Some limitations of our study should be noted. First, our sample drew from parent studies with differing ascertainment criteria and this may have contributed to heterogeneity in those endorsing craving. Our factor analyses, however, indicated that heterogeneity is attributable to the relative frequency of endorsement of craving across the parent studies, but not in its relationship with dependence. Second, as our sample was designed to be enriched for alcohol and drug dependent individuals, results from our study may not generalize to the population. Nonetheless, phenotypic similarities with epidemiological studies are apparent. Third, none of the association signals surpassed genomewide significance and hence, our comparison across craving, dependence without craving and DSM-IV dependence, which relied on p-values, may be confounded with a high rate of false positive findings. Replication will be necessary before we can be confident in these findings. Although 97% of individuals endorsing the craving item were alcohol dependent, the observation that some SNPs uniquely associated with craving while other SNPs were only associated with alcohol dependence demonstrate both the heterogeneity of the alcohol dependence diagnosis and the potential
