the patient record and was excluded from the analysis. Cryptic relatedness was estimated by pairwise identity-by-descent (IBD) analysis implemented in PLINK (http://pngu.mgh.harvard.edu/purcell/plink/). Two pairs of duplicate samples and 104 cryptic first degree relative samples were identified and only one unrelated sample from each family was included. Inbreeding coefficient F is calculated as the ratio of observed vs. expected number of homozygous genotypes using PLINK. One subject with F value > 4 s.d. suggesting excess homozygosity may be due to population substructure and was removed. The final dataset consisted of 1,715 unrelated samples with BMI data for association analysis.
