Abstract
Background
Alcohol and cannabis co-use is associated with negative alcohol consequences and alcohol use disorder. However, mediating and distal effects remain largely unstudied. Co-use is associated with alcohol use disorder/negative consequences even when accounting for drinking levels and personality, suggesting that other person-level characteristics may explain relations between co-use and negative outcomes.
Method
The current study tested whether internalizing symptoms, strong correlates of co-use and alcohol use disorder, explained the effect of co-use on alcohol use disorder. Data from adults (N = 353,000) in the 2008–2019 National Study on Drug Use and Health (NSDUH) were used. Analyses tested whether (1) substance use profiles reduced/dissipated the effect of co-use on alcohol use disorder, (2) internalizing symptoms (anxiety, depression) reduced/dissipated the effect of co-use on alcohol use disorder, and (3) internalizing symptoms were indirectly associated with alcohol use disorder via co-use.
Results
When accounting for frequency/quantity of use, co-use was still associated with higher odds of alcohol use disorder. Anxiety and depression were related to higher odds of an alcohol use disorder, however, the effect of co-use on higher odds of alcohol use disorder remained. Anxiety and depression scores were indirectly associated with higher odds of alcohol use disorder via co-use.
Conclusions
Depressive and anxiety symptoms only accounted for a portion of the variance of co-use on alcohol use disorder, and there were indirect effects of internalizing symptoms through co-use. Future longitudinal research is needed to elucidate other person-level characteristics that drive associations between co-use and alcohol use disorder to target via interventions.
Ethics statement
All participants provided informed consent to the NSDUH methodology team, and all data were deidentified and made publicly available. The current study was exempt from the local institution’s IRB.
Disclosure statement
No potential conflict of interest was reported by the author.
Notes
1 Estimation of missing data via FIML required Monte Carlo Integration, since the models accounted for complex survey design of the NSDUH (variance adjustments and sample weighting). However, models were unchanged when run using listwise deletion.
2 Standardized coefficients are not available for models with a categorical mediator, and thus (1) indirect effects are reported as unstandardized coefficients and odds ratios, and (2) R2 estimates for co-use were not available.