This review provides an overview of the changing US epidemiology of cannabis use and associated problems. Adults and adolescents increasingly view cannabis as harmless, and some can use cannabis without harm. However, potential problems include harms from prenatal exposure and unintentional childhood exposure; decline in educational or occupational functioning after early adolescent use, and in adulthood, impaired driving and vehicle crashes; cannabis use disorders (CUD), cannabis withdrawal, and psychiatric comorbidity. Evidence suggests national increases in cannabis potency, prenatal and unintentional childhood exposure; and in adults, increased use, CUD, cannabis-related emergency room visits, and fatal vehicle crashes. Twenty-nine states have medical marijuana laws (MMLs) and of these, 8 have recreational marijuana laws (RMLs). Many studies indicate that MMLs or their specific provisions did not increase adolescent cannabis use. However, the more limited literature suggests that MMLs have led to increased cannabis potency, unintentional childhood exposures, adult cannabis use, and adult CUD. Ecological-level studies suggest that MMLs have led to substitution of cannabis for opioids, and also possibly for psychiatric medications. Much remains to be determined about cannabis trends and the role of MMLs and RMLs in these trends. The public, health professionals, and policy makers would benefit from education about the risks of cannabis use, the increases in such risks, and the role of marijuana laws in these increases.
Cannabis has been used in the United States since the 1800s, with public attitudes toward its acceptability and potential harmfulness varying over time (Musto, 1991). Since 1996, US state laws about the legal use of cannabis for medical and recreational purposes have changed, as have public attitudes about the safety and acceptability of cannabis use. This review aims to provide a broad overview of the epidemiology of cannabis use and associated problems in the United States. The review begins with a section on epidemiologic and public health findings about adverse behavioral, psychosocial, and psychiatric problems associated with cannabis use. These are presented in life-course order of cannabis exposure: prenatal period, childhood, adolescence, and adulthood. The section includes consideration of whether the associations are causal or not (Hall et al, 2016), eg, ruling out reverse causation by clarifying time order (ensuring that the problem does not precede cannabis use) and confounding (controlling for factors that increase the likelihood of both using cannabis and developing the problem). The second section covers changes in the public perception of potential harms/adverse consequences. In the third section, time trends in cannabis use and behavioral, psychosocial, and psychiatric problems are reviewed. The fourth section covers available evidence about the influence of state medical and recreational marijuana laws on the prevalence of cannabis use and associated problems, as well as on substitution of cannabis for opioids, alcohol, and psychiatric medications. Finally, implications for clinicians, policy makers, and the public are considered, and future research directions suggested.
Trends in public perceptions
Inverse Relationship between Perceived Harmfulness of Cannabis and Likelihood of Use
Perceived harmfulness of cannabis has long been considered an important factor preventing adolescent cannabis use (Janz and Becker, 1984; Keyes et al, 2016; Piontek et al, 2013; Schmidt et al, 2016). MTF, a series of annual national school surveys of US adolescents in eighth, tenth, and twelfth grades (13–14, 15–16, and 17–18 years of age; see Table 2 for more information about the MTF surveys) provided the earliest evidence of an inverse relationship between perceived harmfulness and cannabis use (24,
Increasing Public Perception That Cannabis Is Harmless
Among US students in the MTF surveys of students, the perception that regular use of cannabis is risky declined substantially since 2004–2005 (The Monitoring the Future study and the University of Michigan, 2016); rates of decline ranged from 50 to 80%. In 2016, only ∼30% of twelfth graders perceived such risk. The perception that cannabis use is risky also declined among adults. Among NSDUH adult participants, the perception that cannabis use involves great risk declined from 50 to 33% between 2002 and 2014, whereas the perception that use involves no risk increased from ∼6 to 15% (Compton et al, 2016).
Trends in cannabis potency, use, and associated problems
Time Trends in Cannabis Potency
Over the past four decades, cannabis potency, indicated by the THC content in seized samples, has approximately doubled worldwide (Cascini et al, 2012; Hall et al, 2016), including in the United States (ElSohly et al, 2016). In the early 1990s, the average THC content in confiscated US marijuana samples was ~3.7%, whereas in 2014 it was ~6.1% (National Institute on Drug Abuse, 2017). This increase is relevant to epidemiologic study of time trends in problems associated with cannabis use because cannabis with higher THC concentrations is more likely to increase risks associated with use (Englund et al, 2017). However, in states that pass recreational marijuana laws, the potency of legally purchased cannabis may be more relevant to public health (see below).
Time Trends in Prenatal Cannabis Exposure
Among NSDUH participants aged 18–44 years who were pregnant, the estimated prevalence of past-month cannabis use increased 62% between 2002 and 2014, from 2.4 to 3.9% (Brown et al, 2017). In 2014, the rate of past-month cannabis use was highest among pregnant women aged 18–25 years, 7.5% (Brown et al, 2017).
Time Trends in Childhood Exposure
Using data from the National Poison Data System to investigate accidental cannabis exposures in children <6 years old (Onders et al, 2016), the annual rate of exposures increased from 4.2 per million children in 2000 to 10.4 per million in 2013.
Time Trends in Adolescent Cannabis Use
Among NSDUH participants aged 12–17 years, past-month cannabis users fluctuated between 8.2% in 2002 and 7.4% in 2014 (Azofeifa et al, 2016a, b), with no significant overall increases or decreases. Findings from the MTF student surveys also showed that over the same period, rates of cannabis use fluctuated, with some increases up to around 2010, and either stabilization (twelfth graders) or decreases (eighth and tenth graders) from then until 2016 (The Monitoring the Future study and the University of Michigan, 2016). A study of NSDUH participants aged 12–17 years indicated that between 2002 and 2013, the prevalence of CUD decreased (Grucza et al, 2016b), another indicator that cannabis use/problems has not increased among adolescents in recent years.
