The increasing availability of genome-wide genotype data from public databases of individuals from a wide range of disease cohorts offers an exciting opportunity to increase sample sizes utilized in GWA studies. Provided that samples are from the same population and assuming no unmeasured confounders, power can be increased by expanding the control cohort with disease samples from other studies. However, in the presence of unobserved population structure, such an approach can lead to an increase in the false-positive error rate. Here, we demonstrate by simulation that correction for axes of genetic variation from MDS, obtained from IBS metrics calculated from large numbers of SNPs, genome-wide maintains the correct false-positive error rate, even in the presence of genetic differences between samples in the expanded control group of the magnitude we might expect across Europe or even further afield. In the absence of population structure, there is no loss in power compared to a test that does not correct for axes of genetic variation. Furthermore, there is a substantial increase in power over studies that do not make use of external reference
