(C), and nonshared environmental factors (E). We also calculated genetic and environmental correlations (ie, the correlation of genetic and environmental variance components between the 2 traits). We used the direct-symmetric parameterization, which lowers the risk for type I errors, but can produce negative variance contributions.49 For dichotomous variables, we used a liability-threshold model where the categories (lifetime diagnosis of OCD and substance misuse present vs not present) were assumed to reflect an underlying normal distribution of liability.50 Models were adjusted for sex and birth year. The model with fewer parameters was considered to have the best fit if not significantly worse than the full ACE-ACE model (including A, C, and E components for both phenotypes), as indicated by the Akaike information criterion and P value for the reduction of model fit in the likelihood ratio test. Two-sided P < .05 indicated statistical significance. Analyses were conducted with R, version 4.0.5 (R Core Team).
