The finding that the association of the rs11174811 with DUD liability had opposite directions in the studies based on clinical samples (CEDAR/SADS and COGA) and in the population-based VATSPSUD study requires explanation. Although largely speculative at this stage, several possibilities can be contemplated. The association may be a spurious finding. This possibility conflicts, however, with the fact that the allelic effect of the original finding is supported by the association in another clinical sample, by the mediational effects on the family functioning indicator consistent with the hypothesized gene effect, as well as by the functional data. This “flip-flop” (59) finding may also be possible in the case of true allelic heterogeneity between the studies (60). Elucidation of the source of this heterogeneity might be impossible even with dense genotype data from each of the studies. The difference in the direction of the association may also be related to the differences in the sample ascertainment. In both CEDAR/SADS and COGA samples, the affected individuals, ascertained clinically, have a relatively severe DUD, frequently related to polysubstance abuse, whereas in the population-based Virginia
