in the current study and before age 23 in a study of twins followed longitudinally (van Beek et al. 2012), but were not noted previously in reports of univariate genetic analysis (Heath et al. 1997; Prescott et al. 1999; Knopik et al. 2004). This is probably the consequence of the confounding of genetic non-additivity and shared environmental effects in the twin design, with the former decreasing and the latter increasing the DZ correlation relative to the MZ correlation, creating the potential for non-additivity to mask shared environmental effects and vice versa. In a multivariate analysis, where a second trait shows strong shared environmental influences (as proved to be the case for remission in males), it would not be unexpected to uncover shared environmental effects in the primary traits that were not detectable in a univariate analysis. Shared environmental influences on AUD were found in a twin study that used data from twins who had been hospitalized for alcoholism, defined as abuse, dependence or alcoholic psychosis (Prescott et al. 2007). In that study, twin resemblance for hospitalization may have effectively been a second trait that allowed the discernment of shared environmental influences, even in a univariate analysis.
