Study hypotheses were tested using the General Liner Model (GLM), treating the predictor (glucose levels), as a continuous variable. First, we examined the relationship between baseline (pre-treatment) blood glucose levels and (1) percent heavy drinking days (PHDD); and (2) Craving, each tested separately, at baseline. Second, we performed GLM-based analyses with repeated trials, such that blood glucose level at baseline, treated as a continuous variable, was used to predict during-treatment PHDD at weeks 4, 8, 12, and 16, while controlling for baseline PHDD. Additionally, since blood glucose levels are influenced by both BMI (Ferrannini et al., 2004) and gender (Regitz-Zagrosek et al., 2007), all analyses statistically controlled for both BMI and gender. All analyses were performed using SAS Statistical Software (SAS Inc, 2003) and statistical significance wasset at p < .05
