CHARGE (Cohorts for Heart and Aging Research in Genomic Epidemiology) Consortium Summary Results from Genomic Studies

GWAS have successfully identified genetic loci associated with a variety of conditions such as type 2 diabetes and coronary disease. The large number of statistical tests required in GWAS has posed a special challenge because few studies that have DNA and high-quality phenotype data are sufficiently large to provide adequate statistical power for detecting small to modest effect sizes. Even before the era of GWAS, the requirement for large sample sizes and the importance of replication have served as powerful incentives for collaboration. Meta-analyses combining summary data from multiple sources have improved the ability to detect new loci. The Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium was formed to facilitate GWAS meta-analyses and replication among multiple large and well-phenotyped cohort studies. The design of the CHARGE Consortium was formed initially from 5 prospective cohort studies from the United States and Europe: the Age, Gene/Environment Susceptibility (AGES) - Reykjavik Study, the Atherosclerosis Risk in Communities (ARIC) Study, the Cardiovascular Health Study (CHS), the Framingham Heart Study (FHS), and the Rotterdam Study (RS). Additional studies have expanded the CHARGE consortium based upon the phenotypes and willingness to share information across the research community. In order to facilitate investigators across the world to examine relationships between phenotypes and genetic markers within CHARGE published reports, an open site is made available on dbGaP that provides the rsID and the p-value for inspection. Access to detailed summary statistics (including minor allele frequency, odds ratio/effect size) requires approval of a Data Access Request (DAR).]]>
Association analysis between genotype and alpha-linoleic acid (ALA) was done separately within each study cohort according to a pre-specified plan. All analyses used robust standard errors. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.
]]>Association analysis between genotype and docosahexaenoic acid (DHA) was done separately within each study cohort according to a pre-specified plan. All analyses used robust standard errors. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.
]]>The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. Although originally founded by five cohort studies (Age, Gene, Environment, Susceptibility Study - Reykjavik (AGES), Atherosclerosis Risk in Communities (ARIC) study, Cardiovascular Health Study (CHS), Framingham Heart Study, Rotterdam Study, CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype.Association analysis between genotype and docosahexaenoic acid (DHA) was done separately within each study cohort according to a pre-specified plan. All analyses used robust standard errors. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.Association analysis between genotype and eicosapentaenoic acid (EPA) was done separately within each study cohort according to a pre-specified plan. All analyses used robust standard errors. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.
]]>Association analysis between genotype and white blood cell (WBC) count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and basophil count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and eosinophil count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and lymphocyte count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and monocyte count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and neutrophil count was conducted separately within each participating cohort according to a pre-specified plan. Meta-analyses were performed using inverse-variance weighted fixed-effects models to combine beta coefficients and standard errors from study level regression results for each SNP to derive a combined p-value. Study level results were corrected for genomic inflation factors (λGC) by incorporating study specific λGC estimates into the scaling of the standard errors (SE) of the regression coefficients. Meta-analyses were implemented using METAL.
]]>Association analysis between genotype and common carotid IMT was done separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.
]]> Association analysis between genotype and carotid plaque was done separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL
]]>The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of myocardial infarction (MI). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (ACE inhibitor) with respect to MI was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of myocardial infarction (MI). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (beta blocker) with respect to MI was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of myocardial infarction (MI). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (calcium channel blocker) with respect to MI was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of myocardial infarction (MI). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (thiazide diuretics) with respect to MI was conducted separately within each study cohort according to a pre-specified plan. For each The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of cardiovascular disease (CVD). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (ACE inhibitor) with respect to CVD was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of cardiovascular disease (CVD). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (beta blocker) with respect to CVD was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of cardiovascular disease (CVD). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (calcium channel blocker) with respect to CVD was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of drug-gene interactions on the outcome of cardiovascular disease (CVD). The discovery meta-analysis was conducted in 9 studies with GWAS data with European ancestry participants with treated hypertension. Association analysis between genotype and anti-hypertensive medication (thiazide diuretics) with respect to CVD was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of FEV1 as a mesasure of severity of obstructive lung disease. The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of FEV1 as a mesasure of severity of airflow obstruction in those participants who reported as "ever (current or former) smokers". The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on FEV1 as a mesasure of severity of airflow obstruction in those participants who reported as "never smokers". The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of FEV1/FVC as a mesasure of airflow obstruction independent of lung size. The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of FEV1/FVC as a mesasure of airflow obstruction independent of lung size in those participants who reported as "ever (current or former) smokers ". The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of FEV1/FVC as a mesasure of airflow obstruction independent of lung size in those participants who reported as "never smokers ". The discovery meta-analysis was conducted in 5 studies with GWAS data with European ancestry participants with pulmonary function (FEV1, FVC) measures. Association analysis between genotype and FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP and SNP-smoking interaction effects on the outcome of FEV1 as a measure of severity of obstructive lung disease. Smoking in this analysis was defined as a dichotomy (ever/never). The discovery meta-analysis was conducted in 19 studies with GWAS data with European ancestry participants with FEV1 and smoking data. Association analysis between genotype and genotype-smoking interaction with respect to FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP and SNP-smoking interaction effects on the outcome of FEV1 as a measure of severity of obstructive lung disease. Smoking in this analysis was defined as a continuous variable (pack-years). The discovery meta-analysis was conducted in 19 studies with GWAS data with European ancestry participants with FEV1 and smoking data. Association analysis between genotype and genotype-smoking interaction with respect to FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL. The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP and SNP-smoking interaction effects on the outcome of FEV1/FVC as a measure of obstructive lung disease independent of lung size. Smoking in this analysis was defined as a dichotomy (ever/never). The discovery meta-analysis was conducted in 19 studies with GWAS data with European ancestry participants with FEV1/FVC and smoking data. Association analysis between genotype and genotype-smoking interaction with respect to FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP and SNP-smoking interaction effects on the outcome of FEV1/FVC as a measure of obstructive lung disease independent of lung size. Smoking in this analysis was defined as a continuous variable (pack-years). The discovery meta-analysis was conducted in 19 studies with GWAS data with European ancestry participants with FEV1/FVC and smoking data. Association analysis between genotype and genotype-smoking interaction with respect to FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of forced expiratory volume at 1 second (FEV1) in cohorts of European ancestry. The discovery meta-analysis was conducted in 48,201 subjects from 23 studies with GWAS data. Association analysis between genotype and FEV1 was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of forced vital capacity (FVC) in cohorts of European ancestry. The discovery meta-analysis was conducted in 48,201 subjects from 23 studies with GWAS data. Association analysis between genotype and FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the ratio of forced expiratory volume at 1 second to forced vital capacity (FEV1/FVC) in cohorts of European ancestry. The discovery meta-analysis was conducted in 48,201 subjects from 23 studies with GWAS data. Association analysis between genotype and FEV1/FVC was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of incident CHD in the five prospective cohort studies of European ancestry. The discovery meta-analysis was conducted in 24,024 subjects with 2406 incident CHD cases with GWAS data. Association analysis between genotype and CHD was conducted separately within each study according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of incident MI in the five prospective cohort studies of European ancestry. The discovery meta-analysis was conducted in 24,024 subjects with 1570 incident MI cases with GWAS data. Association analysis between genotype and MI was conducted separately within each study according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of fibrinogen level in cohorts of European ancestry. The discovery meta-analysis was conducted in 120,246 subjects from 34 studies with GWAS data. Association analysis between genotype and log-transformed fibrinogen levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the continuous variable FEV1, taken at multiple time points within each of the 14 cohorts, with measurements in 27,249 adults of European ancestry. The GWAS meta-analysis was conducted in those with repeated measurements using a mixed effects model. Each cohort performed GWAS using a linear mixed effects model with a random intercept and random slope, with fixed effects for time, SNP, SNPxtime interaction, and baseline age, sex, standing height, smoking pattern during follow-up, smokingxtime interaction, baseline smoking pack-years, study site (as needed) and principal components of ancestry (as needed).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of airflow obstruction (a key pathophysiologic characteristic of chronic obstructive pulmonary disease, COPD) in 15 population-based cohorts with 32,875 subjects of European ancestry. The discovery meta-analysis was conducted in those with airflow obstruction defined as FEV1 and FEV1/FVC both less than the lower limit of normal, with unaffected subjects defined as FEV1, FVC, and FEV1/FVC all above the lower limit of normal. Logistic regression models were adjusted for current and former smoking, pack-years of smoking, age, sex, standing height, center/cohort as needed, and principal components for genetic ancestry (as needed). Genome-wide imputation and analysis were performed by cohort investigators, and genome-wide and regional meta-analyses were performed using METAL software. FIve discovery analyses were performed - all cohorts (including ever and never smokers), ever smokers, never smokers, asthma-free participants, and those with FEV1 < 65% predicted (more severe airflow obstruction).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of airflow obstruction (a key pathophysiologic characteristic of chronic obstructive pulmonary disease, COPD) in 15 population-based cohorts with 32,875 subjects of European ancestry. The discovery meta-analysis was conducted in those with airflow obstruction defined as FEV1 and FEV1/FVC both less than the lower limit of normal, with unaffected subjects defined as FEV1, FVC, and FEV1/FVC all above the lower limit of normal. Logistic regression models were adjusted for current and former smoking, pack-years of smoking, age, sex, standing height, center/cohort as needed, and principal components for genetic ancestry (as needed). Genome-wide imputation and analysis were performed by cohort investigators, and genome-wide and regional meta-analyses were performed using METAL software. FIve discovery analyses were performed - all cohorts (including ever and never smokers), ever smokers, never smokers, asthma-free participants, and those with FEV1 < 65% predicted (more severe airflow obstruction).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of airflow obstruction (a key pathophysiologic characteristic of chronic obstructive pulmonary disease, COPD) in 15 population-based cohorts with 32,875 subjects of European ancestry. The discovery meta-analysis was conducted in those with airflow obstruction defined as FEV1 and FEV1/FVC both less than the lower limit of normal, with unaffected subjects defined as FEV1, FVC, and FEV1/FVC all above the lower limit of normal. Logistic regression models were adjusted for current and former smoking, pack-years of smoking, age, sex, standing height, center/cohort as needed, and principal components for genetic ancestry (as needed). Genome-wide imputation and analysis were performed by cohort investigators, and genome-wide and regional meta-analyses were performed using METAL software. FIve discovery analyses were performed - all cohorts (including ever and never smokers), ever smokers, never smokers, asthma-free participants, and those with FEV1 < 65% predicted (more severe airflow obstruction).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of airflow obstruction (a key pathophysiologic characteristic of chronic obstructive pulmonary disease, COPD) in 15 population-based cohorts with 32,875 subjects of European ancestry. The discovery meta-analysis was conducted in those with airflow obstruction defined as FEV1 and FEV1/FVC both less than the lower limit of normal, with unaffected subjects defined as FEV1, FVC, and FEV1/FVC all above the lower limit of normal. Logistic regression models were adjusted for current and former smoking, pack-years of smoking, age, sex, standing height, center/cohort as needed, and principal components for genetic ancestry (as needed). Genome-wide imputation and analysis were performed by cohort investigators, and genome-wide and regional meta-analyses were performed using METAL software. FIve discovery analyses were performed - all cohorts (including ever and never smokers), ever smokers, never smokers, asthma-free participants, and those with FEV1 < 65% predicted (more severe airflow obstruction).