show the same genetic effect. Inclusion of proxy SNPs that are in high LD is still an open issue: some authors recommend this practice as further evidence of a real association [98] while others recommend against it because it is unnecessary [46]. Replication of findings in a population with different genetic backgrounds can strengthen the evidence of true associations and identify variants that are robust to different genetic background and environmental exposures [43]. However, failure to reproduce an association in a genetically different population should not be taken as evidence of a false positive. Also the requirement of a larger size sample for the replication study when compared with the primary study is arguable. The large sample size of the primary population is necessary to achieve sufficient power with genome-wide significance levels. However, replication usually is limited to a small selection of SNPs in which case there is no multiple testing problem and traditional significance levels should be acceptable.
