While these studies provide compelling data about the scope of problems stemming from the co-use of these substances, a fundamental limitation of such population-level, cross-sectional studies is that they are unable to answer questions about event-level patterns of use. That is, the cross-sectional nature of these studies cannot provide information about the pattern and predictive relationship of simultaneous co-use within a given day or drug-use event, which may be especially critical to understanding the co-use of marijuana with other substances. For example, individuals report both using marijuana as a substitute for tobacco or alcohol and in a sequential/simultaneous manner to produce additive or subtractive subjective effects . It is difficult to differentiate such patterns of use unless examining event-level data. The few, recent studies that have used a fine-grained approach to study simultaneous use, while important, have limitations that may affect generalizability. A study examining event-level alcohol and marijuana co-use in adolescents did not report patterns of co-use, only the context of and consequences from simultaneous co-use . Another study that examined daily patterns of marijuana and alcohol co-use, but not cigarette use, in a predominantly male , veteran population found that moderate and heavy-drinking were more likely to occur on days which marijuana was used. Further, while individuals with AUD or comorbid AUD + CUD were more likely to drink heavily on such days, individuals with CUD were less likely to drink heavily, which the authors interpreted as supporting marijuana substitution . Lastly, Gunn et al., found that daily marijuana use was associated with greater alcohol consumption in college students, and this predictive relationship strengthened over a two-year period. However,cannabis grow equipment they did not report on tobacco use nor the influence of sex on the relationship between daily marijuana and alcohol use.

In light of the high rates of marijuana co-use with alcohol and tobacco in epidemiological studies but relatively absent data on event-level patterns of use, the goal of the present study was to examine daily patterns of alcohol, tobacco, and marijuana co- and tri-use in non-treatment seeking drinkers who report regularly using tobacco and marijuana. To our knowledge, no studies have examined the daily co-use or triuse of all three substances at the individual, event-level in the same sample. Because of the strong evidence for the co-use of all three substances, we hypothesized that use of one substance would indiscriminately increase the odds of same-day use of a second substance, and the use of two substances would subsequently increase the odds of using the third. Finally, as an exploratory aim, we examined whether sex moderated any observed daily patterns of co- or tri-use. While men tend to use marijuana, alcohol, and cigarettes earlier, heavier, more frequently, and have greater dependence rates than women , women may have more severe consequences from substance abuse and enter treatment earlier than men . Given the general dearth of event-level substance use studies, sex differences in patterns of simultaneous co-use have obviously not been well characterized. However, at the population level, men have higher rates of marijuana co-use with each alcohol and tobacco and display a more rapid escalation in the frequency of this co-administration than women . The characterization of sex differences in patterns of event-level co-use also may have important implications for understating the etiology and treatment of addiction. All study procedures were approved by the University of California, Los Angeles Institutional Review Board and conducted in accordance with the Declaration of Helsinki. The reported sample draws from baseline data collected as a part of four human laboratory studies. Three studies examined pharmacotherapies for alcohol use: naltrexone in an Asian American sample , ibudilast , and ivermectin .

The fourth study was an alcohol self administration study , resulting in a total sample of 551 participants. Each study recruited a sample of non-treatment seeking, regular drinkers from the Los Angeles area using identical recruitment methods of print and online advertisements. Interested participants completed an initial telephone screening to determine eligibility. During the telephone screening, all participants were asked to report their drinking over the past three months prior to enrollment. The drinking requirement for each study had the following inclusion criteria: naltrexone in Asian Americans – female requirement of > 4 drinks per week and male requirement of > 6 drinks per week, as well as have an Alcohol Use Disorder Identification Test score greater than 8; ibudilast and ivermectin – requirement of > 48 drinks per month and score > 1 on the CAGE questionnaire assessing for alcohol problems; self-administration – female requirement of > 7 drinks per week and male requirement of > 14 drinks per week. Age restrictions for the naltrexone and ibudilast study were between 21–55, whereas participants had to be between 21–65 for the ivermectin study and 21–45 for the self-administration study. Only two studies had ethnicity requirements. Participants in the naltrexone in Asian American study were of East Asian ethnicity and participants in the self-administration study were Caucasian. All studies shared the following exclusion criteria: 1) current involvement in treatment programs for alcohol use or having received treatment in the past 30 days; 2) use of nonprescription drugs or prescription medications for recreational purposes; 3) self-reported history of exclusionary psychiatric disorders assessed during telephone interview; 4) currently using antidepressants, mood stabilizers, sedatives, anti-anxiety medications, seizure medications, or prescription pain killers ; 5) self-reported history of contra-indicated medical conditions or any other medical condition that may interfere with study participation; 6) intense fear of needles or adverse reactions to needle puncture; and 7) if female, pregnancy, nursing, planning to get pregnant in the next 6 months or refusal to use reliable method of birth control.