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of airflow obstruction (a key pathophysiologic characteristic of chronic obstructive pulmonary disease, COPD) in 15 population-based cohorts with 32,875 subjects of European ancestry. The discovery meta-analysis was conducted in those with airflow obstruction defined as FEV1 and FEV1/FVC both less than the lower limit of normal, with unaffected subjects defined as FEV1, FVC, and FEV1/FVC all above the lower limit of normal. Logistic regression models were adjusted for current and former smoking, pack-years of smoking, age, sex, standing height, center/cohort as needed, and principal components for genetic ancestry (as needed). Genome-wide imputation and analysis were performed by cohort investigators, and genome-wide and regional meta-analyses were performed using METAL software. FIve discovery analyses were performed - all cohorts (including ever and never smokers), ever smokers, never smokers, asthma-free participants, and those with FEV1 < 65% predicted (more severe airflow obstruction).The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of African ancestry. The discovery meta-analysis was conducted in 19,896 subjects from 10 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of Europeanancestry. The discovery meta-analysis was conducted in 120,473 Euroean ancestry subjects from 15 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 4586 subjects from 2 studies with Hispanic ancestry and Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 146,562 subjects from 16 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of African ancestry. The discovery meta-analysis was conducted in 19,896 subjects from 10 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of Europeanancestry. The discovery meta-analysis was conducted in 120,473 Euroean ancestry subjects from 15 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 4586 subjects from 2 studies with Hispanic ancestry and Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 146,562 subjects from 16 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of African ancestry. The discovery meta-analysis was conducted in 19,896 subjects from 10 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of Europeanancestry. The discovery meta-analysis was conducted in 120,473 Euroean ancestry subjects from 15 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 4586 subjects from 2 studies with Hispanic ancestry and Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 146,562 subjects from 16 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of African ancestry. The discovery meta-analysis was conducted in 19,896 subjects from 10 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of Europeanancestry. The discovery meta-analysis was conducted in 120,473 Euroean ancestry subjects from 15 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 4586 subjects from 2 studies with Hispanic ancestry and Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 146,562 subjects from 16 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of African ancestry. The discovery meta-analysis was conducted in 19,896 subjects from 10 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of Europeanancestry. The discovery meta-analysis was conducted in 120,473 Euroean ancestry subjects from 15 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 4586 subjects from 2 studies with Hispanic ancestry and Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of blood pressure and hypertension in cohorts of European, African and Hispanic ancestry. The discovery meta-analysis was conducted in 146,562 subjects from 16 studies with Exomechip data. Association analysis between genotype and blood pressure levels and hypertensionn status was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis in seqMeta software. For gene burden tests, SKAT and standard tests as implemented in seqMeta package were used.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGFBP-3 levels in men from cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 8 sources in Stage 1 (n=7,170) and 2 studies from Stage 2 (n=815). Association analysis between genotype and IGFBP-3 level in men was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGF-1 levels in women from cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 14 studies in Stage 1 (n=11,874) and 3 studies from Stage 2 (n=2575); one stduy from Stage 3 (n=1514) provided de novo genotyping) of critical SNPs. Association analysis between genotype and IGF-1 levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGF-1 levels in cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 17 studies in Stage 1 (n=22,007) and 3 studies from Stage 2 (n=4575); one stduy from Stage 3 (n=3364) provided de novo genotyping) of critical SNPs. Association analysis between genotype and IGF-1 levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGFBP-3 levels in men from cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 8 sources in Stage 1 (n=7,170) and 2 studies from Stage 2 (n=815). Association analysis between genotype and IGFBP-3 level in men was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGFBP-3 levels in women from cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 10 sources in Stage 1 (n=9,862) and 2 sources from Stage 2 (n=1,148). Association analysis between genotype and IGFBP-3 level in women was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of IGFBP-3 levels in cohorts of European ancestry. The discovery meta-analysis was conducted in two Stages, with initial GWAS from 11 sources in Stage 1 (n=17,032) and 2 studies from Stage 2 (n=1,963). Association analysis between genotype and IGFBP-3 levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using a sample size-weighted z-score-based meta-analysis implemented in METAL. The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of PAI-1 level in cohorts of European ancestry. The discovery meta-analysis was conducted in 19,599 subjects from 8 studies with GWAS data. Association analysis between genotype and log-transformed PAI-1 levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted fixed-effects meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of tPA level in cohorts of European ancestry. The discovery meta-analysis was conducted in 26,929 subjects from 14 studies with GWAS data. Association analysis between genotype and log-transformed tPA levell was conducted separately within each study cohort according to a pre-specified plan. For each SNP, GWAS-specific results were combined using inverse-variance weighted meta-analysis in METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between QT interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of African ancestry. The discovery meta-analysis was conducted in 10,235 subjects from 5 studies with medication, ECG and GWAS data. TCA-SNP interaction on QT interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between QT interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of European Caucasian ancestry. The discovery meta-analysis was conducted in 45,706 subjects from 14 studies with medication, ECG and GWAS data. TCA-SNP interaction on QT interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between QT interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of Hispanic/Latino ancestry. The discovery meta-analysis was conducted in 13,808 subjects from 2 studies with medication, ECG and GWAS data. TCA-SNP interaction on QT interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between RR interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of African ancestry. The discovery meta-analysis was conducted in 10,235 subjects from 5 studies with medication, ECG and GWAS data. TCA-SNP interaction on RR interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between RR interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of European Caucasian ancestry. The discovery meta-analysis was conducted in 45,706 subjects from 14 studies with medication, ECG and GWAS data. TCA-SNP interaction on RR interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between RR interval (from ECG) and tri/tetracyclic anti-depressant (TCA) use in cohorts of Hispanic/Latino ancestry. The discovery meta-analysis was conducted in 13,808 subjects from 2 studies with medication, ECG and GWAS data. TCA-SNP interaction on RR interval depended on the study design (family or unrelated) and extent of ECG and medication data (longitudinal). Cohorts with longitudinal ECG and medication data used GEE with independent working correlation structure. Family-based studies used LMM that accounted for relatedness, sampling design and heterogeneity of outcome variance by drug use. Cohorts with unrelated subjects and cross-sectional data used linear regression models with robust standard errors. An inverse-variance-weighted meta-analysis was performed with genomic control using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the association between coding-region SNPs (from the Exomechip) with Late-onset Alzheimer's disease (LOAD) in cohorts of European Caucasian ancestry. The discovery meta-analysis was conducted in 1,393 cases and 8,141 cognitively normal controls from 4 studies . The study designs were all population-based cohorts. Analyses used score tests for single variant tests. Each study adjusted for age, sex and those principal components associated with LOAD. Single variant and SKAT tests were performed in R with the seqMeta package in order to meta-analyze study-specific scores and variances/covariances.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the outcome of C-reactive protein (CRP) level in cohorts of European ancestry. The discovery meta-analysis was conducted in 66,185 subjects from 15 studies with European ancestry with SNP and CRP data. Association analysis between genotype and CRP levels was conducted separately within each study cohort according to a pre-specified plan. For each SNP, study-specific results were combined using inverse-variance weighted meta-analysis using METAL software.-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>-6 or cohort-specific criteria, minor allele frequency (MAF) was <1%; imputation information score was <0.5; results were only available from two or fewer cohorts; or total n was <10,000. LD score regression was conducted on the GWAS summary results to examine the degree of inflation in test statistics, and genomic control correction was considered unnecessary. Meta-analyses were performed by METAL or R (v3.2.2) with inverse variance-weighted random-effects model used due to heterogeneity in levels of alcohol consumption and population demographics.
]]>The data are in the public ftp site.
]]>The data are in the public ftp site.
]]>The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QRS interval from ECG with sulfonylurea use in cohorts of African ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 11,731 subjects from 6 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QRS interval from ECG with sulfonylurea use in cohorts of European ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 45,002 subjects from 12 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QRS interval from ECG with sulfonylurea use in cohorts of Hispanic ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 15,124 subjects from 3 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QT interval from ECG with sulfonylurea use in cohorts of African ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 11,731 subjects from 6 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QT interval from ECG with sulfonylurea use in cohorts of African ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 11,731 subjects from 6 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of QT interval from ECG with sulfonylurea use in cohorts of Hispanic ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 15,124 subjects from 3 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of JT interval from ECG with sulfonylurea use in cohorts of African ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 11,731 subjects from 6 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of JT interval from ECG with sulfonylurea use in cohorts of European ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 45,002 subjects from 12 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of SNP effects on the interaction of JT interval from ECG with sulfonylurea use in cohorts of Hispanic ancestry. Drug use was based upon medication inventories conducted at study visits or use of a pharmacy database. The genotyping was conducting at several sites, with QC and analyses conducted at sites based upon a standardized format and design. Single variant meta-analysis was conducted in 15,124 subjects from 3 cohorts with GWAS array data. Association analysis was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of HDL cholesterol level in cohorts of African ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 14,330 subjects from 8 studies with Exomechip data. Association analysis between genotype and HDL level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMeta.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of HDL cholesterol level in cohorts of European ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 42,208 subjects from 12 studies with Exomechip data. Association analysis between genotype and HDL level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMeta.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of LDL cholesterol level in cohorts of African ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 14,330 subjects from 8 studies with Exomechip data. Association analysis between genotype and LDL level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMeta.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of LDL cholesterol level in cohorts of European ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 42,208 subjects from 12 studies with Exomechip data. Association analysis between genotype and LDL level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMeta.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of TG level in cohorts of African ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 14,330 subjects from 8 studies with Exomechip data. Association analysis between genotype and TG level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMeta.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare coding region SNP effects on the outcome of TG level in cohorts of European ancestry. The genotyping was conducting at several sites, with QC at one site. The single variant meta-analysis was conducted in 42,208 subjects from 12 studies with Exomechip data. Association analysis between genotype and TG level was conducted separately within each study cohort according to a pre-specified plan. For each SNP, cohort-specific results were combined using seqMetaThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of African ancestry. The discovery meta-analysis was conducted in 9,919 non-diabetic subjects from 8 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of African ancestry. The discovery meta-analysis was conducted in 9,919 non-diabetic subjects from 8 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of European ancestry. The discovery meta-analysis was conducted in 54,907 non-diabetic subjects from 22 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF> 0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of European ancestry. The discovery meta-analysis was conducted in 54,907 non-diabetic subjects from 22 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of European and African ancestry. The discovery meta-analysis was conducted in 60,564 non-diabetic subjects from 23 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of European and African ancestry. The discovery meta-analysis was conducted in 60,564 non-diabetic subjects from 23 studies with Exomechip data. Association analysis between genotype and fasting glucose level was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of log-transformed fasting plasma insulin level in cohorts of African ancestry. The discovery meta-analysis was conducted in 8,779 non-diabetic subjects from 7 studies with Exomechip data. Association analysis between genotype and fasting insulin level was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, BMI, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.
