Several Genome-Wide Association Studies (GWAS) of LDL-C have been published in populations of European-Ancestry but only one previous GWAS of LDL-C has been published in African Americans. Using phenotype data extracted from repeated lab measurements of LDL-C in electronic medical records (EMR) and the Illumina 1M Duo array we performed a LDL-C GWAS in African Americans in the eMERGE study. Methods: African Americans from eMERGE sites Northwestern University and Vanderbilt University were included in the analysis. All available LDL-C measurements as well as information about statin and hormone prescriptions, cancer, diabetes, and hypo/hyperthyroidism were abstracted from electronic medical records. Two separate samples were created; a sample incorporating exclusions due to medication use and disease diagnoses (n = 618) and a sample without exclusions (n = 1249). After data extraction and cleaning, median LDL-C for each individual was used for both analyses. GWAS analysis of median LDL-C was conducted using PLINK, controlling for age, age2, sex, recruitment site, genotyping batch, and the first three principal components of ancestry. Results: One SNP was significantly (p < 5 X 10-8) associated with LDL-C in both samples; rs7412 in the APOE gene. In the sample excluding measurements taken from those on medication or with prevalent disease, the effect size of this association was large compared to SNPs identified in previous GWAS results for LDL-C (a decrease of 19.9 mg/dl per one additional copy of the minor allele, p = 8.7 x 10 -11). The percent of LDL-C variance accounted for by rs7412 in this sample was 6.7%. The effect size of the rs7412 association was smaller in the sample without exclusions (a decrease of 12.3 mg/dl in LDL-C per allele, p = 1.6 x 10 -9) but still large compared to other LDL-C GWAS findings. Effect sizes for SNPs previously identified in the other African American GWAS of LDL-C were consistent with the earlier report, although there was not sufficient power for these associations to achieve genome-wide significance in this population. Discussion: SNP rs7412 appears to be an important common genetic variant influencing LDL-C levels. Although other signals in APOE have been detected in previous GWAS of European-Ancestry populations and African Americans and this specific SNP has been reported in several previous candidate gene studies, this association has not been identified previously in GWAS likely because rs7412 (or a good proxy) was not included on the genotyping arrays used in these studies. Our results (a) indicate that the association of rs7412 and LDL-C should be examined in other larger African American study populations, (b) further validate the use if EMR-derived traits in genetic research and (c) suggest one should be cautious in concluding that currently published GWAS have discovered all important common genetic variation contributing to certain disease traits.
Return of individual research results from genome-wide association studies: experience of the Electronic Medical Records and Genomics (eMERGE) Network
