Alcohol quantity mediates the association between daily alcohol and cannabis co-use and alcohol consequences
aCenter for Alcohol and Addiction Studies, School of Public Health, Brown University, Providence, RI, United States
bRutgers Addiction Research Center, Rutgers Robert Wood Johnson Medical School, Piscataway, NJ, United States
cPacific Institute for Research and Evaluation, Prevention Research Center, Berkeley, CA, United States
dDepartment of Psychological Sciences, University of Missouri, Columbia, MO, United States
eProvidence Veterans Affairs Medical Center, Providence, RI, United States
*Correspondence to: Center for Alcohol and Addictions Studies, Brown University, G-S121-4, Providence, RI 02912. United States., Rachel_Gunn@brown.edu (R.L. Gunn).Abstract
Background:
Alcohol and cannabis are commonly used substances for young adults, and person-level co-use (i.e., concurrent or simultaneous use of both) is associated with increased likelihood of experiencing positive and negative alcohol-related consequences. However, findings regarding within-person effects (i.e.,day-level) co-use on consequences are mixed, possibly due to inconsistency in including alcohol quantity (i.e., total number of standard drinks consumed) when examining the association between co-use and consequences. In the present study, we examined whether the number of drinks mediates the association between co-use and positive or negative alcohol consequences at the day level.
Methods:
Data from morning reports in a 28-day field-based study of young adults reporting frequent past 60-day alcohol and cannabis use (N = 115) were used to test multilevel mediation models.
Results:
We found significant mediation for both positive and negative alcohol consequences; consuming more alcoholic drinks on co-use days, relative to alcohol-only days, was associated with a higher likelihood of experiencing negative consequences and a lower likelihood of positive consequences. These results suggest that daily number of drinks is a significant driver of the relationship between co-use and alcohol-related consequences at the day-level.
Conclusions:
In the context of increased cannabis use among young adults, this finding provides critical information for prevention and intervention efforts aimed at reducing the alcohol-related consequences associated with co-use days. Overall, reducing total alcohol consumption remains a prominent harm-reduction strategy among this population.
1.Introduction
Rates of alcohol and cannabis co-use are highest in young adulthood (Patrick et al., 2019) and continue to rise (McCabe et al., 2021). Alcohol and cannabis co-use, and in some investigations, simultaneous use (i.e., use of both substances so that the effects overlap), are associated with increased alcohol consumption and related consequences relative to single substance use at the person-level (Lee, 2022). However, questions remain as to how co-use may confer risk for alcohol-related consequences at the daily level. Understanding whether and how co-use relates to acute alcohol consequences may inform prevention efforts among young adults by shedding light on factors that increase risk during co-use events and therefore provide a rationale for tailored harm reduction efforts, such as promoting protective behavioral strategies that may reduce momentary risk.
Day-level examinations of the association between co-use (versus alcohol-only use) and alcohol-related consequences consistently show that co-use is associated with increased likelihood of negative alcohol-related consequences (Fairlie et al., 2023; Linden-Carmichael et al., 2020; Mallett et al., 2017). However, there are inconsistent modeling approaches across these studies for analytically controlling for total alcohol quantity (i.e., the number of drinks at the day level). Some studies demonstrate that the association between co-use and negative alcohol consequences is attenuated when covarying for alcohol quantity (Lee et al., 2020), suggesting that co-use does not confer excess risk for alcohol-related consequences beyond total drinks consumed. In contrast, studies suggest that the association between simultaneous use and negative alcohol consequences is robust when covarying for total alcohol consumed (Fairlie et al., 2023). Moreover, several studies that examined patterns of co-use found consistent and strong associations between alcohol quantity and negative alcohol consequences (Gunn, Sokolovsky, et al., 2021; Howe et al., in press; Wardell et al., 2024). Overall, by controlling for alcohol quantity and finding a robust association with consequences, this research suggests that quantity may explain how co-use increases risk for negative alcohol-related consequences. However, to date, no studies have directly examined whether the number of drinks statistically mediates (i.e., via an indirect effect) the association between daily co-use and negative alcohol consequences.
Among young adults, it is also critical to understand the positive consequences associated with co-use, as they occur more frequently, serve as valuable reinforcers, and motivate continued use (Barnett et al., 2014; Lee et al., 2011; Park, 2004; Usala et al., 2015). Recent work found that, like negative consequences, positive consequences are more likely to occur with co-use relative to alcohol-only days (Boyle et al., 2024; Lee et al., 2020). Interestingly, one study found that these associations persist even when controlling for alcohol quantity, (contrary to the same analysis with negative consequences) (Lee et al., 2020). Furthermore, one study found that simultaneous use was associated with more positive consequences on lighter drinking days, but not on heavier drinking days (Boyle et al., 2024). While this work suggests that total alcohol consumption may partially explain the association between co-use and positive alcohol consequences, no study has examined this mechanism directly.