]]>The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of log-transformed fasting plasma insulin level in cohorts of European and African ancestry. The discovery meta-analysis was conducted in 43,771 non-diabetic subjects from 19 studies with Exomechip data. Association analysis between genotype and fasting insulin level was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, BMI, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma glucose level in cohorts of European ancestry. The discovery meta-analysis was conducted in 43,771 non-diabetic subjects from 19 studies with Exomechip data. Association analysis between genotype and fasting insulin levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, BMI, PCs, and study-specific covariates as needed. At the meta-analysis level, groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of log-transformed fasting plasma insulin level in cohorts of European and African ancestry. The discovery meta-analysis was conducted in 48,118 non-diabetic subjects from 20 studies with Exomechip data. Association analysis between genotype and fasting glucose levell was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, BMI, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of low-frequency and rare variant (from the Exomechip) effect using single variant tests and gene-burden tests on the outcome of fasting plasma insulin level in cohorts of European and African ancestry. The discovery meta-analysis was conducted in 48,118 non-diabetic subjects from 20 studies with Exomechip data. Association analysis between genotype and log-transformed fasting insulin level was conducted using the R package seqMeta, with linear regression. For family-based cohorts, linear mixed effects models were used. All studies used an additive coding of variants to the minor allele observed in the jointly called data set. All analyses were adjusted for age, sex, BMI, PCs, and study-specific covariates as needed. At the meta-analysis level, ancestral groups were analyzed separately and combined; Bonferroni correction was used for significance. all variants with MAF>0.02% were included in single variant tests (P < 3x10-7). For gene-burden tests, SKAT and the Weighted Sum Test was used with MAF<1% with P < 1x10-6.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on cerebral white matter lesion burden in 7 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation was made to the 2.5M non-monomorphic autosomal SNPs in the HapMap CEU panel. Within each study, a linear regression model was used with covariates. Association analysis was conducted separately within each study cohort according to a pre-specified plan; cohort-specific results were combined with fixed-effects meta-analysis using a z-score method with a genomic control parameter to remove residual population stratificationThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on cerebral white matter hyperintensities in 27 discovery cohorts of European, African, Hispanic and Asian ancestry; for analysis, there were two groups, European (EUR) and non-European (AFR, HIS and ASN). The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panel to obtain ~14M non-monomorphic autosomal SNPs for analysis. Within each study, a linear regression model was used with covariates to estimate the effect of each SNP on WMH assuming an additive genetic effect, with phenotype expressed as ln(WMH+1). Analysis was conducted separately within each study cohort according to a pre-specified plan; cohort-specific results were combined for EUR and (AFR, HIS, ASN) with fixed-effects meta-analysis using a z-score method, followed by meta-analysis of the two EUR and non-EUR groups using METAL software.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on cerebral white matter hyperintensities in 20 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panel to obtain ~14M non-monomorphic autosomal SNPs for analysis. Within each study, a linear regression model was used with covariates to estimate the effect of each SNP on WMH assuming an additive genetic effect, with phenotype expressed as ln(WMH+1). Analysis was conducted separately within each study cohort according to a pre-specified plan; cohort-specific results were combined with fixed-effects meta-analysis using a z-score method with a genomic control parameter to remove residual population stratification.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of SNP effect on paragraph recall in 8 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the HapMap CEU reference panel was performed obtain ~2.7M non-monomorphic autosomal SNPs for analysis. Meta-analyses with METAL using inverse-variance weighted meta-analysis to combine GWAS for the paragraph recall memory test.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of verbal declarative memory (combining multiple tests of word list recall or paragraph recall) in 19 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the HapMap CEU reference panel was performed obtain ~2.7M non-monomorphic autosomal SNPs for analysis. Meta-analyses with METAL using inverse-variance weighted meta-analysis to combine GWAS for the same memory test and effective sample size weighted meta-analysis to combine GWAS for nonidentical memory tests. For each SNP, the Z-statistic was weighted by the effective sample size. A combined estimate was obtained by summing the weighted Z statistics and dividing by the summed weights.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of hippocampal volume in 10 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the HapMap CEU reference panel was performed obtain ~2.5M non-monomorphic autosomal SNPs for analysis, as long as the minor allele frequency was > 1% and the SNP was reported in at least 2 studies. Meta-analyses with METAL using inverse-variance weighted meta-analysis to combine GWAS for the same memory test and effective sample size weighted meta-analysis to combine GWAS for nonidentical memory tests. For each SNP, the Z-statistic was weighted by the effective sample size. A combined estimate was obtained by summing the weighted Z statistics and dividing by the summed weights.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of silent brain infarcts (from MRI) in 7 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, followed by imputation to the HapMap CEU reference panel to generate ~2.2M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using logistic regression on phenotype (having or not at least 1 MRI infarct) with the same additive genetic model of dosages of the risk allele with a 1-df trend test. Meta-analyses used fixed effects inverse-variance weighting to combine GWAS summary statistics across the 7 cohorts. The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of brain volume in 6 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the HapMap CEU reference panel was performed obtain ~2.2M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using linear regression with the same basic additive genetic model of dosages of the risk allele with a 1-df trend test. Meta-analyses used inverse-variance weighting to combine GWAS summary statistics. The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of intracranial volume in 5 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the HapMap CEU reference panel was performed obtain ~2.2M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using linear regression with the same basic additive genetic model of dosages of the risk allele with a 1-df trend test. Meta-analyses used inverse-variance weighting to combine GWAS summary statistics. The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on incident stroke in 18 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panel (Health ABC imputed to HapMap release 22, build 36) to obtain ~14M non-monomorphic autosomal SNPs for analysis. Within each study, genome-wide multivariable Cox regression was used to test the association of SNPs with incident stroke (all stroke, ischemic stroke, cardioembolic stroke, non-cardioembolic stroke) undder an additive model, with covariates. Meta-analysis of study-specific association statistics used inverse variance weighted approach using METAL.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on incident ischemic stroke in 16 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panel to obtain ~14M non-monomorphic autosomal SNPs for analysis. Within each study, genome-wide multivariable Cox regression was used to test the association of SNPs with incident ischemic stroke under an additive model, with covariates. Meta-analysis of study-specific association statistics used inverse variance weighted approach using METALThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. CHARGE has incorporated many genomic and phenotypic data from other cohorts, depending on the phenotype. The basic design was estimation of single variant tests on incident ischemic stroke in 4 discovery cohorts of European ancestry. The genotyping was conducted with different platforms, so imputation was made to the ~2.5M non-monomorphic autosomal SNPs in the HapMap CEU panel. Within each study, a Cox proportional hazards model was used to evaluate time to first ischemic stroke; each study fit an additive genetic model with genotype dosage to total stroke; summary statistics were combined using inverse-variance weighting fixed-effects meta-analysis; SNPs with P < 5x10-8 was considered genome-wide significanceThe Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of heart rate from ECG in 3 discovery cohorts of Hispanic/Latino ancestry. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panels was performed obtain ~17M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using mixed models accounting for familial relationships (HCHS/SOL using R/Bioconductor GENESIS) and linear regression (MESA using SNPTEST2, WHI using ProbAbel). Results were combined using summary statistics from each cohort employing inverse variance-weighted meta-analysis.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of heart rate variability from 12-lead ECG in 3 discovery cohorts of Hispanic/Latino ancestry. The phenotype is the root mean squared difference in successive, normal-to-normal interbeat intervals (RMSSD, ms), natural log-transformed. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panels was performed obtain ~17M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using mixed models accounting for familial relationships (HCHS/SOL using R/Bioconductor GENESIS) and linear regression (MESA using SNPTEST2, WHI using ProbAbel). Results were combined using summary statistics from each cohort employing inverse variance-weighted meta-analysis.The Cohorts for Heart and Aging Research in Genomic Epimiology (CHARGE) consortium was formed to facilitate genome-wide association study meta-analyses and replication among large, well-phenotyped cohort studies. The basic design was estimation of heart rate variability from 12-lead ECG in 3 discovery cohorts of Hispanic/Latino ancestry. The phenotype is the standard deviation of normal-to-normal interbeat intervals (SDNN, ms), natural log-transformed. The genotyping was conducted with different platforms, so imputation to the 1000 Genomes Project reference panels was performed obtain ~17M non-monomorphic autosomal SNPs for analysis. Each study performed GWAS tests using mixed models accounting for familial relationships (HCHS/SOL using R/Bioconductor GENESIS) and linear regression (MESA using SNPTEST2, WHI using ProbAbel). Results were combined using summary statistics from each cohort employing inverse variance-weighted meta-analysis.EPA+DHA consumption was evaluated as the dependent variable using linear regression with robust standard error, adjusted for covariates. SNPs with low MAF (<1%), low imputation quality (MACH: R2<0.3; or IMPUTE: proper info <0.4), were excluded. Quality control for cohort-level GWAS results was performed to ensure correct specification of the minor allele and agreement in frequencies with the reference population (HapMap CEU), consistent distribution of effect sizes and standard error, and examination of QQ plots to assess any large inflation of test statistics. Results across studies were combined using fixed-effect meta-analysis with inverse variance weights (METAL software). The association results from individual studies as well as meta-analyses were adjusted for genomic control. A meta-analysis within cohorts from Europe and USA was performed. Genome-wide significance for the threshold of P<5x10-8. Fish consumption was evaluated as the dependent variable using linear regression with robust standard error, adjusted for covariates. SNPs with low MAF (<1%), low imputation quality (MACH: R2<0.3; or IMPUTE: proper info <0.4), were excluded. Quality control for cohort-level GWAS results was performed to ensure correct specification of the minor allele and agreement in frequencies with the reference population (HapMap CEU), consistent distribution of effect sizes and standard error, and examination of QQ plots to assess any large inflation of test statistics. Results across studies were combined using fixed-effect meta-analysis with inverse variance weights (METAL software). The association results from individual studies as well as meta-analyses were adjusted for genomic control. A meta-analysis within cohorts from Europe and USA was performed. Genome-wide significance for the threshold of P<5x10-8. Genome-wide association analyses were conducted by each study, with BMI regressed on covariates to obtain residuals, and the residuals were inverse-normally transformed to obtain a standard normal distribution with mean zero and standard deviation one. For studies with unrelated subjects, each SNP was tested assuming an additive genetic model by regressing the transformed residuals on the number of copies of the SNP effect allele. The analyses were stratified by sex and case-control status (if needed). For studies with related subjects, family-based association tests were used, with sex-stratified, case-control stratified and combined analyses performed. Association results were combined across studies in sex-combined and sex-stratified samples using inverse-variance weighted fixed effect meta-analysis (METAL). BMI analyses used ~18M SNPs for BMI, with genome-wide significance attaining a P<5x10-8; variants with P<10-4 were carried to replication. Separate analyses were conducted for all subjects, men only, and women only.Genome-wide association analyses were conducted in men by each study, with BMI regressed on covariates to obtain residuals, and the residuals were inverse-normally transformed to obtain a standard normal distribution with mean zero and standard deviation one. For studies with unrelated subjects, each SNP was tested assuming an additive genetic model by regressing the transformed residuals on the number of copies of the SNP effect allele. The analyses were for men only and case-control status (if needed). For studies with related subjects, family-based association tests were used, with sex-stratified, case-control stratified and combined analyses performed. Association results were combined across studies in men, using inverse-variance weighted fixed effect meta-analysis (METAL). BMI analyses used ~18M SNPs for BMI, with genome-wide significance attaining a P<5x10-8; variants with P<10-4 were carried to replication. Genome-wide association analyses were conducted by each study in women only, with BMI regressed on covariates to obtain residuals, and the residuals were inverse-normally transformed to obtain a standard normal distribution with mean zero and standard deviation one. For studies with unrelated subjects, each SNP was tested assuming an additive genetic model by regressing the transformed residuals on the number of copies of the SNP effect allele. The analyses were stratified by case-control status (if needed). For studies with related subjects, family-based association tests were used, case-control stratified and combined analyses performed. Association results were combined across studies in women using inverse-variance weighted fixed effect meta-analysis (METAL). BMI analyses used ~18M SNPs for BMI, with genome-wide significance attaining a P<5x10-8; variants with P<10-4 were carried to replication. Genome-wide association analyses were conducted by each study, with WHR regressed on covariates to obtain residuals, and the residuals were inverse-normally transformed to obtain a standard normal distribution with mean zero and standard deviation one. For studies with unrelated subjects, each SNP was tested assuming an additive genetic model by regressing the transformed residuals on the number of copies of the SNP effect allele. The analyses were stratified by sex and case-control status (if needed). For studies with related subjects, family-based association tests were used, with sex-stratified, case-control stratified and combined analyses performed. Association results were combined across studies in sex-combined and sex-stratified samples using inverse-variance weighted fixed effect meta-analysis (METAL). Analyses used ~20.5M SNPs for WHRadjBMI with genome-wide significance attaining a P<5x10-8; variants with P<10-4 were carried to replication. Separate analyses were conducted for all subjects, men only, and women only.The analysis of WHRadjBMI in African ancestry male participants utilized GWAS data imputed to 1000 Genomes Project Phase 1 v3 data. Discovery was based upon 8 studies. Genotyping in each study was conducted with Affymetrix or Illumina arrays, with quality control and imputation performed by each study (~20.5M variants). WHR residuals were normalized and were included in inverse-variance weighted fixed effect meta-analysis. A novel locus was declared if teh most significant SNP was >500 kb from an established lead SNP in a previous study. A locus was named by the closest gene to the most associated variant.Genome-wide association analyses were conducted by each study, with WHR in women regressed on covariates to obtain residuals, and the residuals were inverse-normally transformed to obtain a standard normal distribution with mean zero and standard deviation one. For studies with unrelated subjects, each SNP was tested assuming an additive genetic model by regressing the transformed residuals on the number of copies of the SNP effect allele. The analyses were stratified by case-control status (if needed). For studies with related subjects, family-based association tests were used, case-control stratified and combined analyses performed. Association results were combined across studies in women using inverse-variance weighted fixed effect meta-analysis (METAL). Analyses used ~20.5M SNPs for WHRadjBMI with genome-wide significance attaining a P<5x10-8; variants with P<10-4 were carried to replication. Stage 1 discovery samples came from 17 T2D studies (FIND and WFSM were combined into a single analysis group), with association results combined by a fixed effect model with inverse variance weighted method (METAL). Genomic control correction was applied to each study prior to meta-analysis. Results from SNPs genotyped in <10,000 samples and those with minor allele frequency difference >0.3 across studies were excluded, resulting in ~2.5M SNPs for analysis. Following discovery, replications samples were used and combined in meta-analysis, with a final effect esimation performed with all samples. Genome-wide significance was based upon P<5x10-8 criterion from meta-analysis of the combined discovery and replication samples. Logistic regression was used for samples from unrelated individuals; GEE or SOLAR were used for samles of related individuals. The data are in the public ftp site.
]]>The data are in the public ftp site.
]]>The data are in the public ftp site.
]]>The data are in the public ftp site.
