The prevalence of alcohol and drug dependence were compared across ancestral groups using chi-square tests. The number of DSM-IV criteria endorsed by individuals in AA and EA families were compared (total number and for each drug) using an ordinary least squares regression that accounted for sex. Birth cohorts (1890–1929, 1930–1949, 1950–1969, ≥1970) were included in all COGA analyses as covariates to account for secular trends (see37) across this wide range of birth years. Over and above birth cohort, age was not a significant predictor of ANYDEP (p > 0.3). A confirmatory factor analysis of substance dependence diagnoses was fitted to the data separately for EAs and AAs, to determine phenotypic patterns of comorbidity using MPLUSv838. The comparative fit index (CFI), standardized root mean square residual (SRMR), and the root mean square error of approximation (RMSEA) were used to assess model fit. Heritability for ANYDEP and drug_noalc was estimated using SOLAR39 in the EA and AA families separately, using familial relatedness (but not GWAS data) alone.