1.1.Current study
This study leveraged naturalistic data to examine whether daily alcohol quantity (i.e., number of alcoholic beverages consumed each day) mediated the association between daily co-use (relative to alcohol-only) and positive and negative alcohol-related consequences experienced the same day. We expected that co-use would be positively associated with more total drinks and, in turn, more daily alcohol-related negative consequences. In contrast, we expected that lower total drinking on co-use days would be associated with more daily positive alcohol consequences.
2.Methods
2.1.Study design
This report presents the primary aims of a study of young adults (ages 18–30) who frequently use alcohol and cannabis (Gunn et al., 2024). Participants were recruited via community flyers and social media, followed by a confirmation of eligibility via telephone. Eligibility criteria included alcohol use (≥2x/week, with at least one heavy drinking occasion/week), cannabis use (≥1x/week), at least one instance of simultaneous use in the past 30 days, English fluency, smartphone access, no current mania or psychosis, and not seeking substance use treatment. Eligible participants (N = 115) completed a baseline session, either in person or via Zoom, during which they completed baseline measures and received training on using the mobile assessment application.
2.2.Data collection procedures
Following baseline, participants began a 28-day ecological momentary assessment (EMA) study, which included event-contingent (i.e., participant-initiated based on substance use), signal-contingent (i.e., random), and daily ‘morning’ surveys, starting at 9:00 AM and available until noon. The morning survey was exclusively used in the present analyses and asked participants to report on details of their prior-day alcohol and cannabis use and related consequences. Participants completed 88 % of morning reports.
2.3.Measures
Substance use days were queried each morning with the item, “Did you use alcohol and cannabis yesterday so that their effects overlapped?” and categorically coded as simultaneous (“I used alcohol and cannabis and the effects overlapped”), concurrent (“I used alcohol and cannabis but the effects did not overlap”), alcohol-only (“I only used alcohol”), cannabis-only (“I only used cannabis”), or non-use days (“I did not use alcohol or cannabis”). We compared alcohol-only days to co-use days (simultaneous and concurrent). We combined the co-use types because research shows limited differences between simultaneous and concurrent use on consequences at the day-level (Sokolovsky et al., 2020).
Alcohol quantity was assessed via numeric entry of the number of standard alcoholic drinks consumed the prior day (“How many total standard drinks did you have yesterday?”).
Alcohol-related consequences were captured each morning using 14 negative and 8 positive items (“Did you experience any of the following yesterday as a result of your alcohol use? Select all that apply”). Consequences were selected from prior studies based on items with the highest endorsement and identified as “positive” or “negative” (Lee et al., 2017, 2020; Linden-Carmichael et al., 2020). Items were summed to calculate the total number of negative or positive consequences experienced that day.
Person-level alcohol and cannabis use was measured at baseline using a 60-day Timeline Follow-back interview (Sobell and Sobell, 1992) to characterize the sample.
Covariates included self-reported sex assigned at birth (reference group: Male), and weekend (1 ={Friday, Saturday} 0 ={all other days}).
2.4.Data Analysis
Descriptive statistics were computed at the person-level or aggregated across all person-days for daily measures. Repeated-measures correlations were run to examine associations between primary variables of interest (Bakdash et al., 2017). To examine the indirect effect of alcohol and cannabis use patterns on negative and positive consequences via alcohol quantity on days where alcohol use was reported, we first fit mediator and outcome models with a series of linear and generalized linear mixed models (LMEM, GLMM respectively) (Hedeker, 2005). For the mediator model, we regressed daily alcohol quantity on daily co-use, disaggregated into between (i.e., person-level proportion of co-use days) and within-subjects (i.e., person-mean centering, computed as deviation of daily co-use from proportion of co-use days) (Enders and Tofighi, 2007). Alcohol quantity was fit to a Gaussian distribution. The outcome models regressed daily negative or positive consequences (level 1) on the disaggregated co-use variables (level 1) and the mediator (alcohol quantity, level 1) in two separate models. Both models were fit to a poisson distribution. All models included unstructured covariance matrices, random intercepts, and covariates. The outcome model also covaried for person-level mean alcohol quantity. We grand-mean-centered the person-level proportion of co-use days (all models) and mean alcohol quantity (outcome models) to increase the interpretability of model intercepts. Mediation analyses based on a potential outcomes framework (Imai et al., 2011) were conducted using the package mediation (Tingley et al., 2014); a sensitivity analysis examined the robustness of the indirect effects to post-treatment confounding, a violation of the sequential ignorability assumption (e.g., confounded mediating mechanisms (Imai et al., 2011). All analyses were conducted in R 4.4.1 (R Core Team, 2017).
