by population stratification or other third variables,52 it is highly unlikely in the present study because the participants in each of the data sets were predominantly white (approximately 95%) and the findings were consistent across multiple studies and levels of analysis. We did not use genomic controls because the cost of assaying hundreds of extra markers was high, the benefits of using genomic controls in a sample that is 95% white were uncertain,53 and a reasonable alternative was available (ie, using self-reported ancestry as a covariate). Finally, future studies may benefit from testing additional loci in CNR1. A recent study26 used a haplotype approach to identify 2 SNPs that are associated with alcohol and drug dependence, 1 of which is linkage disequilibrium with rs2023239.
