Descriptive statistics for covariates were calculated by AUD/remission status. The bivariate association of each covariate with AUD/remission status was tested using multinomial regression analysis with persistent AUD as the reference category. Concordance rates for AUD/remission status in related and unrelated pairs were calculated. Relatives’ AUD/remission status was the dependent variable in a multivariable multinomial logistic regression, with relatives’ persistent AUD as the reference category and non-abstinent and abstinent remission as the two outcome categories. The primary independent variables were proband non-abstinent and abstinent remission; their association with relatives’ remission status was adjusted for the covariates listed above. The interactions of proband non-abstinent and abstinent remission with the dummy variable representing related pairs were tested one at a time in the fully adjusted regression to determine whether the association of probands’ AUD/remission status with relatives’ AUD/remission status varied in related and unrelated pairs. The final regression was calculated separately in related and unrelated pairs. The Huber-White robust variance estimator was used to adjust for the clustering of family data. Datasets and variables were created using SAS statistical software, version 9.2 (49). Analyses were performed using Stata Statistical Software, version 14.2 (50).
