Latent class analysis favored three-class solutions in both samples (Table 4). In the URS, all classes showed higher criterion endorsement probabilities and expectancy values for dependence than for abuse criteria, indicating quantitative differences in severity across classes (Figure 2). In the RS, however, the pattern of class 2 diverged from the other classes, representing qualitative differences. Table 5 illustrates the transition of cases according to their class membership in the URS and the RS. The majority of cases in class 3 in the URS were excluded in the RS because of no criterion endorsement (99.2%). A separate “abuse class” only emerged in the RS and was composed of individuals categorized as the “nearly zero class” within the URS (96.8%). A distinct “dependence class” was visible only within the RS and was formed by individuals categorized as class 1 in the URS (88.5%).