A considerable literature has amassed documenting strong associations of substance use, abuse and dependence with a range of mental disorders in community samples (1
). These forms of comorbidity have been observed for diverse substances and have negative consequences for both the persistence and severity of these disorders (13
). The reasons for these associations have been widely debated (21
), including the possibility that they may result in part from causal effects of primary mental disorders. If such effects could be documented rigorously, they would have important implications for refining substance use prevention and treatment strategies. However, the majority of existing epidemiologic research has been based on analyses of syndrome severity changes attributable to comorbidity (18
) or, more commonly, on retrospective estimates concerning the order of disorder onset (6
). Such estimates are susceptible to forward telescoping (25
) or other memory biases, and therefore do not offer definitive evidence that specific mental disorders are risk factors for the development of substance use, abuse or dependence.
Longitudinal investigations capable of confirming the order of onset of these conditions are limited in number, but several have found that certain mental disorders predict the later onset of smoking or nicotine dependence (26
), alcohol abuse or dependence (29
), and drug abuse or dependence (29
). Associations have also been found in the opposite direction (12
), but retrospective and prospective studies both indicate that mental disorders have a temporally primary age of onset in the majority of these forms of comorbidity (13
). Although the full impact of primary mental disorders is unknown, simulation studies have estimated that their early treatment or prevention might reduce 15% to 40% of cases of secondary substance dependence (8
). These simulations provide upper bound estimates due to the fact that mental-substance disorder comorbidity might be attributable, at least in part, to shared etiologic factors rather than solely to the causal effects of mental disorders. To date, these estimates have not been based on prospective community surveys and it is therefore unclear to what extent they may be distorted by retrospective dating of disorder onset or to biases associated with longitudinal assessments in clinical samples.
An additional concern for comorbidity research is the lack of information regarding the specific stages of substance use trajectories that are most strongly associated with pre-existing mental disorders. The common approach has been to examine mental disorders as predictors of substance dependence among all individuals in a given sample, ignoring the possibility that mental disorders may more strongly predict substance use, or the transition from use to abuse, rather than the onset of dependence alone. An alternative approach would be to examine predictors of these different transitions, thereby gaining information about the precise stage at which mental disorders have their greatest predictive effects. This strategy that has recently been applied to investigate the influence of sociodemographic predictors of the transitions between categories substance use, abuse and dependence (36
) and has provided novel information that is inaccessible to classic analytic approaches. Application of the same approach to mental-substance comorbidity would refine our understanding of the prospective associations between primary mental disorders and the subsequent onset of substance use, abuse and dependence.
Using data from a nationally representative two-wave panel survey of the US population spanning a 10-year period, the current investigation examines the risk of pre-existing mental disorders for the initial onset of use, abuse, and dependence relative to three classes of substances: nicotine, alcohol and illicit drugs. Baseline predictors include lifetime history of mood disorders, anxiety disorders, behavior disorders and additional substance use conditions. In order to provide comparability with the previous literature (1
), unconditional analyses examine mental disorders as risk factors for the first onset of nicotine, alcohol and drug dependence, while conditional analyses then decompose these aggregate associations by examining the predictive effects of mental disorders on transitions from nonuse to use, from use to abuse, and from abuse to dependence for each substance class.