An important limitation is that although consistent with recent work (Dick et al., 2006), we tested only a single SNP in the GABRA2 gene. Although this SNP was the most strongly associated with AD by Edenburg et al., (2004) and was included in the three SNP haplotypes tested in the adult COGA sample, inclusion of other nearby SNPs or haplotype-based analyses might provide more consistent results; our findings do not preclude significant associations between conduct disorder and other SNPs in the GABRA2 gene, or nearby genes such as GABRG1. In addition, although we had adequate power to detect a similar effect size to that seen by Dick et al., 2006 in our case-control analyses, under a dominant model with an odds ratio of 2.0, our family-based analyses had only 66% power; when considering perhaps more realistic effect sizes, the current study is underpowered. Additional work is needed to confirm the findings of Dick et al., 2006.