]]>Single Nucleotide Variant (SNV) effect size estimates were calculated using standard inverse variance weighted meta-analysis of results provided by each cohort from a linear model with JT interval as the dependent variable, including covariates. Significance (P-value) was determined by first inverse rank normal transforming residuals from a linear model with JT as the outcome using covariates, followed by a standard IVW meta-analysis on a linear association model with the transformed residuals as the outcome using cohort specific adjustments as covariates. JT loci are considered discovered if passing aBonferroni correction, P< 0.05 / 208,917 SNVs (P<2E-07). Variants with minor allele counts less than 10 were excluded from the meta-analysis. SKAT tests were performed using the R package seqMeta with rare variants (MAF ≤ 0.01) from each gene. Variants were filtered to those that alter protein coding: frame-shift, nonsynonymous, stop-gain, stop-loss, or splicing. In a second analysis, the nonsynonymous variants were further filtered to those predicted to be damaging by at least two prediction algorithms (Polyphen2, LRT, SIFT, Mutation Taster1). Genes with only a single variant were excluded. Bonferroni corrected ExomeChip-wide significance was defined by P<0.05 divided by 29,366 genes for JT interval.Single Nucleotide Variant (SNV) effect size estimates were calculated using standard inverse variance weighted meta-analysis of results provided by each cohort from a linear model with JT interval as the dependent variable, including covariates. Significance (P-value) was determined by first inverse rank normal transforming residuals from a linear model with JT as the outcome using covariates, followed by a standard IVW meta-analysis on a linear association model with the transformed residuals as the outcome using cohort specific adjustments as covariates. QT loci are considered discovered if passing a Bonferroni correction, P< 0.05 / 208,917 SNVs (P<2E-07). Variants with minor allele counts less than 10 were excluded from the meta-analysis. SKAT tests were performed using the R package seqMeta with rare variants (MAF ≤ 0.01) from each gene. Variants were filtered to those that alter protein coding: frame-shift, nonsynonymous, stop-gain, stop-loss, or splicing. In a second analysis, the nonsynonymous variants were further filtered to those predicted to be damaging by at least two prediction algorithms (Polyphen2, LRT, SIFT, Mutation Taster1). Genes with only a single variant were excluded. Bonferroni corrected ExomeChip-wide significance was defined by P<0.05 divided by 29,368 genes for QT interval.In each study, PR interval was linearly regressed on allele dosage (0?2, either measured directly or imputed), adjusting for covariates and related individuals (in FHS and SardiNIA) and case-control sampling probabilities (in CHS). Associations at P<5E-08 are considered significant. Meta-analysis used inverse variance weighting in METAL.In cohorts with multiple ECGs per participant over time (ARIC, WHI), ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In studies with one ECG per participant (MESA, CHS), associations were estimated using logistic regression (SNPTEST, R geeglm package). Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).In cohorts with multiple ECGs per participant over time (ARIC, WHI), ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In studies with one ECG per participant (MESA, CHS), associations were estimated using logistic regression (SNPTEST, R geeglm package). Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).In WHI, a cohort with multiple ECGs per participant over time, ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In MESA, with one ECG per participant, associations were estimated using logistic regression (SNPTEST, R geeglm package). In HCHS/SOL, associations were estimated among unrelated (at the 3rd degree level) participants (one per household) using a generalized linear model and a Firth test to account for small numbers of cases (R logistf package), assuming Census block group effects were negligible. Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).In cohorts with multiple ECGs per participant over time (ARIC, WHI), ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In studies with one ECG per participant (MESA, CHS), associations were estimated using logistic regression (SNPTEST, R geeglm package). In HCHS/SOL, associations were estimated among unrelated (at the 3rd degree level) participants (one per household) using a generalized linear model and a Firth test to account for small numbers of cases (R logistf package), assuming Census block group effects were negligible. Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Trans-ethnic Bayesian meta-analysis (MANTRA) was used to account for allelic heterogeneity among ancestry groups. MANTRA clustered similar populations according to allele frequencies, allowed for varying allele effects across populations, and produced Bayes' factors (BFs) for each ectopy-SNP association and its posterior probability of heterogeneity (Phet).In cohorts with multiple ECGs per participant over time (ARIC, WHI), ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In studies with one ECG per participant (MESA, CHS), associations were estimated using logistic regression (SNPTEST, R geeglm package). Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).In cohorts with multiple ECGs per participant over time (ARIC, WHI), ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In studies with one ECG per participant (MESA, CHS), associations were estimated using logistic regression (SNPTEST, R geeglm package). Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).In WHI, a cohort with multiple ECGs per participant over time, ectopy-SNP associations were estimated using generalized estimating equation (GEE) methods, a logit link, and an exchangeable working correlation structure to control for correlation of repeated measures (R geepack package). In HCHS/SOL, associations were estimated among unrelated (at the 3rd degree level) participants (one per household) using a generalized linear model and a Firth test to account for small numbers of cases (R logistf package), assuming Census block group effects were negligible. Within subgroups, observed P-values were compared for each SNP with expected values from a Χ2 distribution using quantile-quantile (Q-Q) plots and genomic inflation factors (lambda). To eliminate statistical artifacts at low allele and ectopy frequencies, the comparisons excluded SNPs with an effective number of minor alleles present in exposed participants (defined as 2 x number of exposed participants x minor allele frequency x imputation quality) <10 or a log odds ratio >10. Analyses involved combining ancestry-specific summary results in fixed-effects, inverse-variance-weighted meta-analyses with genomic control (METAL).The data are in
the public ftp site.
]]>The association of aPTT with genotyped or imputed SNPs was tested with the use of an additive genetic model in accordance with a unified analysis plan. ARIC was the discovery cohort and other studies provided replication. Outliers or users of anticoagulants were excluded. Analysis was conducted with linear regression models for studies of unrelated individuals. A score test was used for testing the genetic associations while accounting for familial correlations in MICROS, a population study that contains related individuals in extended families. All analyses were adjusted for age, sex (except CaPS), and other cohort-specific covariates and were performed using ProbABEL-GenABEL, Mach2QTL or PLINK. For aPTT, an effective-sample-size weighted meta-analysis was performed to combine p-values and the direction of genetic effects from the replication (MICROS,LBC1921, LBC1936, and CaPS) and discovery (ARIC) cohorts. The p value-based approach was chosen because of mixed aPTT and aPTT-ratio measurements across cohorts. The genomic-control correction was applied to each study before the meta-analysis. The genomic-control inflation factor l was around 1 (0.98?1.03) for aPTT in all cohorts and in the meta-analyses. The meta-analyses were conducted with the METAL package. The MARTHA data were not pooled with the community-based cohorts because of the possibility of different allele frequencies, linkage disequilibria (LD), and effect sizes in individuals with VTE compared to the general population.The association of aPTT with genotyped or imputed SNPs was tested with the use of an additive genetic model in accordance with a unified analysis plan. ARIC was the discovery cohort and other studies provided replication. Outliers or users of anticoagulants were excluded. Analysis was conducted with linear regression models for studies of unrelated individuals. A score test was used for testing the genetic associations while accounting for familial correlations in MICROS, a population study that contains related individuals in extended families. All analyses were adjusted for age, sex (except CaPS), and other cohort-specific covariates and were performed using ProbABEL-GenABEL, Mach2QTL or PLINK. For aPTT, an effective-sample-size weighted meta-analysis was performed to combine p-values and the direction of genetic effects from the replication (MICROS,LBC1921, LBC1936, and CaPS) and discovery (ARIC) cohorts. The p value-based approach was chosen because of mixed aPTT and aPTT-ratio measurements across cohorts. The genomic-control correction was applied to each study before the meta-analysis. The genomic-control inflation factor l was around 1 (0.98?1.03) for aPTT in all cohorts and in the meta-analyses. The meta-analyses were conducted with the METAL package. The MARTHA data were not pooled with the community-based cohorts because of the possibility of different allele frequencies, linkage disequilibria (LD), and effect sizes in individuals with VTE compared to the general population.The association of D-dimer with genotyped SNPs was tested with the use of an additive genetic model in accordance with a unified analysis plan. All studies used linear regression to conduct association analyses between measured and imputed SNPs and natural-log?transformed D-dimer measures except for CROATIA-Vis, FHS, and ORCADES, which used a linear mixed-effects model to account for family relationships, and TwinsUK that used a score test and variance components method asvimplemented in Merlin. For each analysis, a genomic control coefficient, that estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. Within-study findings were combined across studies to produce summary results using standard meta-analytic approaches. A fixed-effects, inverse-variance weighted meta-analysis was performed, and summary P-values and beta-coefficients were calculated. All meta-analyses were conducted with MetABEL. The a priori threshold of genome-wide significance was set at a value of P<5.0x10-8.Each study independently analyzed their genotype-phenotype data. Except for FHS, which has a family structure, all studies conducted analyses of all directly genotyped and imputed SNPs using linear regression on untransformed fibrinogen measures using an additive genetic model adjusted for covariates. In FHS, a linear mixed effects model was used with a fixed additive effect for the SNP genotype, fixed covariate effects, random family-specific additive residual polygenic effects to account for within family correlations, and a random environment effect. In addition, FHS adjusted for population stratification using principal components of the directly measured SNPs, which were computed using Eigenstrat. To account for residual stratification, P-values were adjusted for genomic inflation which was small for all studies Using the studyspecific results, a fixed effect model meta-analysis based on inverse-variance weighting was used. MetABEL, a package running under R, was used to perform the meta-analysis. Bonferroni correction was used to deal with the problem of multiple testing. The a priori threshold of genome-wide significance was set at P<5.0x10-8.Within each study, linear regression was used to model the additive effect of variants on FEV1, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for African ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FEV1, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Asian ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FEV1, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size of 20,184 for European ancestry. Significance was defined as P<5x10-8.Within each Hispanic/Latino subgroup, linear regression was used to model the additive effect of variants on FEV1, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Hispanic/Latino ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FEV1, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Multiethnic random effects meta-analyses using the four ancestry-specific fixed effects meta-analysis results were conducted using the Han-Eskin model in METASOFT. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models.Within each study, linear regression was used to model the additive effect of variants on FVC, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for African ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FVC, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Asian ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FVC, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size of 20,184 for European ancestry. Significance was defined as P<5x10-8.Within each Hispanic/Latino subgroup, linear regression was used to model the additive effect of variants on FVC, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Hispanic/Latino ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on FVC, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Multiethnic random effects meta-analyses using the four ancestry-specific fixed effects meta-analysis results were conducted using the Han-Eskin model in METASOFT. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models.Within each study, linear regression was used to model the additive effect of variants on the FEV1/FVC ratio, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for African ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on the FEV1/FVC ratio, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Asian ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on the FEV1/FVC ratio, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size of 20,184 for European ancestry. Significance was defined as P<5x10-8Within each Hispanic/Latino subgroup, linear regression was used to model the additive effect of variants on the FEV1/FVC ratio, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Variants with imputation quality scores (r2) less than 0.3 and/or a minor allele count (MAC) less than 20 were excluded from each study prior to meta-analysis. Following meta-analysis, variants were excluded with less than one-third the total sample size or less than the sample size of the largest study for a given meta-analysis to achieve the minimal sample size for Hispanic/Latino ancestry. Significance was defined as P<5x10-8.Within each study, linear regression was used to model the additive effect of variants on the FEV1/FVC ratio, modeled as milliliters. Studies were asked to adjust analyses for covariates. Analyses were conducted using ProbAbel, PLINK, FAST, or the R kinship package. Ancestry-specific and multiethnic fixed effects meta-analyses using inverse variance weighting of study-specific results with genomic control correction were conducted in METAL. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models. Multiethnic random effects meta-analyses using the four ancestry-specific fixed effects meta-analysis results were conducted using the Han-Eskin model in METASOFT. Only variants with p-values for association <0.05 or p-values for heterogeneity <0.1 from fixed-effects models were included in the random-effects models.Each study performed GWAS according to a uniform analysis plan by regressing covariate-adjusted residuals of the natural logarithm of eGFRcrea on the allelic dosage levels. All GWAS files underwent quality control using the GWAtoolbox package. GWAS meta-analyses for eGFRcrea were performed using the software METAL assuming fixed effects across studies and using inverse-variance weighting, excluding variants with imputation quality IQ ? 0.4 or variants present in less than 50% of the 110,517 subjects (yielding 10,971,307 variants). Genomic-control (GC) correction was applied to p-values and SEs in case of ? > 1 (1st GC correction). To limit the possibility of false positives, a second GC correction on the aggregated results was applied after the meta-analysis. Between-study heterogeneity was assessed with the I2 statistic.Each study performed GWAS according to a uniform analysis plan by regressing covariate-adjusted residuals of the natural logarithm of eGFRcys on the allelic dosage levels. All GWAS files underwent quality control using the GWAtoolbox package. GWAS meta-analyses for eGFRcys were performed using the software METAL assuming fixed effects across studies and using inverse-variance weighting, excluding variants with imputation quality IQ ? 