The Relative Harms of Marijuana and Alcohol

David A Gorelick. American Journal of Drug & Alcohol Abuse. Volume 40, Issue 6, Nov 2014.

The US is currently engaged in a vigorous debate over the legal status of cannabis (marijuana). One factor raised in this debate is the relative harm posed by cannabis use compared to use of other psychoactive substances, both those currently legal (e.g. alcohol, tobacco) and those currently illegal (e.g. heroin, cocaine). Some proponents of loosening current legal restrictions on cannabis use (e.g. decriminalization, legal for medical purposes only, or legal for both medical and recreational purposes) argue that cannabis is no more harmful, or perhaps safer, than some legal substances, and certainly safer than many other illegal substances.

Special emphasis has been put on comparisons of cannabis with alcohol, perhaps for two reasons. First, alcohol is the second most widely used psychoactive substance (after caffeine) in the US. Second, the US has experienced major shifts in the legal status of alcohol over the past century, from legal and largely unregulated to illegal (Prohibition) and then to legal for those above a minimum age. Even the current President of the US has offered an opinion on this issue, stating earlier this year that “I don’t think it [marijuana] is more dangerous than alcohol.”

Regardless of how much weight one gives to the comparative harms argument within the overall cannabis debate, it is relevant to evaluate the scientific evidence behind it. The scientifically most rigorous study would be a prospective interventional trial in which large representative groups of substance-naïve individuals, matched on all relevant baseline characteristics, were randomly assigned to exposure to either alcohol or marijuana (but not both) and then evaluated over a period of time for substance-associated adverse effects. Random assignment to experiment-imposed substance exposure are key design features of a rigorous study because they serve to minimize bias from pre-existing (baseline) subject characteristics that might influence substance use and susceptibility or resistance to substance-associated adverse effects. Of course, such an ideal trial could not be conducted. The next most rigorous study design would be a prospective longitudinal observational trial in which subjects were selected at baseline to match ideal eligibility criteria as closely as possible (i.e. use of only one psychoactive substance) and evaluated for major potential confounding variables (e.g. socioeconomic status, psychiatric comorbidity). I am not aware of any such published study. Given the common use of both alcohol and marijuana, especially among adolescents (one-quarter of US high school students reported simultaneous use of alcohol and marijuana at least once in 2011), and the large sample size needed to have adequate statistical power to detect group differences (or have confidence that there were no differences), assembling the requisite study groups would be expensive and time-consuming. Therefore, I doubt that such a study will be conducted in the foreseeable future.

Current evidence is limited to retrospective studies, both cross-sectional and longitudinal, using a variety of data sources, variables, and analytical methods, which have compared the adverse effects associated with use of various psychoactive substances. A major flaw in many of these studies is the failure to distinguish users of a single substance from users of multiple substances. This leads to two potential confounds which limit interpretation of the findings: (i) If more than one substance has been used, with which substance should any observed adverse effect be associated? (ii) Users of substance A may have baseline characteristics, not found in non-users of substance A, that alter their risk of adverse effects if they use substance B.

The study by Palamar et al. in this issue of The American Journal of Drug and Alcohol Abuse offers new evidence on this issue and addresses some of the study design limitations mentioned above. The authors made creative use of community-based epidemiological data collected by the nation-wide US Monitoring the Future annual survey. This survey provided self-reports on 15 possible psychosocial harms (adverse effects) associated with substance use by adolescents, an age group particularly susceptible to such harms. By combining data from the five most recent surveys (2007–2011), the authors created a nationally representative cohort of 7437 high school seniors comprising three groups: 2949 adolescents who had used only alcohol, 240 who had used only marijuana, and 4249 who had used both alcohol and marijuana. The first two groups allowed a more confident association between substance use and adverse effects, as comparing them directly illustrated the comparative harms. The third group intrinsically controlled for any potential baseline differences between alcohol and marijuana users that might otherwise confound cross-substance comparisons. Furthermore, pair-wise comparison of each of the first two groups with the third provided a measure of the bias introduced by polysubstance use into comparisons between two substances. Another advantage of this study is that the sample size was large enough to allow evaluation, for at least some comparisons, of the influence of important covariates on the prevalence of adverse effects, such as race, ethnicity, sex, and frequency of lifetime use. Thus, a close reading of the study’s findings (particularly Table 2) sheds significant light on the comparative harms issue.