3.Results
Descriptive statistics from the baseline assessment are presented in Table 1. One participant did not complete any morning reports and thus was omitted from analyses. On average across the 28 EMA days, participants reported 2.86 (SD=3.96) alcohol-only use days, 6.67 (SD=5.44) cannabis-only days, 10.51 (SD=7.06) co-use days, and 4.92 (SD=5.61) non-use days. On alcohol-only days, participants consumed M= 4.48 (SD=3.16) standard drinks and experienced M= 0.67 (SD=0.98) negative and M= 1.53 (SD=1.73) positive consequences. On co-use days, participants consumed M= 5.08 (SD=3.41) standard drinks and experienced M= 0.64 (SD=1.03) negative and M= 1.61 (SD=1.76) positive consequences. Supplementary Table 1 presents repeated measures correlations. Table 2 presents results for the mediator, outcome, and full mediation models. In the mediator model, the intraclass correlation (ICC) was 0.28. Day-level co-use exerted a significant positive effect on total drinks consumed, suggesting that co-use was associated with heavier daily drinking.
3.1.Negative consequences mediation
In the outcome model for negative consequences, the ICC was 0.41. A significant positive association was observed between daily total drinks and negative consequences. As hypothesized, in the mediation model, the indirect effect of daily co-use on negative consequences via total drinks was significant and positive. The direct effect was non-significant.
3.2.Positive consequences mediation
In the outcome model for positive consequences, the ICC was 0.38. A significant negative relationship was observed between daily total drinks and positive consequences. As hypothesized, in the mediation model, the indirect effect of daily co-use on positive consequences via total drinks was significant and negative, indicating that drinking quantity was one mechanism by which co-use reduced positive alcohol-related consequences. The direct effect was non-significant.
3.3.Sensitivity analysis
We conducted a sensitivity analysis varying the level of correlated error between the mediator and outcome models, and calculating the indirect effect at each level.1 For negative consequences, a substantial violation (ρ = 0.4) was required for the confidence interval of the indirect effect to include 0, suggesting robustness to unobserved confounding. For positive consequences, the confidence interval for the indirect effect always included 0 at low levels of ρ and the effect could switch directions with larger violations, suggesting the need for interpretive caution and replication for this outcome (see Supplemental Materials for sensitivity analysis plots).
4.Discussion
This study examined whether the quantity of alcohol young adults consumed when they engaged in alcohol-cannabis co-use explained the well-documented association between daily co-use (relative to alcohol-only) and positive and negative alcohol-related consequences experienced the same day. Results suggested a significant mediation for both positive and negative consequences, such that drinking more on co-use days was associated with experiencing more negative and fewer positive consequences.
Findings provide evidence for the complementary hypothesis (Gunn et al., 2021; Risso et al., 2020; Subbaraman, 2016), in that they suggest that co-use in a day is associated with higher volume of alcohol consumption (compared to alcohol-only days), which in turn leads to greater negative alcohol-related consequences. While previous studies identified the strength of the association between total alcohol consumed in a day and negative alcohol-related consequences in young adults, and even alluded to volume of alcohol consumed as a potential mechanism (Fairlie et al., 2023; Lee et al., 2011; Sokolovsky et al., 2020), no work explicitly tested this hypothesis. Our findings found, indeed, that the amount of alcohol consumed mediates the relationship between co-use and negative alcohol consequences.
We also found evidence for alcohol quantity as a mechanism in the association between co-use and positive consequences in a day. Specifically, we found an inverse relationship between alcohol quantity and positive consequences, such that heavier drinking on co-use days (versus alcohol-only days) was associated with fewer positive consequences. While sensitivity analyses suggest that these results should be replicated and confirmed, findings are consistent with those for negative consequences, such that when more alcohol is consumed on co-use days, the individual’s experience is negatively affected, as reflected in lower positive and higher negative consequences. These findings provide valuable insights for prevention and intervention efforts, such as personalized interventions reflecting how using cannabis and alcohol together may serve to increase drinking, and in turn, reduce the likelihood of experiencing positive consequences and increase the likelihood of negative consequences during critical developmental substance use phases, such as young adulthood. Evidence-based practices, such as promoting protective behavioral strategies (Borden et al., 2011; Martens et al., 2004), are effective in similar intervention and prevention frameworks and could be tailored to young adults who engage in regular co-use via just-in-time adaptive interventions (Nahum-Shani et al., 2017).
