Higher level analyses were conducted on the alcohol versus water cue contrast images transformed into standard space. This model included these contrast images to determine pre- vs. post- effects, modeling each participant’s mean separately in order account for within-subject variability. Z-statistic images were thresholded with cluster-based corrections for multiple comparisons based on the theory of Gaussian Random Fields with a cluster-forming threshold of Z>2.3 (reduced to Z>1.96 for exploratory correlational analyses) and a probability threshold of p<0.05 (26). Anatomical localization within each cluster (maximum Z statistics and MNI coordinates) was obtained by searching within maximum likelihood regions from the FSL Harvard-Oxford probabilistic atlas. To account for the trend level difference in alcoholism severity factor scores between genotype groups, alcoholism severity was modeled as the explanatory variable in whole-brain group-level analyses. OPRM1 genotype (i.e., AA and AG/GG) was also entered as a predictor variable and examined in relation to the computed contrasts. These analyses were considered exploratory due to the relatively small sample sizes for genetic and correlational analyses.