0.4 or variants present in less than 50% of the 24,063 subjects. Genomic-control (GC) correction was applied to p-values and SEs in case of ? > 1 (1st GC correction). To limit the possibility of false positives, a second GC correction on the aggregated results was applied after the meta-analysis. Between-study heterogeneity was assessed with the I2 statistic.For each gene in the curated gene list (258 genes), the stage 1 meta-analysis results were examined for association between eGFR and all common SNPs in the gene region, restricted to SNPs with MAF>5% Each gene region was defined as 10 kb upstream and 5 kb downstream of the known transcription start and end sites of the gene (build 36). Each of the participating cohorts of the CKDGen consortium performed GWAS of eGFR with linear regression. An additive genetic effect model for the genotype dosages was used, adjusting for covariates. Study-specific GWAS data were subjected to inverse-variance weighted fixed-effects meta-analysis using METAL Bonferroni gene-specific significance thresholds were defined as 0.05/(the number of independent tests within the gene), where the denominator is the number of independent LD blocks (independent blocks identifiedby grouping SNPs in LD (r2>0.20, based on HapMap Phase 2 release 21).SNPs were modeled as allelic dosages in all analyses. Genome-wide analyses of magnesium concentration were conducted within the R package ProbABEL for ARIC and RS, or using linear mixed effects regression models in the R kinship package to account for pedigree structure in FHS. SNP-magnesium associations were adjusted for covariates. Genomic control correction based on median chi-square was used within each study to adjust for inflation of the test statistics prior to meta-analysis, as well as applied to the combined results after the meta-analysis. Inverse variance-weighted fixed-effects meta-analyses were carried out by two independent analysts using the software METAL After meta-analysis, results were filtered to remove SNPs with low minor allele frequency (<0.01). Statistical heterogeneity was evaluated using Cochrane's x2 test (Q-test). P-values < 5x10-8 were used to indicate genome-wide significant results.Individual GWAS analyses encompassing approximately 2.5 million imputed SNPs were performed within each of the 12 discovery CKDGen population-based cohorts.For cohorts consisting of only unrelated individuals, we performed logistic regression for microalbuminuria. For family-based studies, we applied logistic regression via generalized estimating equations (GEE) for microalbuminuria to account for the familial relatedness. Meta-analyses combining study-specific microalbuminuria GWAS analysis results used the inverse-variance weighted fixed effect model for combining beta estimates. From these sets of meta-analysis results, a list of independent SNPs were selected (pairwise r2 < 0.2) with a P < 1x10-6 and minor allele frequencies (MAF) < 5%.Individual GWAS analyses encompassing approximately 2.5 million imputed SNPs were performed within each of the 11 discovery CKDGen population-based cohorts with UACR measured. UACR was log-transformed, and sex-specific residuals were computed by regression on age and, in multicenter studies, on study center. For cohorts consisting of only unrelated individuals, we performed linear regression for UACR. For family-based studies, we applied linear regression via generalized estimating equations (GEE) for UACR to account for the familial relatedness. Meta-analyses combining study-specific UACR GWAS analysis results used the inverse-variance weighted fixed effect model for combining beta estimates. From these sets of meta-analysis results, a list of independent SNPs were selected (pairwise r2 < 0.2) with a P < 1x10-6 and minor allele frequencies (MAF) < 5%.All GWAS were performed following a standardized analysis protocol. Sex-specific residuals were obtained from linear regression models of ln (UACR) on covariates. The continuous sex-specific residuals were then combined and used as the dependent variable that was regressed on imputed allelic dosages for each SNP in the GWAS. Before the meta-analyses, all study-specific GWAS summary files underwent quality control using the GWAtoolbox. Genomic-control (GC) correction was applied when the GC factor was >1. Inverse variance weighted fixed-effects meta-analyses were conducted using METAL. The I2 statistic was used to evaluate between-study heterogeneity. All meta-analyses were performed in duplicate by two independent researchers. After meta-analysis, SNPs with average minor allele frequency (MAF) <0.01 were excluded, and another GC correction was applied. There were 2,191,945 SNPs with average MAF >0.05 and present in >50% of the studies. Genome-wide significance was defined by P < 5x10-8.All GWAS were performed following a standardized analysis protocol. Sex-specific residuals were obtained from linear regression models of ln (UACR) on covariates. The continuous sex-specific residuals were then combined and used as the dependent variable that was regressed on imputed allelic dosages for each SNP in the GWAS. Before the meta-analyses, all study-specific GWAS summary files underwent quality control using the GWAtoolbox. Genomic-control (GC) correction was applied when the GC factor was >1. Inverse variance weighted fixed-effects meta-analyses were conducted using METAL. The I2 statistic was used to evaluate between-study heterogeneity. All meta-analyses were performed in duplicate by two independent researchers. After meta-analysis, SNPs with average minor allele frequency (MAF) <0.01 were excluded, and another GC correction was applied. There were 2,191,945 SNPs with average MAF >0.05 and present in >50% of the studies. Genome-wide significance was defined by P < 5x10-8.All GWAS were performed following a standardized analysis protocol restricted to subjects with type 2 diabetes. Sex-specific residuals were obtained from linear regression models of ln (UACR) on covariates. The continuous sex-specific residuals were then combined and used as the dependent variable that was regressed on imputed allelic dosages for each SNP in the GWAS. Before the meta-analyses, all study-specific GWAS summary files underwent quality control using the GWAtoolbox. Genomic-control (GC) correction was applied when the GC factor was >1. Inverse variance weighted fixed-effects meta-analyses were conducted using METAL. The I2 statistic was used to evaluate between-study heterogeneity. All meta-analyses were performed in duplicate by two independent researchers. After meta-analysis, SNPs with average minor allele frequency (MAF) <0.01 were excluded, and another GC correction was applied. There were 2,191,945 SNPs with average MAF >0.05 and present in >50% of the studies. Genome-wide significance was defined by P < 5x10-8.All GWAS were performed following a standardized analysis protocol in subjects wtihout diagnosed (or self-reported) diabetes. Sex-specific residuals were obtained from linear regression models of ln (UACR) on covariates. The continuous sex-specific residuals were then combined and used as the dependent variable that was regressed on imputed allelic dosages for each SNP in the GWAS. Before the meta-analyses, all study-specific GWAS summary files underwent quality control using the GWAtoolbox. Genomic-control (GC) correction was applied when the GC factor was >1. Inverse variance weighted fixed-effects meta-analyses were conducted using METAL. The I2 statistic was used to evaluate between-study heterogeneity. All meta-analyses were performed in duplicate by two independent researchers. After meta-analysis, SNPs with average minor allele frequency (MAF) <0.01 were excluded, and another GC correction was applied. There were 2,191,945 SNPs with average MAF >0.05 and present in >50% of the studies. Genome-wide significance was defined by P < 5x10-8.Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The dichotomous trait (CKDi25) was analyzed by logistic regression in those with a decline in kidney function of at least 25% at follow-up and progressing to the clinical outcome CKD stage 3 or higher in cases (CKDi25) with controls having no evidence of CKD at baseline or followup. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 15 population-based GWAS of individuals of European ancestry were used. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The dichotomous trait (CKDi) was analyzed by logistic regression in those with a decline in kidney function to the clinical outcome CKD stage 3 or higher in cases (CKDi) with controls having no evidence of CKD at baseline or followup. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 15 population-based GWASs of individuals of European ancestry were used. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The continuous trait (eGFRchange) was analyzed by linear regression in subjects without evidence of CKD at baseline. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 15 population-based GWASs of individuals of European ancestry were used for eGFRchange. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The continuous trait (eGFRchange) was analyzed by linear regression. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 16 population-based GWASs of individuals of European ancestry were used for eGFRchange overall. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The continuous trait (eGFRchange) was analyzed by linear regression in those with eGFR> ml/min per 1.73ml at baseline (eGFRchange withCKD). Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 15 population-based GWASs of individuals of European ancestry were used for eGFRchange. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The dichotomous trait (Rapid Decline in eGFR) was analyzed by logistic regression. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 16 population-based GWAS of individuals of European ancestry were used. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 GWAS meta-analysis was performed in all samples. All participating studies used a uniform analysis plan and each trait was created using standard programming commands that were provided to collaborating studies. The dichotomous trait (Rapid Decline in eGFR) was analyzed by logistic regression. Models included the allelic dosage at each marker from imputed study data consisting of 2.5M HapMap-II SNPs on average, based on imputations with different programs and reference panels. We used the additive genetic model, adjusted for covariates. For stage 1 analysis, aggregated statistics of 16 population-based GWAS of individuals of European ancestry were used. All input files underwent quality control using the GWAtoolbox package in R before including them into the meta-analysis. Study data were meta-analyzed assuming fixed effects and using inverse-variance weighting. The meta-analyses were performed by METAL. We performed genomic control correction if the inflation factor in the study files was greater than 1 (first genomic control correction) or if it was greater than 1 in the meta-analysis result (second genomic control correction). Stage 1 discovery GWAS meta-analysis of eGFRcrea were contributed by 48 studies (total sample size, N=133,413. All GWAS files underwent quality control using the GWAtoolbox package in R, before including them into the meta-analysis. Genome-wide meta-analysis was performed with the software METAL, assuming fixed effects and using inverse-variance weighting. The genomic inflation factor lambda was estimated for each study as the ratio between the median of all observed test statistics (beta/s.e.)**2 and the expected median of a chi-square with 1 degree of freedom, with beta and s.e. representing the effect of each SNP on the phenotype and its standard error, respectively. Genomic-control (GC) correction was applied to P values and s.e.'s in case of lambda>1 (first GC correction). SNPs with an average minor allele frequency (MAF) of >0.01 were used for the meta-analysis. After the meta-analysis, a second GC correction on the aggregated results was applied. Between-study heterogeneity was assessed through the I2 statistic. By following a centralized analysis plan, each study regressed sex- and age-adjusted residuals of the logarithm of eGFRcrea on SNP dosage levels. Adjustment for covariates was included in the regression and family-based studies appropriately accounted for relatedness.Stage 1 discovery GWAS meta-analysis of eGFRcrea were contributed by 45 studies in those participants with diabetes. All GWAS files underwent quality control using the GWAtoolbox package in R, before including them into the meta-analysis. Genome-wide meta-analysis was performed with the software METAL, assuming fixed effects and using inverse-variance weighting. The genomic inflation factor lambda was estimated for each study as the ratio between the median of all observed test statistics (beta/s.e.)**2 and the expected median of a chi-square with 1 degree of freedom, with beta and s.e. representing the effect of each SNP on the phenotype and its standard error, respectively. Genomic-control (GC) correction was applied to P values and s.e.'s in case of lambda>1 (first GC correction). SNPs with an average minor allele frequency (MAF) of >0.01 were used for the meta-analysis. After the meta-analysis, a second GC correction on the aggregated results was applied. Between-study heterogeneity was assessed through the I2 statistic. By following a centralized analysis plan, each study regressed sex- and age-adjusted residuals of the logarithm of eGFRcrea on SNP dosage levels. Adjustment for covariates was included in the regression and family-based studies appropriately accounted for relatedness.Stage 1 discovery GWAS meta-analysis of eGFRcrea were contributed by 44 studies (total sample size. All GWAS files underwent quality control using the GWAtoolbox package in R, before including them into the meta-analysis. Genome-wide meta-analysis was performed with the software METAL, assuming fixed effects and using inverse-variance weighting. The genomic inflation factor lambda was estimated for each study as the ratio between the median of all observed test statistics (beta/s.e.)**2 and the expected median of a chi-square with 1 degree of freedom, with beta and s.e. representing the effect of each SNP on the phenotype and its standard error, respectively. Genomic-control (GC) correction was applied to P values and s.e.?s in case of lambda>1 (first GC correction). SNPs with an average minor allele frequency (MAF) of >0.01 were used for the meta-analysis. After the meta-analysis, a second GC correction on the aggregated results was applied. Between-study heterogeneity was assessed through the I2 statistic. By following a centralized analysis plan, each study regressed sex- and age-adjusted residuals of the logarithm of eGFRcrea on SNP dosage levels. Adjustment for covariates was included in the regression and family-based studies appropriately accounted for relatedness.Stage 1 discovery GWAS meta-analysis of GWAS of CKD were comprised by 43 studies, for a total sample size of 117,165, including 12,385 CKD cases. All GWAS files underwent quality control using the GWAtoolbox package in R, before including them into the meta-analysis. Genome-wide meta-analysis was performed with the software METAL, assuming fixed effects and using inverse-variance weighting. The genomic inflation factor lambda was estimated for each study as the ratio between the median of all observed test statistics (beta/s.e.)**2 and the expected median of a chi-square with 1 degree of freedom, with beta and s.e. representing the effect of each SNP on the phenotype and its standard error, respectively. Genomic-control (GC) correction was applied to P values and s.e.*s in case of lambda>1 (first GC correction). SNPs with an average minor allele frequency (MAF) of >0.01 were used for the meta-analysis. After the meta-analysis, a second GC correction on the aggregated results was applied. Between-study heterogeneity was assessed through the I2 statistic. By following a centralized analysis plan, each study regressed sex- and age-adjusted residuals of the logarithm of eGFRcrea on SNP dosage levels. Adjustment for covariates was included in the regression and family-based studies appropriately accounted for relatedness.Stage 1 discovery GWAS was performed using fixed-effect meta-analysis of the genome-wide association data from 12 African ancestry studies (17,721 subjects) with imputation to HapMap reference panel, based on inverse-variance weighting using METAL. Only SNPs with MAF >0.01 and imputation quality r2>0.3 were considered for the meta-analysis. Statistical significance was assessed at the standard threshold of P<5.0x10-8. Genomic control correction was applied at both the individual study level before metaanalysis and after the meta-analysis.Stage 1 discovery GWAS meta-analysis of eGFRcys were contributed by 16 studies (total sample size, N=32,904. All GWAS files underwent quality control using the GWAtoolbox package in R, before including them into the meta-analysis. Genome-wide meta-analysis was performed with the software METAL, assuming fixed effects and using inverse-variance weighting. The genomic inflation factor lambda was estimated for each study as the ratio between the median of all observed test statistics (beta/s.e.)**2 and the expected median of a chi-square with 1 degree of freedom, with beta and s.e. representing the effect of each SNP on the phenotype and its standard error, respectively. Genomic-control (GC) correction was applied to P values and s.e.'s in case of lambda>1 (first GC correction). SNPs with an average minor allele frequency (MAF) of >0.01 were used for the meta-analysis. After the meta-analysis, a second GC correction on the aggregated results was applied. Between-study heterogeneity was assessed through the I2 statistic. By following a centralized analysis plan, each study regressed sex- and age-adjusted residuals of the logarithm of eGFRcrea on SNP dosage levels. Adjustment for covariates was included in the regression and family-based studies appropriately accounted for relatedness.Stage 1 discovery meta-analysis of UACR used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses were focused on UACR in subjects of African ancestry. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed UACR. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of UACR used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The primary meta-analyses were focused on the European ancestry population. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcrea used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. Meta-analyses were performed on the African ancestry population with diabetes. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bi-directional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 utosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcrea used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses were conduted on those with diabetes of European ancestry. