This study was approved by the Rutgers University IRB. We used Genomic SEM16 to estimate SNP effects in two models (Figure 1). In our two-factor model, the SUD factor included problematic alcohol use PAU)49, problematic tobacco use (PTU)9, opioid use disorder (OUD)24, and cannabis use disorder (CUD)25. The remaining non-SUD indicators from the original externalizing model (attention deficit hyperactivity disorder [ADHD]19, risk taking [RISK]22, number of sexual partners [NSEX]22, age at first sex [FSEX]22, smoking initiation [SMOK]21, and cannabis initiation [CANN]20) loaded onto a behavioral disinhibition factor (boxes B and C in Figure 1). In our common factor model, all behavioral disinhibition phenotypes and SUDs loaded onto a common externalizing factor (Box A in Figure 1). We retained the same GWAS as were used in the respective original multivariate GWAS except in the case of OUD, for which a better powered GWAS was available.
