This assessment of discriminatory accuracy shows that these panels of SNPs currently have limited clinical utility. One reason that many of the candidate gene SNPs did not replicate in the independent samples used to assess for clinical validity could be due to heterogeneity across samples; different genetic variants may contribute to risk in different populations containing varying subsets of alcohol-dependent individuals. Therefore, genetic risk could be unique to the samples used in these association analyses. For example, several variants have been found to have stronger association with AD in individuals with co-occurring drug dependence. Dick et al. showed that CHRM2 is associated with a form of AD that is comorbid with drug dependence, but not with AD alone (Dick et al., 2007a). In another case, Foroud et al. found that SNPs in TACR3 that were associated with AD in EA COGA families had the strongest association in individuals with more severe AD and comorbid cocaine dependence (Foroud et al., 2008). Furthermore, Agrawal et al. showed that GABRA2 is associated with AD only in individuals with comorbid drug dependence. When these
