Since COGA is a family cohort, mixed models were used with a random effect to adjust for the family relationships. Linear mixed models were used to compare PRS between AUD cases and controls, as well as in individuals with different counts of DSM-5 AUD diagnostic criterion in COGA full sample, FH+, FH?, and FH− subsamples. To evaluate the risk for AUD and AUD severity, generalized linear mixture models were fit by using the logit link function and the log link function for the status of AUD (yes or no) and DSM-5 AUD diagnostic criterion count (range 0–11), respectively. PRS were dichotomized to facilitate comparison with binary family history measures. As the prevalence of AUD ranges from 3.5% to 14.9% depending on sex and country (World Health Organization, 2018), we assumed an average prevalence of 10% and defined those top 10% of individuals with the highest PRS as the high PRS group and compared with the remaining 90% of individuals. For comparison purposes, we also performed analyses using the original continuous PRS scores (i.e., not dichotomized). For testing the associations between
