We separate the analysis into two parts: (1) global analysis and (2) regional GMV and WM microstructure analysis, including 514 IDPs in total (139 GMV IDPs, 375 WM microstructure IDPs). The interactions of alcohol intake variables with sex and age are not significant in the global analysis (p > 0.001), so we exclude them from the regional analyses. To control the family-wise error rate in the regional GMV and WM microstructure analysis, we determine the significance thresholds for all regressions using the Holm method65, ensuring a family-wise error rate below 5%. When testing for M hypotheses, this method orders the corresponding p-values from lowest to highest: p0,…, pM, and identifies the minimal index k such that pk > 0.05/(M + 1 − k). All hypotheses with an index m < k are then considered to be statistically significant. In our application, the significance threshold was determined to be 1.64 × 10−4.
