In both the AA and EA models, the estimated proportion of variance in age at first drink accounted for by additive genetic influences fell within the (broad) range reported in the literature. The difference between the models, 44% vs. 26% for the AA and EA subsamples, respectively, is noteworthy, but given the absence of previous research in this area, it is not easily interpreted. A lower heritability estimate might be expected if age at first drink is a robust marker for genetic liability to a highly heritable problem drinking phenotype in European Americans but not in African Americans. However, the estimated heritability of problem use in the EA model was a modest 21%, compared with estimates of 50–60% for the more severe phenotype of alcohol dependence reported in previous studies based on primarily EA individuals (Heath et al., 1997; Kendler et al., 1994; Knopik et al., 2004; Reed et al., 1996; van den Bree et al., 1998). Also unlike AD, the problem use phenotype in our sample did not share a substantial degree of genetic variance with alcohol use initiation.
