(none, early, sustained) was created to test whether early and sustained remission were statistically distinct categories. This variable was regressed on two dummy variables representing co-twin status on early and sustained remission in a multinomial logistic regression (with no remission as the reference category) and planned post-hoc tests were used to test whether co-twin early and sustained remission were differentially associated with twin early and sustained remission. Co-twin early and sustained remission were not differentially associated with twin early or sustained remission from two or more [Wald χ2(3)=0.67, p=0.88] or from three or more AUD symptoms [Wald χ2(3)=0.39, p=0.94]. Remission was coded as binary thereafter, with 1 representing individuals with any remission (early or sustained). Individuals who did not endorse two or more AUD symptoms (for broadly defined remission, n=3343) or three or more (n=4380) were coded as missing on remission, consistent with the two-stage model used to decompose variance into genetic and environmental components, described below.
