observed a nominal association (P = 0.007) with the F11 region. Our BMI and VTE results confirm that GWAS analysis of secondary traits in this data is valid and provides a platform for future studies of secondary traits. We ran the BMI and VTE analyses twice, the first time without removing duplicates between the datasets (total of 444 pairs), and the second time with the duplicates removed. Although the 444 pairs constitute less than 5% of our total sample size, including them had an impact on the genomic inflation factor (for BMI, the genomic inflation factor went from 1.09 to 1.05 and for VTE, the genomic inflation factor went from 1.02 to 1.00). These results are especially interesting as it is often difficult to identify duplicates across studies when raw data from all participating studies are not available. Care should be taken to remove overlapping subjects across GWAS contributing to a meta-analysis, but any remaining cryptic overlap may inflate association statistics. In that case, statistical adjustment procedures like LD score regression [55] can be used to account for cryptic overlap.
