Because a control variable cannot be a confounder unless it is related to both exposure and outcome, association with either exposure or outcome can be used as a criterion for statistical adjustment (Schlesselman, 1982). In this study, control variables were selected for inclusion in the regression analyses based on their relations to outcome measures (Kleinbaum et al, 1988). All control variables that were even modestly related to each outcome (p<0.10) were adjusted statistically by regressing the outcome on prenatal alcohol exposure level and the control variables related to that outcome. Gestational age was used as a control variable in analyses of other birth outcomes (i.e., birth weight, birth length and head circumference). Pearson’s ‘r’ correlations were used to examine the relations of each control variable to each outcome. After controlling for potential confounders, the associations of the specific neurobehavioral outcomes with prenatal alcohol exposure measures and the “at-risk” alcohol metrics were evaluated with one-tailed tests with a significant alpha set at p<0.05. The one-tailed test and this level of significance were used because we had specific a priori, directional hypotheses