The study found that the vast majority (83–89%) of adolescents in each group reported one or more psychosocial adverse effects associated with their use of either alcohol or marijuana, such as interfering with ability to think clearly; hurting relationships with parents, significant others, friends, or teachers; reducing emotional stability; and causing unsafe driving. As expected, there was a significant positive association between frequency of use and prevalence of most adverse effects for both alcohol and marijuana, i.e. the more frequent users had a higher prevalence of adverse effects. These findings support the sensitivity and internal validity of the study.

Among adolescents who used only one substance, there was no consistent pattern of greater prevalence of adverse effects associated with alcohol vs. marijuana. Rather, some adverse effects were more often associated with alcohol use (e.g. “caused you to behave in ways that you later regretted” [23.7% vs. 5.6%, respectively], “interfered with your ability to think clearly” [15.5% vs. 6.7%, respectively]), while others were more often associated with marijuana use (e.g. “involved you with people you think are a bad influence on you” [12.1% vs. 8.2%, respectively], “caused you to have less energy” [9.3% vs. 5.2%, respectively]). Especially relevant to the current debate, marijuana use was as likely as alcohol use to be associated with no self-reported psychosocial adverse effects (both 13.0%). These findings are consistent with the argument that marijuana is no more dangerous (but also not safer) than alcohol.

The finding of approximately equivalent psychosocial adverse effects from adolescent use of alcohol and marijuana is not consistent with the findings of some previous studies suggesting that alcohol was substantially more harmful than marijuana. For example, two recent UK studies using different types of data and statistical approaches both found alcohol substantially more harmful that cannabis. In a convenience sample of 1501 adult substance users who answered an online survey, respondents ranked alcohol as fifth most harmful among the 21 substances rated, while cannabis was ranked 18th. A panel of British experts, applying a pre-specified statistical model to country-level data, ranked alcohol as fifth most harmful among the 20 substances rated, while marijuana was 11th. There was high overall agreement between the user and expert rankings (r = 0.90, p < 0.001), suggesting good convergent validity for the findings.

The reason for this difference in comparative harmfulness ranking is unclear, but probably is not due to national differences in substance-related adverse effects. This pattern of findings illustrates the importance of relying on converging evidence from different types of studies before drawing firm conclusions that affect public health and public policy. Thus, while the argument that marijuana use is no more harmful than alcohol use is consistently supported by the available evidence, including Palamar et al., the further argument that marijuana use is substantially less harmful than alcohol use is not always supported.

In Palamar et al., adolescents who used both substances also showed no consistent pattern of greater prevalence of adverse effects associated with alcohol use vs. marijuana use, and the specific adverse effects more associated with one substance than the other were similar to those among adolescents who used only one substance. However, all 15 of the alcohol-associated adverse effects and eight of the 15 marijuana-associated adverse effects were significantly more prevalent among users of both alcohol and marijuana than among users of a single substance. These findings may be due to potentiation of adverse effects by the use of two substances and/or differences between users of one substance vs. two substances that make the latter more susceptible to substance-associated adverse effects. In either case, these findings support the need for caution when interpreting the results of observational studies among polysubstance users, especially when making causal inferences.

The findings of this study certainly do not settle the discussion about comparative harms of marijuana vs. alcohol, even within the particular population that was studied. Several limitations of study design, appropriately acknowledged by the authors, preclude such certainty. These include data collection by retrospective self-report with no corroboration by collateral informants or objective evidence, lack of data on several common potential adverse effects such as substance abuse and unprotected sex, lack of data on the magnitude or consequences of the adverse effects (i.e. only data on prevalence were collected), and lack of information on several potentially important confounding variables such as subjects’ socioeconomic status and psychiatric comorbidity. Future studies with more rigorous design features and systematic data collection are needed to fill in these knowledge gaps. Meanwhile, the study by Palamar et al. adds a useful new perspective to the growing body of data on the comparative harmfulness of marijuana use vs. alcohol use, which, to date, has consistently demonstrated that marijuana is no more harmful than alcohol.