We had BMI and GWAS data for 20,283 individuals (n = 6,762 for HumanHap, n = 5,844 for OmniExpress, n = 7,677 for AffyMetrix) within NHS, NHSII, HPFS and PHS. Platform-specific QQ-plots (S4A–S4C Fig) showed no indication of systematic bias (genomic inflation factor λ = 1.00–1.02). The results from the meta-analysis are shown in Figs 3 and 4. We observed a tail of strongly associated SNPs with the top SNPs located in the known BMI FTO locus (strongest associated SNP: rs55872725, β = 0.45, p = 3.48x10-22). Given that the FTO locus has also been associated with Type 2 Diabetes, we reran the analysis excluding all Type 2 Diabetes cases (n = 2,540), The association for the FTO SNP rs55872725 remained strongly significant (β = 0.41, p = 4.25x10-18). We also observed genome-wide significant associations for the previously identified TMEM18 (strongest associated SNP: rs7563362, β = -0.36, p = 1.76x10-8) and FANCL loci (strongest associated SNP: rs980183, β = -0.26, p = 2.73x10-8). None of the SNPs that were originally reported were the top SNP in our data. However, for
