All 20 CNVRs with evidence for an association with obesity in both GWAS discovery samples were followed up in 281 independent obesity trios (a subsample of the 365 independent obesity trios analysed at the two tagging SNPs) that were very recently genotyped (Affymetrix Genome-Wide Human SNP Array 6.0). Six of the 20 CNVRs again showed evidence for a directionally consistent association with obesity (one-sided CNV-FBAT P-value ≤ 0.025; Table 2). These six CNVRs involve those two CNVRs that are also supported by the tagging SNP approach. Furthermore, association signals for only these two CNVRs were found at exactly the same CNV markers as in the family-based discovery GWAS sample. Neither in the family-based discovery nor in the array-based replication sample did any of the CNVRs achieve significant CNVR-adjusted P-values (all >0.025; data not shown). However, in a combined analysis of both family-based samples, three CNVRs presented at least nominally significant CNVR-adjusted one-sided P-values (≤0.025; Table 2). None of these CNVR associations remained significant after strict Bonferroni correction, which may potentially be due to an insufficient sample size. Among the three
