Our primary interest is in determining whether the ALDH2 effect on drinking outcomes was moderated by each environmental indicator (as indicated by inclusion of the two-way interaction term in model F in the tables) and further whether this effect was moderated by age (by inclusion of the three-way interaction effect in model H). Regarding the parental combined alcohol problem index, for the drinking index and alcohol abuse and dependence symptom count outcomes, but not the binary alcohol abuse or dependence diagnosis outcome, the AIC-determined best-fitting model included the hypothesized interaction between the parental combined alcohol problem index and ALDH2 polymorphism (table 4, Model F for each outcome). However, by the likelihood ratio test, for the alcohol abuse and dependence symptom count outcome, models including the influence of the parental combined alcohol problem index did not differ significantly from the best fitting model without the parental combined alcohol problem index (Model C*). In no case was there evidence for the three-way ALDH2 by Parent combined alcohol problem index by Age interaction effect.
