

GERMIFS III (KORA): Zouhair Aherrahrou, Petra Bruse, Angela Doering, Jeanette Erdmann, Christian Hengstenberg, Thomas Illig, Norman Klopp, Inke R. König, Patrick Linsel-Nitschke, Christina Loley, Anja Medack, Christina Meisinger, Thomas Meitinger, Janja Nahrstaedt, Annette Peters, Michael Preuss, Klaus Stark, Arnika K. Wagner, H.-Erich Wichmann, Christina Willenborg, Andreas Ziegler, Heribert Schunkert

LURIC/AtheroRemo: Bernhard O. Böhm, Harald Dobnig, Tanja B. Grammer, Eran Halperin, Michael M. Hoffmann, Marcus Kleber, Reijo Laaksonen, Winfried März, Andreas Meinitzer, Bernhard R. Winkelmann, Stefan Pilz, Wilfried Renner, Hubert Scharnagl, Tatjana Stojakovic, Andreas Tomaschitz, Karl Winkler

MIGen: Benjamin F. Voight, Kiran Musunuru, Candace Guiducci, Noel Burtt, Stacey B. Gabriel, David S. Siscovick, Christopher J. O’Donnell, Roberto Elosua, Leena Peltonen, Veikko Salomaa, Stephen M. Schwartz, Olle Melander, David Altshuler, Sekar Kathiresan

OHGS: Alexandre F. R. Stewart, Li Chen, Sonny Dandona, George A. Wells, Olga Jarinova, Ruth McPherson, Robert Roberts

PennCATH/MedStar: Muredach P. Reilly, Mingyao Li, Liming Qu, Robert Wilensky, William Matthai, Hakon H. Hakonarson, Joe Devaney, Mary Susan Burnett, Augusto D. Pichard, Kenneth M. Kent, Lowell Satler, Joseph M. Lindsay, Ron Waksman, Christopher W. Knouff, Dawn M. Waterworth, Max C. Walker, Vincent Mooser, Stephen E. Epstein, Daniel J. Rader

WTCCC: Nilesh J. Samani, John R. Thompson, Peter S. Braund, Christopher P. Nelson, Benjamin J. Wright, Anthony J. Balmforth, Stephen G. Ball, Alistair S. Hall (members of the WTCCC are listed in the online-only Data Supplement)

Disclosures

Dr Absher reports receiving an NIH research grant for the ADVANCE study. Dr Assimes reports receiving an NIH research grant for the ADVANCE study. Dr Blankenberg reports receiving research grants from NGFNplus for Atherogenomics and from BMBF for CADomics. Dr Boerwinkle received research support from NIH/National Human Genome Research Institute (NHGRI), GWA for gene-environment interaction effects influencing CGD; NIH/NHLBI, Molecular epidemiology of essential hypertension; NIH/NHLBI, Genome-wide association for loci influencing coronary heart disease; NIH/NHLBI, Genetics of hypertension-associated treatment; NIH/NHLBI, Modeling DNA diversity in reverse cholesterol transport; NIH/NHLBI, 20-year changes in fitness and cardiovascular disease risk; NIH/NHLBI, Genetic epidemiology of sodium-lithium countertransport; NIH/National Institute of General Medical Sciences (NIGMS), Pharmacogenomic evaluation of antihypertensive responses; NIH/NIGMS, Genomic approaches to common chronic disease; NIH/NHLBI, Genes of the CYP450-derived eicosanoids in subclinical atherosclerosis; NIH/NHGRI-University of North Carolina, Chapel Hill, Genetic epidemiology of causal variants across the life course; and NIH/NHLBI, Building on GWAS for NHLBI-diseases: the CHARGE consortium. Dr Cupples reports receiving research grants from NIH/NHLBI, The Framingham Heart Study; NIH/NHLBI, Genome-wide association study of cardiac structure and function; NIH/NHLBI, Functional evaluation of GWAS loci for cardiovascular intermediate phenotypes; and NIH/NHLBI, Building on GWAS for NHLBI-diseases: the CHARGE consortium. Dr Halperin reports receiving research grants from NIH, subcontract Genome-wide association study of Non Hodgkin’s lymphoma; ISF, Efficient design and analysis of disease association studies; EU, consultant AtheroRemo; NSF, Methods for sequencing based associations; BSF, Searching for causal genetic variants in breast cancer and honoraria from Scripps Institute, UCLA. Dr Halperin also reports ownership interest in Navigenics. Dr Hengstenberg reports receiving research grants for EU Cardiogenics. Dr Holm reports receiving a research grant from NIH; providing expert witness consultation for the district court of Reykjavik; serving as member of the editorial board for decodeme, a service provided by deCODE Genetics; and employment with deCODE Genetics. Dr Li reports receiving research grant R01HG004517 and other research support in the form of coinvestigator on several NIH-funded grants and receiving honoraria from National Cancer Institute Division of Cancer Epidemiology and Genetics. Dr McPherson reports receiving research grants from Heart & Stroke Funds Ontario, CIHR, and CFI. Dr Rader reports receiving research grant support from GlaxoSmithKline. Dr Roberts reports receiving research grants from the Cystic Fibrosis Foundation, NIH, and Cancer Immunology and Hematology Branch; membership on the speakers bureau for AstraZeneca; receiving honoraria from Several; and serving as consultant/advisory board member for Celera. Dr Stewart reports receiving research grant support from CIHR, Genome-wide scan to identify coronary artery disease genes, and CIHR, Genetic basis of salt-sensitive hypertension in humans; other research support from CFI: Infrastructure support; and honoraria from the Institute for Biomedical Sciences, Academia Sinica, Taipei, Taiwan. Dr Thorleifsson is an employee of deCODE Genetics. Dr Thorsteinsdottir reports receiving research grants from NIH and EU; serving as an expert witness for a US trial; having stock options at deCODE Genetics; and having employment with deCODE Genetics. Dr Kathiresan reports receiving research grants from Pfizer, Discovery of type 2 diabetes genes, and Alnylam, Function of new lipid genes, and serving as consultant/advisory board member for DAIICHI SANKYO Merck. Dr Reilly reports receiving research grant support from GlaxoSmithKline. Dr Schunkert reports receiving research grants from the EU, project Cardiogenics; NGFN, project Atherogenomics; and CADnet BMBF. M. Preuss, L. Chen, and Drs König, Thompson, Erdmann, Hall, Laaksonen, März, Musunuru, Nelson, Burnett, Epstein, O’Donnell, Quertermous, Schillert, Stefansson, Voight, Wells, Ziegler, and Samani have no conflicts to disclose.

Genotyping of PennCATH and MedStar was supported by Glaxo-SmithKline. Dawn M. Waterworth, Max C. Walker, and Vincent Mooser are employees of GlaxoSmithKline. PennCath/MedStar investigators acknowledge the support of Eliot Ohlstein, Dan Burns and Allen Roses at GlaxoSmithKline.