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bi-directional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcrea used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses were focused on eGFRcrea in subjects without diabetes of African ancestry. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcrea in subjects without diabetes used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses were focused on the European ancestry non-diabetic population. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6.Stage 1 discovery meta-analysis of eGFRcrea used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. Meta-analyses were performed on the African ancestry population. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcrea used a centralized analysis plan. Each study performed two setsof analyses: single-variant analysis and gene-based analysis. The primarymeta-analyses were focused on the European ancestry population. Typically, R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyseswere performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effortwithin the CHARGE consortium, and (4) with minor allele count >20 across all cohorts . Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcys used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses used data from the European ancestry populations. R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts. Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Stage 1 discovery meta-analysis of eGFRcys used a centralized analysis plan. Each study performed two sets of analyses: single-variant analysis and gene-based analysis. The meta-analyses used data from the European ancestry populations. R software was used for data management, statistical analyses, and graphing. Each study performed association analyses of ln-transformed eGFRcrea. These association analyses were based on linear regression models adjusting for covariates. All analyses were performed assuming an additive genetic effect. For single-variant meta-analysis, study-specific results were combined in a fixed-effects model using METAL, restricted to (1) autosomal variants, (2) polymorphic variants, (3) variants existing in the joint calling effort within the CHARGE consortium, and (4) with minor allele count >20 across all cohorts. Bonferroni correction was used to set the significance threshold for chip-wide significance (P<3.7x10-7). For gene-based meta-analysis, study-specific results were combined using the seqMeta package for R. Two gene-based tests were used for aggregated analysis of SNVs: (1) the T1 test, which is more powerful when all variants within the gene region affect the phenotype in the same direction; and (2) SKAT, which allows for bidirectional effects and is more powerful when there are both protective and deleterious variants within the same gene. Both gene-based tests were restricted to variants with MAF<1% and variants likely to exert a major effect on the gene product (stop gain/loss, nonsynonymous, or splice-site variants on the basis of annotation with dbNSFP [v.2.0]). Genes containing at least two variants with a cumulative minor allele count >20 were included in the analysis. In total, 9990 autosomal genes were tested. Bonferroni correction for the number of genes and tests performed was used, corresponding to P<2.5x10-6. Each study independently analyzed its genotype-phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs over the autosomes and untransformed factor VII activity (FHS) and factor VII antigen (ARIC, CHS, RS). FHS used a linear mixed-effects model to account for family relationships. An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches. FVII results combined analysis of activity and antigen measures, so effective-sample-size?weighted meta-analysis was performed to estimate summary P-values only. All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype-phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs and untransformed phenotype measures except for FHS, which used a linear mixed-effects model to account for family relationships. An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. For females, genotypes were coded as 0, 1, and 2 representing the number of copies of the coded alleles; for males, genotypes were coded as 0 and 2. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches. FVII level combined analysis of activity and antigen measures, so effective-sample-size?weighted meta-analysis was performed to estimate summary P-values only. All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype-phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs and untransformed phenotype measures An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches. For factor VIII activity levels, fixed-effects, inverse-variance?weighted meta-analysis was performed, and summary P-values and beta-coefficients were calculated. All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype- phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs and untransformed phenotype measures. An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. For females, genotypes were coded as 0, 1, and 2 representing the number of copies of the coded alleles; for males, genotypes were coded as 0 and 2. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches. For factor VIII activity levels, fixed-effects, inverse-variance?weighted meta-analysis was performed, and summary P-values and beta-coefficients were calculated. All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype-phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs and untransformed vWF antigen except for FHS, which used a linear mixed-effects model to account for family relationships. An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches.A fixed-effects, inverse-variance?weighted meta-analysis was performed, and summary P-values and beta-coefficients were calculated All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype- phenotype data. All studies used linear regression to conduct association analyses between measured and imputed SNPs and untransformed vWF antigen except for FHS, which used a linear mixed-effects model to account for family relationships. An additive genetic model with 1 degree of freedom was adjusted for covariates. For each analysis, a genomic control coefficient, which estimated the extent of underlying population structure on the basis of test-statistic inflation, was used to adjust standard errors. For females, genotypes were coded as 0, 1, and 2, representing the number of copies of the coded alleles; for males, genotypes were coded as 0 and 2. Within-study findings were combined across studies to produce summary results by use of standard meta-analytic approaches.A fixed-effects, inverse-variance?weighted meta-analysis was performed, and summary P-values and beta-coefficients were calculated All meta-analyses were conducted with METAL software The a priori threshold of genome-wide significance was set at P<5.0x10-8.Within each study, we used linear and logistic regression to model cIMT and an additive genetic model (SNP dosage) adjusted for covariates. Summary estimates from each study were combined using an inverse variance weighted meta-analysis. Additional filters were applied during meta-analyses including imputation quality (MACH r2 <0.3 and IMPUTE info <0.4), a minor allele frequency (MAF) < 0.01, and SNPs that were not present in at least four studies. The genome wide significance threshold was considered at p<5.0 x 10-8.Within each study, we used linear and logistic regression to model carotid plaque and an additive genetic model (SNP dosage) adjusted for covariates. Summary estimates from each study were combined using an inverse variance weighted meta-analysis. Additional filters were applied during meta-analyses including imputation quality (MACH r2 <0.3 and IMPUTE info <0.4), a minor allele frequency (MAF) < 0.01, and SNPs that were not present in at least four studies. The genome wide significance threshold was considered at p<5.0 x 10-8.In each study of the discovery GWAS, natural log-transformed FGF23 was analyzed using linear regression with covariates. Genomic control parameters were estimated for each cohort, and genomic control correction was applied to input statistics before performing meta-analysis to correct for residual cryptic relatedness or population stratification. A fixed-effects inverse-variance weighted meta-analysis using METAL was performed across cohorts on the beta coefficient/SEM from each cohort.Second, random-effects DerSimonian and Laird models were performed. Genetic differentiation was estimated using the Weir unbiased estimator of the fixation index, calculated using the variance in allele frequencies among European and African ancestry samples from the 1000 Genomes and standardized according to the mean allele frequency in the combined sample. Individual centers performed GWAS analyses using linear regression of natural log?transformed PTH concentrations as the dependent variable, and genotypes (SNPallelic dosage) as predictors, under an additive genetic model. Genomic control parameters were estimated for each cohort and appropriate genomic control correction was applied to input statistics before performing meta-analysis to correct for residual cryptic relatedness or population stratification. Genetic differentiation was estimated using the Weir unbiased estimator of the fixation index, calculated using the variance in allele frequencies among European and African samples from the 1000 Genomes and standardized according to themeanallele frequency in the combined sample. Fixed-effects inverse variance?weighted meta-analysis was conducted within each cohort. Study-specific effect estimates and SEMs were combined using METAL. Analyses of the X chromosome were carried out separately in men and women, and the studies were meta-analyzed separately by sex using an inverse-variance model with fixed effects. The sex-specific meta-analysis results were then combined using a sample-size weighted model.All participating cohorts implemented a pre-specified study plag. Within each study, sex-specific and age-adjusted residuals of ABI were created from linear regression models and used as phenotypes in the analysis. No transformation of the ABI measure was performed before analysis. Each SNP was tested for association with ABI in additive genetic models using linear regression, with family-based studies using linear mixed effects or GEE models to account for familial correlations. A genome-wide meta-analysis using a fixed effects approach with inverse variance weighting was then conducted in METAL, on 2.7M SNPs (genotyped and HapMap-imputed). SNP associations were considered to be significant on a genome-wide level at P < 5x10-8, with suggestive evidence of associations at P< 1x10-5. SNPs were excluded from association results if the total meta-analysis sample was less than 20 000 and if the average minor allele frequency of the SNP was less than 5%. Replication analysis was conducted on independent SNP with significant (1 region) and suggestive (5 regions) evidence of association with ABI. All participating cohorts implemented a pre-specified study plag. Within each study, sex-specific and age-adjusted residuals of ABI were created from linear regression models and used as phenotypes in the analysis. No transformation of the ABI measure was performed before analysis. Each SNP was tested for association with ABI in additive genetic models using linear regression, with family-based studies using linear mixed effects or GEE models to account for familial correlations. A genome-wide meta-analysis using a fixed effects approach with inverse variance weighting was then conducted in METAL, on 2.7M SNPs (genotyped and HapMap-imputed). SNP associations were considered to be significant on a genome-wide level at P<5x10-8, with suggestive evidence of associations at P< 1x10-5. SNPs were excluded from association results if the total meta-analysis sample was less than 20,000 and if the average minor allele frequency of the SNP was less than 5%. Replication analysis was conducted on independent SNP with significant (1 region) and suggestive (5 regions) evidence of association with ABI. All participating cohorts implemented a pre-specified study plag. Logistic regression adjusting for age and sex was used to test each SNP for association with the PAD phenotype. Family-based studies used models to account for familial correlations. A genome-wide meta-analysis using a fixed effects approach with inverse variance weighting was then conducted in METAL, on 2.7M SNPs (genotyped and HapMap-imputed). SNP associations were considered to be significant on a genome-wide level at P<5x10-8, with suggestive evidence of associations at P< 1x10-5. SNPs were excluded from association results if the total meta-analysis sample was less than 20,000 and if the average minor allele frequency of the SNP was less than 5%. All participating cohorts implemented a pre-specified study plag. Logistic regression adjusting for age and sex was used to test each SNP for association with the PAD phenotype. Family-based studies used models to account for familial correlations. A genome-wide meta-analysis using a fixed effects approach with inverse variance weighting was then conducted in METAL, on 2.7M SNPs (genotyped and HapMap-imputed). SNP associations were considered to be significant on a genome-wide level at P<5x10-8, with suggestive evidence of associations at P< 1x10-5. SNPs were excluded from association results if the total meta-analysis sample was less than 20,000 and if the average minor allele frequency of the SNP was less than 5%. All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G+B10*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G+B10*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G+B10*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G+B10*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G+B10*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*??+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*??+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*??+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*??+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*??+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is LDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is HDL, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 MbAll participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the SNPxPA interaction term alone using the METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.All participating studies used the following model to test for interaction: E[Y] = Β0 + (ΒE)*PA + (ΒG)*G+ (ΒINT)*G*PA + (ΒC)*C, where E[Y] is TG, PA is the Physical Activity variable (0=active, 1=inactive), G is the dosage of the imputed SNP coded additively from 0 to 2, C the vector of covariates with ΒG, ΒINT and ΒC the estimated effects. Using these estimates, inverse variance-weighted meta-analysis was performed for the 2 df joint meta-analyses of the SNP effect and SNPxPA interaction combined (Manning AK, et al. Genet Epidemiol 35, 11-18, 2011), using METAL meta-analysis software. Genomic control correction was applied twice in Stage 1, first for study-specific GWAS results and again for meta-analysis results, whereas genomic control correction was not applied to the Stage 2 results as interaction testing was only performed at select variants. All variants that reached 2-sided P<1x10-6 in any Stage 1 analyis (all or ancestry-specific) were included in Stage 2. In Stage 2, a variant that reached two-sided P<5x10-8 in the meta-analysis for the interaction term alone, either in the ancestry-specific analyses or in all ancestries combined, as genome-wide significant. The loci were defined as independent if the distance between the lead variants was >1 Mb.In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately.Association of the SNPs was analyzed using linear regression for FT4. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for FT4 were initially sex-stratified and meta-analyzed as a second step and adjusted for covariates. The family-based cohorts (GARP, SardiNIA, MICROS, TwinsUK, LLS) conducted additional analyses on the men and women combined sample, with additional adjustment for sex, to properly account for their family relatedness. All meta-analyses were carried out in duplicate by three independent analysts using a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs was analyzed using linear regression for FT4. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for FT4 in men were adjusted for covariates. The family-based cohorts (GARP, SardiNIA, MICROS, TwinsUK, LLS) conducted additional analyses on the male sample, accounting for family relatedness. Meta-analyses were carried out using a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs was analyzed in women using linear regression for FT4. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for FT4 in women were meta-analyzed and adjusted for covariates. The family-based cohorts (GARP, SardiNIA, MICROS, TwinsUK, LLS) conducted additional analyses on the women, accounting for their family relatedness. All meta-analyses were carried out using a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs was analyzed using linear regression for TSH. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for TSH were initially sex-stratified and meta-analyzed as a second step and adjusted for covariates. The family-based cohorts (GARP, SardiNIA, Val Borbera, MICROS, TwinsUK, LLS and FHS) conducted additional analyses on the men and women combined sample, with additional adjustment for sex, to properly account for their family relatedness. All meta-analyses were carried out in duplicate by three independent analysts using a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs in males was analyzed using linear regression for TSH. