GWAS was conducted using linear and logistic regressions in PLINK,35 for CD symptom counts and case status, respectively. Genotypes for each of the 917 694 autosomal SNPs were coded log additively (0, 1, 2 copies of the minor allele). Covariates representing sex, age (defined using quartiles as three dummy measures representing younger than 34 years, 35–39 years, 40–44 years, with older than 45 years as the reference group, as in other SAGE publications27), study site, and the two aforementioned principal component scores were included as covariates.