4.1.Limitations and future directions
This work should be considered in the context of several limitations. First, the results may not generalize to other populations, such as adolescents, those seeking treatment for AUD, or minoritized groups. Second, in consequence items, we asked if the consequences were experienced “as a result of your alcohol use”, which may have led to lower endorsement rates for those who did not attribute their consequences to substance use. Third, given that the variables were assessed at the same time, and because consequences could have occurred in the midst of a drinking event, full temporal mediation cannot be inferred. Fourth, although we did conduct sensitivity analyses to examine the potential impact of unaccounted-for confounding between the mediator and outcome, we did not control for all possible “pre-treatment” confounds, such as social context, which may impact both alcohol consumption and consequences. Future work should explore environmental and contextual factors that interrelate to predict substance use behavior. Finally, we were unable to control for cannabis use (day-level) quantity in analyses due to rank-deficiency that would be present in modeling. Future work should examine how the quantity of cannabis use on co-use days might contribute to or explain the association between co-use and negative consequences. Despite these limitations, this work contributes an important finding to the alcohol cannabis co-use literature, that co-use on a day leads to greater drinking and, in turn, experience of more alcohol-related negative consequences and less positive consequences. Understanding this critical mechanism begins to clarify competing literature and may contribute to prevention and intervention development efforts to reduce the harms associated with co-use.
Supplementary Material
Acknowledgements
The National Institutes of Health supported this project: K08 AA027551 (Gunn), T32 DA016184 (Tidey), K24 AA026326 (Miranda), K08 DA048137.
Appendix Group
Supporting information
Supplementary data associated with this article can be found in the online version at doi:10.1016/j.drugalcdep.2026.113043.
| N (%) / Mean (sd) | |
|---|---|
| Age | 23.24 (3.4) |
| Sex at birth: Female | 66 (57 %) |
| Race | |
| White | 92 (80 %) |
| Black | 19 (17 %) |
| Asian | 7 (6 %) |
| Other | 7 (6 %) |
| American Indian/Alaskan Native | 2 (2% |
| Native Hawaiian/Pacific Islander | 1 (1 %) |
| Ethnicity: Hispanic/Latinx | 19 (25 %) |
| Substance Use (60-day TLFB) | |
| % Alcohol use days | 54.43 (21.13) |
| % Cannabis Use days | 76.21 (27.84) |
| Path | Negative Consequences | Positive Consequences | ||||||
|---|---|---|---|---|---|---|---|---|
| Estimate | SE | p | Estimate | SE | p | |||
| Model 1: Mediator (Quantity) | ||||||||
| Intercept | 5.84 | 0.28 | < .001 | * | ||||
| Co-use (within) | 0.69 | 0.25 | .005 | |||||
| Co-use (between) | 0.22 | 0.59 | .706 | |||||
| Sex (ref: Male) | −1.90 | 0.36 | < .001 | |||||
| Weekend | 0.32 | 0.15 | .039 | |||||
| Model 2: Outcome (Consequences) | ||||||||
| 95 % CI | 95% CI | |||||||
| IRR | Low | High | p | IRR | Low | High | p | |
| Intercept | 0.15 | 0.11 | 0.21 | < .001 | 1.33 | 1.08 | 1.63 | .007 |
| Total drinks | 1.20 | 1.17 | 1.23 | < .001 | 0.98 | 0.96 | 0.99 | .004 |
| Mean drinks | 0.94 | 0.86 | 1.04 | .229 | 1.04 | 0.98 | 1.11 | .212 |
| Co-use (within) | 1.01 | 0.80 | 1.27 | .957 | 1.13 | 0.98 | 1.31 | .084 |
| Co-use (between) | 0.63 | 0.37 | 1.08 | .095 | 1.05 | 0.72 | 1.52 | .816 |
| Sex (ref: Male) | 1.34 | 0.92 | 1.97 | .131 | 1.14 | 0.88 | 1.48 | .304 |
| Weekend | 1.01 | 0.88 | 1.16 | .835 | 1.14 | 1.04 | 1.24 | .004 |
| Mediated Effects | ||||||||
| 95 % CI | 95% CI | |||||||
| Estimate | Low | High | p | Estimate | Low | High | p | |
| Direct Effect | 0.01 | −0.15 | 0.19 | .988 | 0.22 | −0.03 | 0.50 | .098 |
| Indirect Effect | 0.09 | 0.02 | 0.17 | .008 | −0.03 | −0.06 | −0.006 | .006 |
| Total Effect | 0.10 | −0.07 | 0.31 | .296 | 0.19 | −0.06 | 0.48 | .160 |