Specific to the ivermectin study, participants were excluded if they had a Body Mass Index less than 18.5 or greater than 30. For the self-administration study, participants were excluded if they weighed over 265 pounds. If participants were eligible following the telephone screening, they completed an in-person screening visit where written informed consent was obtained. During the screening visit, participants were required to produce a breath alcohol concentration of 0.000 g/dl, and test negative for pregnancy and drug use, except for marijuana, on a urine toxicology screening.Participants completed a battery of measures at the screening visit. A demographics questionnaire assessed age, sex, and ethnicity. The Timeline Follow-Back queried daily alcohol consumption in standard drinks, number of cigarettes smoked per day,mobile grow system and marijuana use during the previous thirty days. Marijuana use was assessed in a dichotomous fashion ; route of marijuana administration was not recorded. Alcohol and cigarette use was assessed as a continuous variable. Use of other tobacco products, e.g. snus or chewing tobacco, was not assessed. The Fagerström Test of Nicotine Dependence queried extent of nicotine dependence. The AUDIT was administered to evaluate severity of drinking. The Cannabis Use Disorders Identification Test , a reliable and valid adaptation of the AUDIT, was given to assess marijuana use severityOf the 551 total subjects who were screened for the four studies from which we culled data, 541 reported using alcohol on the AUDIT and/or TLFB, 296 reported using marijuana on the CUDIT and/or TLFB, and 260 reported using cigarettes on the FTND. As this study aimed to understand patterns of co- and tri-use among all three substances, we included only participants who reported using alcohol, cigarettes, and marijuana on a monthly basis. This selection resulted in a final sample of N = 179 participants. While this represents a significant decrease in the number of subjects, statistical power is still quite high for these analyses. The proposed analyses test the association between drug use on a per-day basis . Thus, the sample size for this study is properly conceptualized in terms of both the number of subjects, 179, but also the number of Level 1 observations which is 5390 total days. To confirm that this study is well powered, GPower 3.1.9.2. was used to conduct a power analysis. Based on a simplified repeated measures approach, a small effect size and a nominal α = 0.05 threshold, power for this study was exceptionally high . To explore patterns of marijuana, alcohol, and/or cigarette co-use a series of multilevel logistic models were run on 30-day timeline follow-back drug use data. Owing to the one on-one clinical interview nature of data collection for the key variables, there was no missing data in this study. Only individuals who reported using all three substances at least once per month were analyzed.

Multilevel logistic modeling was chosen because the data structure is nested with days nested within subjects which is appropriately modeled with a multilevel modeling approach and the outcome variable of whether a given drug was used on a given day is binary necessitating the logistic modeling approach. Multilevel logistic models were run via PROC GLIMMIX in SAS version 9.4 with a binomial dependent variable distribution and a logit link function. Models were run with cigarette use and marijuana use as the dependent variable and the other drug classes treated as predictor variables with main effects and interactions to test for potentially synergistic effects of combined use on the likelihood on the third drug use , or a suppressive effect where use of both drugs is associated with the same risk as singular use . Variables tested were Drink, a binary Level 1 variable coding whether alcohol was consumed on a given day, Smoke, a binary Level 1 variable coding whether cigarettes were smoked on a given day, and Marijuana, a binary Level 1 variable coding whether marijuana was used on a given day. To disentangle within-person effects from between person effects , person-means for each predictor variable were entered into models as Level 2 variables. To further ensure that the effects reported are within-subject effects, all Level 1 variables were treated as random at Level 2, meaning the effects were allowed to vary between subjects. Where interactions were observed, analysis of simple slopes were conducted through a recentering scheme to test the lower-order effects at specific levels of the interacting variables. In accordance with the NIH policy on considering sex as a biological variable and given the sex differences in the prevalence of marijuana, tobacco, and alcohol use in the US , we also tested for sex differences in the propensity for drug co-use. To test the robustness of these results several covariates were explored including: age, ethnicity, and source study, all of which were included as Level 2 variables. Ethnicity was examined as a covariate because one of the source studies was completely composed of individuals of East Asian descent , and age was included due to findings that patterns of co-use may differ by age . In line with the recommendations of , where discrepancies between models which included vs. omitted covariates were observed, we report the results of both models.Please see supplemental information for summary tables of all results. To first test whether average drug use frequency across these three drugs of abuse were correlated at the subject level, a series of linear regressions were conducted analyzing the correlation between proportion of days using each drug from the TLFB. Both drinking frequency and marijuana use frequency were found to independently predict cigarette smoking frequency , but drinking frequency did not predict marijuana use frequency . While these results suggest that subjects who use cigarettes and drink alcohol more often also use marijuana more frequently, these analyses do not address the central question posed in this paper of whether use of one substance on a particular day increases the likelihood of co-use or tri-use on that same day.To test whether use of one drug increases the likelihood of same-day co-use, a series of multilevel models were run with daily use of each drug included as Level 1 variables and drug use frequency person means also included as covariates to disentangle the between subject effects summarized in 3.1 from same-day effects. such that the effect of combined use of alcohol and marijuana on a given day was sub-additive . This interaction was such that on non-drinking days, marijuana use was associated with relatively large increases in the likelihood to smoke cigarettes .