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for TSH were meta-analyzed and adjusted for covariates. The family-based cohorts (GARP, SardiNIA, Val Borbera, MICROS, TwinsUK, LLS and FHS) conducted additional analyses on the men with additional adjustment to properly account for family relatedness. All meta-analyses used a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs in women was analyzed using linear regression for TSH. The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The analyses for TSH in women were meta-analyzed and adjusted for covariates. The family-based cohorts (GARP, SardiNIA, Val Borbera, MICROS, TwinsUK, LLS and FHS) conducted analyses on women , with additional adjustment to properly account for their family relatedness. All meta-analyses used a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs was analyzed using logistic regression for hyperthyroidism (decreased TSH). The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The family-based cohorts (GARP, SardiNIA, Val Borbera, MICROS, TwinsUK, LLS and FHS) conducted additional analyses to properly account for their family relatedness. All meta-analyses used a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. In each study of the discovery GWAS, genotyping was performed on genome-wide arrays. Genome-wide data were imputed to the 1000 Genomes, phase 1 version 3 (March 2012) ALL populations reference panel, including the X chromosome. Quality control before imputation was applied in each study separately. Association of the SNPs was analyzed using logistic regression for hypothyroidism (increased TSH). The genotype-phenotype association was conducted using an additive genetic model on SNP dosages. The family-based cohorts (GARP, SardiNIA, Val Borbera, MICROS, TwinsUK, LLS and FHS) conducted additional analyses to properly account for their family relatedness. All meta-analyses used a fixed-effect meta-analysis applying inverse variance weighting as implemented in METAL. SNPs with minor allele frequency ≤0.005 or an imputation quality score ≤0.4 were excluded prior to the meta-analyses. Study-specific results were corrected by their specific ΛGC if >1. SNPs that were present in <75% of the total sample size contributing to the respective meta-analysis (separately for autosomal and X-chromosomal SNPs) or with a MAF ≤0.01 were excluded from subsequent analyses. Each GWAS was imputed up to Phase 1 integrated 1000 Genomes Project reference panel and each SNP that passed QC was tested for association with eGFR adjusted for covariates. eGFR summary statistics were aggregated across studies through trans-ethnic meta-analysis. Stouffer's method (implemented in METAL) was used because allelic effect sizes were reported on different scales in each of the three sources contributing to the meta-analysis. Genome-wide significant evidence of association with eGFR was considered for p < 5 x 10-8. The observed scale heritability of eGFR was estimated by LD Score regression. Three sources were used - 19 studies of diverse ancestry (COGENT-Kidney Consortium), published meta-analysis of 33 European ancestry studies (CKDGen Consortium), and a published East Asian ancestry study from the Biobank Japan project. The trans-ethnic meta-analysis effect and standard error were obtained from the COGENT-Kidney Consortium (up to 81,829 subjects) under a fixed-effects model with inverse variance weighting. Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction termDiscovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking).Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction termDiscovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 8 cohorts. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 13 cohorts of African ancestry and up to 23,748 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 9 cohorts of Asian ancestry and up to 13,171 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 23 cohorts of European ancestry and up to 90,272 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction termDiscovery genome-wide interaction analyses included 3 cohorts of Hispanic ancestry and up to 6,705 individuals. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 2 degree of freedom joint test of main and interaction effects (2df) was employed. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.Discovery genome-wide interaction analyses included 29 cohorts contributing data from 51 study/ancestry groups and up to 133,805 individuals of EUR, AFR, ASN, and HISP ancestry. All cohorts ran three models in all individuals: a main effect model, a model adjusted for smoking, and an interaction model that included a multiplicative interaction term between the variant and smoking status. The main effect model was run stratified by smoking exposure. All models were run for 3 lipids traits (HDL, LDL, and triglycerides) and 2 smoking exposures (current smoking and ever smoking). Studies including participants from multiple ancestry groups conducted and reported analyses separately by ancestry. Participating studies provided the estimated genetic main effect and robust estimates of standard error for all requested models. In addition, for the models with an interaction term, studies also reported the interaction effects and robust estimates of their standard errors, and a robust estimate of the corresponding covariance matrix between the main and interaction effects. To obtain robust estimates of covariance matrices and robust standard errors, studies with only unrelated participants used R packages; either sandwich or ProbABEL. If the study included related individuals, either generalized estimating equations (R package geepack) or linear mixed models (GenABEL, MMAP, or R) were used. Meta-analyses were conducted for all models using the inverse variance-weighted fixed effects method as implemented in METAL. A 1df Wald test was used to evaluate the 1df interaction, as well as the main effect and the smoking-adjusted main effect in models without an interaction term. Variants that were associated in any analysis at p ≤ 10-6 were carried forward for analysis in Stage 2. Variants selected for Stage 2 were evaluated in 50 cohorts. In addition to the 4 Stage 1 ancestry groups, Stage 2 analyses also included studies of Brazilian (BR) individuals that were only included in the trans-ancestry meta-analyses. In Stage 2, variants were evaluated only in a model with the interaction term.To assess the association between genotype and QRS duration, individual cohort studies used additive genetic linear regression models, either in a regression model (MESA, WHI, Starr County) or a mixed model (HCHS/SOL, to account for relatedness and shared environment between individuals). Genotypes and imputed genotypes from 1000 Genomes Project had individual cohort QC filters applied, which excluded SNPs with low imputation quality (<0.30) or small effective sample sizes for each individual SNP (effN <30), with effN = 2 x MAF x (MAF-1) x N x Imputation Quality; where N is the number of participants. Results were combined using fixed-effects inverse variance meta-analysis using METAL software with genomic control for summary statistics to reduce test-statistic inflation. Study heterogeneity was evaluated using the Cochran Q test. Results were considered genome-wide significant for P-values < 5x10-8. Secondary signals were identified using iterative rounds of conditional analysis, with adjustment for additional Hispanic/Latino index SNPs in the model, until there were no SNPs found to be genome-wide significant.Natural log-transformed FVII was analyzed within each cohort. Participants with values 3 standard deviations (SDs) above or below the population mean were removed prior to cohort-level analysis and any individuals on anticoagulant therapy were also excluded. Ancestry-stratified, study-specific regression analyses using an additive genetic model were performed between genome-wide 1000G imputed variant dosages and phenotype levels, adjusted for covariates. X chromosome analyses were stratified by sex, with variants in males coded as 0/2. The discovery trans-ancestry meta-analysis was conducted in 2 steps. First, METAL software was used to perform ancestry-specific inverse-variance weighted meta analysis. The same method was used to meta-analyze the ancestry-specific results. A genomewide significance threshold of P< 2.5 x 10-8 was employed. A locus was defined as 61 Mb from the variant with the lowest P value. The primary trans-ancestry meta-analysis included the 7 studies that measured FVII activity, and not the 2 studies (FHS and PROCARDIS) that measured FVII antigen.Natural log-transformed FVII was analyzed within each cohort. Participants with values 3 standard deviations (SDs) above or below the population mean were removed prior to cohort-level analysis and any individuals on anticoagulant therapy were also excluded. Ancestry-stratified, study-specific regression analyses using an additive genetic model were performed between genome-wide 1000G imputed variant dosages and phenotype levels, adjusted for covariates. X chromosome analyses were stratified by sex, with variants in males coded as 0/2. The discovery trans-ancestry meta-analysis was conducted in 2 steps. First, METAL software was used to perform ancestry-specific inverse-variance weighted meta analysis. The same method was used to meta-analyze the ancestry-specific results. A genomewide significance threshold of P< 2.5 x 10-8 was employed. A locus was defined as 61 Mb from the variant with the lowest P value. The primary trans-ancestry meta-analysis included the 7 studies that measured FVII activity, and not the 2 studies (FHS and PROCARDIS) that measured FVII antigen.Within each study, Cox proportional hazards regression models were used to test the association between each SNP and time to incident HF, while adjusting for covariates. Time of entry to the analysis was the date of cohort entry and failure time was the time of HF diagnosis. Participants without incident HF were censored at the time of death, last date of contact, or at the end of follow-up, whichever came first. For each SNP, additive genetic models were used to estimate the regression coefficient for the hazard ratio (HR) for allele dosage and its respective standard error. For each analysis, a genomic control coefficient was calculated, which estimated the extent of underlying population structure. Meta-analyses were performed using MetABEL software. Fixed-effect meta-analyses combined regression coefficients and standard errors across the 2 cohorts for each SNP to produce an overall beta coefficient, standard error, and P-value. SNPs were excluded for which the post-meta-analysis population-size-weighted MAF<0.015. The a priori threshold of genome-wide significance was set at P < 5.0 x10-7. When more than 1 SNP clustered at a locus, we picked the SNP with the smallest P value as the locus marker.Within each study, Cox proportional hazards regression models were used to test the association between each SNP and time to incident HF, while adjusting for covariates. Time of entry to the analysis was the date of cohort entry (ARIC, CHS, RS) or DNA collection (FHS), and failure time was the time of HF diagnosis. Participants without incident HF were censored at the time of death, last date of contact, or at the end of follow-up, whichever came first. For each SNP, additive genetic models were used to estimate the regression coefficient for the hazard ratio (HR) for allele dosage and its respective standard error. In ARIC and CHS, analyses were conducted separately in 2 ancestry groups, European and African. For each analysis, a genomic control coefficient was calculated, which estimated the extent of underlying population structure. Meta-analyses were performed using MetABEL software. For the populations of European ancestry, fixed-effect meta-analyses combined regression coefficients and standard errors across the 4 cohorts for each SNP to produce an overall beta coefficient, standard error, and P-value. SNPs were excluded for which the post-meta-analysis population-size-weighted MAF<0.015. The a priori threshold of genome-wide significance was set at P < 5.0 x10-7. When more than 1 SNP clustered at a locus, we picked the SNP with the smallest P value as the locus marker.The association between genomic variation and time to death among individuals with incident HF was evaluated separately in each cohort by Cox proportional hazards models. The follow-up time interval was defined as the time between the date of incident HF and the date of death, last contact if lost to follow-up, or the end of follow-up, whichever came first. Each model was evaluated separately in individuals with HF of European or African ancestry and included adjustment for covariates at the time of the HF event. All SNPs were evaluated for association under an additive genetic model. Across the 4 study populations of European ancestry and separately the 2 populations of African ancestry, fixed-effect meta-analyses combined b-coefficients and study-specific P-values using genomic control lambdas to adjust for remaining population substructure. Meta-analyses were performed using MetABEL and METAL software for confirmation. Genome-wide significance was set at P < 5 x 10-8. The association between genomic variation and time to death among individuals with incident HF was evaluated separately in each cohort by Cox proportional hazards models. The follow-up time interval was defined as the time between the date of incident HF and the date of death, last contact if lost to follow-up, or the end of follow-up, whichever came first. Each model was evaluated separately in individuals with HF of European or African ancestry and included adjustment for covariates at the time of the HF event. All SNPs were evaluated for association under an additive genetic model. Across the 4 study populations of European ancestry and separately the 2 populations of African ancestry, fixed-effect meta-analyses combined b-coefficients and study-specific P-values using genomic control lambdas to adjust for remaining population substructure. Meta-analyses were performed using MetABEL and METAL software for confirmation. Genome-wide significance was set at P < 5 x 10-8. Genotyping was performed using commercially available assays for genome-wide SNP detection. Imputation of non-genotyped SNPs was performed using CEU reference panels of SNP correlations from the HapMap project phase II. All-cause mortality following initial HF diagnosis was examined for association with additive allele dosage of each genotyped or imputed SNP using Cox proportional hazards models, with censoring at the end of or loss to follow-up. In the family-based FHS, Cox models were implemented with clustering on pedigree to account for relatedness. Genomic control was applied to results from each cohort. Cohort-specific GWAS results were combined using fixed effects meta-analysis with inverse variance weights. SNPs passing a P-value threshold defined a priori as P < 5.0x10-7 in the genome-wide stage 1 were carried forward to the second stage with targeted genotyping in 1,870 HF patients from four independent cohorts.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All traits were analyzed as continuous traits, with the exception of LVSD, DDpEF, and HFpEF. LVSD was defined as an EF <50%, fractional shortening (FS) <29% or reduced (poor or impaired) EF by visual estimation. Aggregate binary phenotypes were defined for asymptomatic participants with echocardiographic evidence of LV DDpEF and for those with HFpEF based on information on classes of HF according to the New York Heart Association (NYHA) and medication for HF in addition to echocardiography. All participating cohorts implemented a pre-specified study plag. Linear regression adjusting for covariates was used to test each SNP for association with the phenotype. Continuous echocardiographic traits were related to genotype dosage (0-2 copies of the effect allele) for each autosomal SNP using linear regression assuming additive genetic models adjusted for age, sex, height, weight, and study site (only applicable for the Cardiovascular Health Study [CHS] and Anglo-Scandinavian Cardiac Outcomes Trial [ASCOT]). For binary traits, we used logistic regression models with the same adjustments. In the Framingham Heart Study (FHS), linear mixed-effects models were applied to account for familial correlations. The associations of genotypes with echocardiographic traits were quantified by beta estimates, SEM, and P values. After verifying strand alignment across studies and applying genomic control to each study, inverse variance-weighted fixed-effects meta-analysis was applied across the discovery cohorts with METAL (structural and the systolic function traits) or the R package MetABEL (diastolic traits). After meta-analysis, SNPs were excluded with MAF < 0.5% for diastolic function traits and MAF < 1% for structural traits, and FS and MAF < 3% for LVSD. Genome-wide significance was set at P < 5 x 10-8.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. A GWAS analysis on the MRI-derived volumes of subcortical structures used the 1000 Genomes Project reference panel (phase 1; version 3) for imputation of missing variants in CHARGE and ENIGMA. UK Biobank performed imputation of variants using the Haplotype Reference Consortium (HRC) reference panel. The sample comprised up to 37,741 individuals of European ancestry from 48 study samples across CHARGE, ENIGMA and UK Biobank. Each study examined the association between genetic variants with a MAF≥1% and the volumes of subcortical structures (average volume for bilateral structures) using additive genetic models adjusted for covariates. After quality control, meta-analyses combined all samples using sample-size-weighted fixed-effects methods in METAL.All participating cohorts implemented a pre-specified study plan. Association analyses of allele dosage of imputed SNPs with VTE were performed separately in each study by using logistic or Cox proportional regression analyses adjusted for study-specific covariates. All SNPs with acceptable imputation quality (r2 >0.3, or r2 > 0.7 for eMERGE) in all discovery studies and minor allele count (MAC) greater than 20 in either cases or controls were entered into a meta-analysis. For the meta-analysis, a fixed-effects model based on the inverse-variance weighting was employed as implemented in METAL software. A statistical threshold of P < 5 x 10-8 was applied to declare genome-wide significance. The data are in the public ftp site.
