This study was conducted between October 1, 2023, and May 21, 2024. The study design is shown in the Figure. We chose the 2-stage design by following the pioneering work of Khera et al,29,30 who developed the first potentially clinically applicable PGS. Multiple PGSs were tested in the screening stage and then the PGS with the highest estimability moved to the testing stage, avoiding multiple testing problems and reducing computational burdens. In this study, we calculated multiple PGSs by using different sets of concordant SNVs in the screening stage. Concordant SNVs were identified from 3 large-scale genome-wide association studies (GWASs) of AUD-related phenotypes: the Million Veteran Program (MVP),36 the UK Biobank (UKBB),37 and the FinnGen Consortium (FinnGen).38 The Collaborative Study on the Genetics of Alcoholism (COGA) and 2 datasets from the Database of Genotypes and Phenotypes (dbGaP) (phs000092.v1.p1 from the Study of Addiction: Genetics and Environment [SAGE], and phs000181.v1.p1 from the Australian Twin-Family Study of Alcohol Use Disorder [OZALC]) were used as the screening datasets. Two independent datasets, the All of Us Research Program (AOU) and the Indiana Biobank (IB),25,34
