Premium
Impact of repeated measures and sample selection on genome‐wide association studies of fasting glucose
Author(s) -
RasmussenTorvik Laura J.,
Alonso Alvaro,
Li Man,
Kao Wen,
Köttgen Anna,
Yan Yu,
Couper David,
Boerwinkle Eric,
Bielinski Suzette J.,
Pankow James S.
Publication year - 2010
Publication title -
genetic epidemiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.301
H-Index - 98
eISSN - 1098-2272
pISSN - 0741-0395
DOI - 10.1002/gepi.20525
Subject(s) - genome wide association study , single nucleotide polymorphism , diabetes mellitus , genotyping , medicine , genetic association , snp , impaired fasting glucose , type 2 diabetes , biology , genotype , genetics , endocrinology , impaired glucose tolerance , gene
Abstract Although genome‐wide association studies (GWAS) have been performed in longitudinal studies, most used only a single trait measure. GWAS of fasting glucose have generally included only normoglycemic individuals. We examined the impact of both repeated measures and sample selection on GWAS in Atherosclerosis Risk In Communities (ARIC), a study which obtained four longitudinal measures of fasting glucose and included both individuals with and without prevalent diabetes. The sample included Caucasians and the Affymetrix 6.0 chip was used for genotyping. Sample sizes for GWAS analyses ranged from 8,372 (first study visit) to 5,782 (average fasting glucose). Candidate SNP analyses with SNPs identified through fasting glucose or diabetes GWAS were conducted in 9,133 individuals, including 761 with prevalent diabetes. For a constant sample size, smaller P ‐values were obtained for the average measure of fasting glucose compared to values at any single visit, and two additional significant GWAS signals were detected. For four candidate SNPs (rs780094, rs10830963, rs7903146, and rs4607517), the strength of association between genotype and glucose was significantly ( P ‐interaction<0.05) different in those with and without prevalent diabetes, and for all five fasting glucose candidate SNPs (rs780094, rs10830963, rs560887, rs4607517, and rs13266634) the association with measured fasting glucose was more significant in the smaller sample without prevalent diabetes than in the larger combined sample of those with and without diabetes. This analysis demonstrates the potential utility of averaging trait values in GWAS studies and explores the advantage of using only individuals without prevalent diabetes in GWAS of fasting glucose. Genet. Epidemiol . 34: 665‐673, 2010. © 2010 Wiley‐Liss, Inc.