There are a number of limitations. First, the cross-sectional data makes it difficult to disentangle developmental and cohort effects. Although similar age-varying trends in substance use have been documented using longitudinal data (Chen and Jacobson, 2012; Evans-Polce et al., 2015), cohort effects have also been found (Lanza et al., 2015; Johnson et al., 2016). The higher prevalence of NUD among midlife adults and high rates of CUD among young adults may reflect generational differences in the acceptability and use of these substances. Second, the proportion of participants reporting a past-year OUD was small, and we were unable to look at other SUDs due to low prevalence. We were also unable to examine age trends in smaller racial/ethnic groups. Our SUD measures were self-reported symptoms, and thus do not represent clinical diagnoses. In addition, we only examined SUDs, and it is possible that differences by subgroup would differ if other aspects of substance use (e.g., frequency of use) were examined. Future research should examine this possibility.
