The GIANT meta-analysis identified 180 loci that were associated with height3. When we fitted all 180 hit SNPs simultaneously in a joint analysis, the majority of them seemed to be independently associated, because they had been deliberately selected to be at least 1 Mb away from each other to render them unlikely to be in strong LD. However, there was an exception. Two SNPs, rs1814175 and rs5017948, were reported as independently associated SNPs by the GIANT Consortium3 with P values in the discovery set of 1.9 × 10–8 and 4.6 × 10–8, respectively (Table 3). These SNPs are ~1.76 Mb distant but in substantial LD (r = 0.61 and 0.59 in the ARIC and QIMR cohorts, respectively), suggesting that, in some specific cases, the commonly used 1-Mb window is not big enough to guarantee that two SNPs are independently associated with a trait and that the stepwise conditional analysis is a more general approach to refine association signals and to identify additional associated SNPs. Given either of these SNPs in the model, the other SNP is not found to reach