]]>To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme. To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme. Those with stroke or dementia were excluded from analysis.To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme.To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme. Those with dementia or stroke were excluded from analysis. To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme.To assess the association between genotype and BMB, individual studies used additive genetic logistic regression models, adjusted for covariates. Genotypes and imputed genotypes were filtered by imputation quality (INFO or r2> 0.5, MAF> 0.005, MAF*Ncases*imputation quality> 5). CHARGE consortium results were meta-analyzed with UKB results suing an inverse variance-weighted fixed-effects model (using METAL) with the standard error analysis scheme. Those with stroke or dementia were excluded from analysis.The response to statin treatment was defined as the difference between the natural log-transformed on- and off-treatment LDL-C levels. The beta of the regression reflects the fraction of differential LDL lowering in carriers versus non-carriers of the SNP. For each individual, at least one off-treatment LDL-C measurement and at least one on-treatment LDL-C measurement were required. When multiple on- or off-treatment measurements were available, the mean of LDL-C cholesterol was used. Each study independently performed the GWAS on the difference between natural log-transformed on- and off-treatment LDL-C levels. An additive genetic model was assumed and tested using a linear regression model. For imputed SNPs, regression analysis was performed onto expected allele dosage. Analyses in the observational studies were, if available, additionally adjusted for the statin dose by the natural logarithm of the dose equivalent. Within each study, SNPs with MAF<1% or imputation quality<0.3 were excluded from the analysis. The software package METAL was used for performing the meta-analysis using a fixed effects, inverse variance weighted approach. To correct for possible population stratification, genomic control was performed by adjusting the within-study findings and the meta-analysis results for the genomic inflation factor.To assess the impact of statin therapy on the association between genotype and incident MI, the GIST consortium included both randomized clinical trials (RCTs) and observational studies in two analyses of GWAS data. For imputed SNPs, regression analyses were performed on allelic dosage. For RCTs, an additive genetic model was assumed and a Cox proportional hazards model included an interaction term of statin use * SNP dosage on incident MI. For observational studies, a case-only design with binary variable for statin use and SNP dosage under binary logistic regression model. SNPs with minor allele frequency (MAF) < 1% or imputation R2 <0.3 were excluded. The METAL software package was used for fixed effects inverse variance weighted meta-analysis. The interaction beta (from RCTs) and the case-only beta from observational studies were meta-analyzed. All SNPs with P<5x10-4 in stage 1 (8 studies, n=10,769; 4,212 cases) were selected for follow-up by meta-analysis of stage 1 and stage 2 results, with significance of P<5x10-8. Gene-based analysis: variants were aggregated by GENCODE genes (v24). Variants within a gene were filtered to retain a set of rare variants (minor allele frequency [MAF]<1%) that were predicted as loss-of-function variants (LoF), protein altering small deletions/insertions (indels) or synonymous SNVs that have a deleterious functional annotation (FATMM-MKL score>0.5 or MetaSVM score > 0 for missense SNVs). Variants in a 5 kb window promoter region (upstream of transcription start site and in a FANTOM5 [Functional ANalyses Through Hidden Markov models] peak) and variants at the first intron of genes were also included. Genes with at least 10 individuals with at least one copy of any alternative allele were included. Burden and SKAT tests were performed and significance was p < 1.6x 10-6 based on a Bonferroni correction for two tests on 16,054 genes . To identify the contribution of one or more variants within genes with a gene-based significant association, the association of each single variant within the aggregate gene unit was tested, with a leave-one-variant-out analysis with variants aggregated within a gene for gene-based tests.
]]>High-sensitivity cardiac troponin I (cTnI) levels were inverse-normalized for analyses. In each ARIC, linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. Variants wtih MAF<1% were excluded after meta-analyses. Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.High-sensitivity cardiac troponin I (cTnI) levels were inverse-normalized for analyses. In each study and ancestry, linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. The METAL software package was used for fixed effects inverse variance weighted meta-analysis for variants included in at least two studies. Variants wtih MAF<1% were excluded after meta-analyses. Cross-study heterogeneity was assessed using Cochran's Q-test (excluding those variants with p<0.05). Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.High-sensitivity cardiac troponin I (cTnTI) levels were inverse-normalized for analyses. In each study and ancestry, linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. The METAL software package was used for fixed effects inverse variance weighted meta-analysis for variants included in at least two studies. Variants wtih MAF<1% were excluded after meta-analyses. Cross-study heterogeneity was assessed using Cochran's Q-test (excluding those variants with p<0.05). Variants with High-sensitivity cardiac troponin T (cTnT) levels were inverse-normalized for analyses; linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. Variants wtih MAF<1% were excluded. Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.High-sensitivity cardiac troponin T (cTnT) levels were inverse-normalized for analyses. In each study and ancestry, linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. The METAL software package was used for fixed effects inverse variance weighted meta-analysis for variants included in at least two studies. Variants wtih MAF<1% were excluded after meta-analyses. Cross-study heterogeneity was assessed using Cochran's Q-test (excluding those variants with p<0.05). Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.High-sensitivity cardiac troponin T (cTnT) levels were inverse-normalized for analyses. In each study and ancestry, linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. The METAL software package was used for fixed effects inverse variance weighted meta-analysis for variants included in at least two studies. Variants wtih MAF<1% were excluded after meta-analyses. Cross-study heterogeneity was assessed using Cochran's Q-test (excluding those variants with p<0.05). Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.High-sensitivity cardiac troponin T (cTnT) levels were inverse-normalized for analyses; linear regression models were used to test the association between genetic variants and inverse-normalized troponin levels. Variants wtih MAF<1% were excluded after meta-analyses. Variants with association p<5x10-8 were considered genome-wide significant; variants with p<1x10-5 were conisdered as “suggestive” of association. Gene-based association analyses used MAGMA v 1.07, implemented in FUMA.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype (with imputation) and each fatty acid within self-reported race/ancestry, individual cohort studies used additive genetic linear regression models. Log-transformation was applied for ALA, EPA and GLA, and all fatty acid results had outliers winsorized to the value (median +/- 3.5*MAD'). Genotypes and imputed genotypes from 1000 Genomes Project were filtered using EasyQC (minor allele count >6 and imputation R-squared >0.3). Results were combined using weighted sum of Z-scores with METAL software and filtered by effHET >60. Results were considered genome-wide significant for P-values < 5x10-8.To assess the association between genotype and PVS, each study conducted logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. To assess the association between genotype and PVS, each study conducted an ancestry-specific logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. Additional GWAS meta analyses were conducted on EUR-only participants.To assess the association between genotype and PVS, each study conducted a logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. To assess the association between genotype and PVS, each study conducted an ancestry-specific logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. Additional GWAS meta analyses were conducted on EUR-only participants.To assess the association between genotype and PVS, each study conducted a logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. To assess the association between genotype and PVS, each study conducted an ancestry-specific logistic regression analysis under an additive genetic model adjusting for covariates. In each study, rare variants, variants with low imputation accuracy, and those with extensive effect size (beta>5 or beta<-5) were removed. A sample size-weighted GWAS meta-analysis used METAL software, with variants having an effective allele count <10 as well as those with significant heterogeneity (Phet<5.0x10-8) removed. LD clumping permitted the retention of most significant SNPs with removal of those within 1 Mb and r2>0.1; cross-ancestry meta-analysis used MR-MEGA software. Additional GWAS meta analyses were conducted on EUR-only participants.Each study independently analyzed its genotype-phenotype data. Studies performed logistic regression or generalized mixed models for case-control studies and Cox regression for cohort studies with study-specific covariates. For the X chromosome, sex-stratified analyses were performed, followed by male-female meta-analyses (using 0/2 allele coding for males and 0/1/2 for females). All meta-analyses were conducted with METAL software using a fixed-effects inverse-variance weighted model. All variants were included as defined by each study and, for discovery, INVENT-2019 (three studies)and MVP were meta-analyzed. For replication, five GWAS studies were included: BioME, MESA, MGB, Upenn, and UKBB. Discovery genome-wide significance was set at P<5.0x10-8, and variants were retained if a concordant effect direction was observed in 2 or more studies and grouped in the same locus if within 1Mb. Each study independently analyzed its genotype-phenotype data. Studies performed logistic regression or generalized mixed models for case-control studies and Cox regression for cohort studies with study-specific covariates. For the X chromosome, sex-stratified analyses were performed, followed by male-female meta-analyses (using 0/2 allele coding for males and 0/1/2 for females). Meta-analysis was conducted with METAL software using a fixed-effects inverse-variance weighted model. All variants were included as defined by each study and, for discovery, four consortia/studies were grouped and meta-analyzed - INVENT-2019, the Estonian BioBank, FinnGen, and MVP. For replication, European-ancestry data from nine studies were included: BioME, FARIVE, GAIT2, MARTHA12, MESA, MGB, RETROVE, Upenn, and UKBB. Combined European-ancestry GWAS meta-analysis used variants with MAF >= 0.01 with significance claimed with P < 5.0x10-8.Each study independently analyzed its genotype-phenotype data. Studies performed logistic regression or generalized mixed models for case-control studies and Cox regression for cohort studies with study-specific covariates. For the X chromosome, sex-stratified analyses were performed, followed by male-female meta-analyses (using 0/2 allele coding for males and 0/1/2 for females). All meta-analyses were conducted with METAL software using a fixed-effects inverse-variance weighted model. Genome-wide significance was set at P<5.0x10-8.Each study independently analyzed its genotype-phenotype data. Studies performed logistic regression or generalized mixed models for case-control studies and Cox regression for cohort studies with study-specific covariates. For the X chromosome, sex-stratified analyses were performed, followed by male-female meta-analyses (using 0/2 allele coding for males and 0/1/2 for females). All meta-analyses were conducted with METAL software using a fixed-effects inverse-variance weighted model. All variants were included as defined by each study and, for discovery, four consortia/studies were grouped and meta-analyzed - INVENT-2019, the Estonian BioBank, FinnGen, and MVP. For replication, ten GWAS studies were included: BBJ, BioME, FARIVE, GAIT2, MARTHA12, MESA, MGB, RETROVE, Upenn, and UKBB. Discovery genome-wide significance was set at P<5.0x10-8, and variants were retained if a concordant effect direction was observed in 2 or more studies and grouped in the same locus if within 1Mb. Combined cross-ancestry GWAS meta-analysis used variants with MAF >= 0.01 with significance claimed with P < 5.0x10-8.Each study independently analyzed its genotype-phenotype data. Studies performed logistic regression or generalized mixed models for case-control studies and Cox regression for cohort studies with study-specific covariates. For the X chromosome, sex-stratified analyses were performed, followed by male-female meta-analyses (using 0/2 allele coding for males and 0/1/2 for females). All meta-analyses were conducted with METAL software using a fixed-effects inverse-variance weighted model. All variants were included as defined by each study and, for discovery, four consortia/studies were grouped and meta-analyzed - INVENT-2019, the Estonian BioBank, FinnGen, and MVP. Discovery genome-wide significance was set at P<5.0x10-8, and variants were retained if a concordant effect direction was observed in 2 or more studies and grouped in the same locus if within 1Mb.To be included in a CHARGE study, a cohort has to provide summary statistics for meta-analyses according to the specifications of the CHARGE Working Group. The summary statistics are based upon an agreed-upon analytic plan that provides guidance for all protocols and software.]]>
A major goal of the CHARGE Consortium is to collaboratively produce a series of jointly coordinated, high-impact publications that describes the collaborative results of genome-wide association scans (GWAS) for a number of cardiovascular, lung, blood and aging phenotypes. The focal point of the design, analysis, interpretation and publication of research results for each specific phenotype is one of several collaborative phenotype Working Groups (WGs) that have been convened by the CHARGE Research Steering Committee (RSC). One analytic strategy that is being undertaken by many of the CHARGE WGs is to conduct a meta-analysis of GWAS results for single phenotypes common to multiple CHARGE cohorts. The actions of members of CHARGE WGs from each of the individual CHARGE cohorts and other collaborating cohorts should support this research goal. The principles are meant to encourage collaborating investigators to conduct their research in the spirit of collaboration and trust and with transparency regarding potentially sensitive issues about the use of shared data prior to publication.]]>