Time Trends in Adult Cannabis Use, CUD, and Other Indicators of Cannabis-Related Problems
Between 2002 and 2014, past-month cannabis use in NSDUH participants aged 18–25 years increased from 17.3% in 2002 to 19.6% in 2014, a significant overall increase (Azofeifa et al, 2016b), with most of the increase occurring since 2007. Among NSDUH participants 26 years, past-month use also increased significantly between 2002 and 2014 (Azofeifa et al, 2016b), from 4.0 to 6.6% in 2014. In this age group, the increase started in 2008 (Compton et al, 2016). Increases occurred across gender, region, educational level, and employment status (Azofeifa et al, 2016b), with higher rates in males and unemployed participants. Other indicators of adult cannabis use also showed significant increases in NSDUH participants during this period, including first-time use and daily/near-daily use (Compton et al, 2016). Additional studies of NSDUH data showed greater increases in cannabis use among males, especially those with lower incomes (Carliner et al, 2017). However, although daily or near-daily use increased in the entire sample and higher frequency of use is correlated with the risk for a substance use disorder (Compton et al, 2009), NSDUH data did not show increases in the prevalence of adult CUD (Compton et al, 2016) between 2002 and 2014.
Another source of data on adult trends were surveys conducted by the National Institute on Alcohol Abuse and Alcoholism (NIAAA): the NESARC, conducted in 2001–2002, and the NESARC-III, conducted in 2012–2013 (see Table 2 for more information about the NIAAA surveys). The past-year prevalence of cannabis use in 2001–2002 and 2012–2013 was 4.1% and 9.5%, respectively, a significant increase (Hasin et al, 2015a, 2017). Significant increases in use were also found across demographic subgroups (sex, age, race/ethnicity, education, marital status, income, urban/rural, and region). The past-year prevalence of DSM-IV CUD was 1.5% in 2001–2002 and 2.9% in 2012–2013, also a significant increase (Hasin et al, 2015a, 2017). Increases in the prevalence of CUD between 2001–2002 and 2012–2013 were also statistically significant across demographic subgroups.
The NSDUH and NESARC surveys were consistent in showing increases in adult cannabis use. However, their findings differed regarding time trends in the prevalence of CUD. Methodological differences between the surveys have been discussed as one possible explanation of the discrepant findings (Compton et al, 2016; Grucza et al, 2016a, c). Another approach to understanding the differences is to determine the consistency of the findings of these surveys with other national indicators of cannabis-related problems or CUD over the same time period (Hasin and Grant, 2016). The NESARC to NESARC-III findings on increases in the prevalence of cannabis use disorders are consistent with these other national trends (see next paragraph), suggesting that the NESARC findings on increased rates of CUD are valid (Hasin and Grant, 2016).
Information on increasing time trends in rates of cannabis-related problems and CUD is available from a wide variety of large-scale data sources, many of them national in scope (Table 3). These show increasing prevalence of ICD-9-CM cannabis use disorders among patients seen in emergency rooms (Zhu and Wu, 2016), hospital inpatients (Charilaou et al, 2017; Shi, 2017), gastroenterology patients (Gubatan et al, 2016), burn treatment patients (Jehle et al, 2015), and in United States veterans treated in the Veterans Health Administration (Bonn-Miller et al, 2012). Of note, the burn treatment patients testing positive for cannabis were younger, less likely to be insured, had larger burns, and required more intensive treatment. Rates of cannabis detected among decedents of fatal motor vehicle (Brady and Li, 2014) and airplane crashes (McKay and Groff, 2016) also increased, as did rates of cannabis detected in roadside surveys of weekend nighttime drivers (Johnson et al, 2012).
Medical and recreational marijuana laws
Medical Marijuana Laws
In 1970, the Federal Drug Enforcement Agency (DEA) defined cannabis as a Schedule 1 substance, meaning no accepted medical use and high abuse potential (U.S. Department of Justice and Drug Enforcement Administration, 2017). However, since 1996, when California passed the first state medical marijuana law (MML), a total of 29 states and the District of Columbia have passed laws legalizing the use of marijuana for medical purposes (Figure 2). State MMLs share the common feature that they permit legal use of cannabis to treat medical conditions if the user has obtained medical authorization. However, the specific provisions of MMLs vary considerably (Pacula et al, 2015), eg, regarding the range and specificity of the permitted medical conditions, the provisions for distribution through dispensaries, permitted amounts per patient, and so on. The restrictiveness or ‘medicalization (238,
Concerns about MMLs include their potential to increase problematic use of cannabis in the general population, hypothesized to occur through several mechanisms, including reduced perceived harmfulness and normalization of use (Pacula et al, 2015; Wen et al, 2015). Additional posited mechanisms include greater access via dispensaries and home or caregiver cultivation. Greater access increases availability, and can also normalize the idea of use and reduce perceived harm (Pacula et al, 2015; Wen et al, 2015). Table 5 shows that the percentage of Americans living in MML states has increased over time. Because early and later MMLs were passed in differing national normative contexts, their effects may vary over time (Hasin et al, 2017).
Before 2009, the discrepancy between the Federal and state positions meant that individuals using marijuana as specified by state MMLs were still vulnerable to federal prosecution. However, in 2009, the US Attorney General issued a memorandum instructing federal prosecutors not to prioritize prosecution of individuals compliant with state MMLs (Ogden, 2009). This gave more flexibility to MML states (Cambron et al, 2017), with particular impact in Colorado (Davis et al, 2016b; Hasin et al, 2017; