The robustness of associations between maternal maximum consumption in these models might stem from the fact that it is a quasi-continuous measure while maternal alcohol dependence is dichotomous. We therefore considered a log-transformed count of alcohol abuse and dependence symptoms in place of the diagnostic variable. The dependent measures were again those consistently associated with maternal maximum consumption across cohorts: conduct disorder, ODD in females, any disruptive disorder, and all substance-related measures. Adjusting for the symptom count variable had inconsistent effects on the coefficients associated with maximum consumption; partial coefficients increased in magnitude over unadjusted ones as often as they decreased while remaining significant.
