Furthermore, we were unable to replicate our results in the GADD sample, and for rs1044394 the “risk” allele was different. We chose the GADD as a replication sample because it was ascertained in a manner similar to the CADD with recruitment being through drug abuse treatment centers. Overall, the samples were phenotypically similar (see Table 1). However, the entire CADD sample was ascertained in the Denver metro area, while the GADD sample was recruited in both Denver and San Diego. Furthermore, there were fewer informative families in the GADD analysis compared to the CADD, thus there was less power to detect an association in this sample (see Table 5). In the GADD non-Hispanic EA sample there were only 14 – 43 informative families per SNP (compared to 32 - 116 for the same SNPs in CADD). This is quite a small number of informative families and would only provide limited power to detect an association. Given this limitation, future studies that have greater power to detect an association are warranted.
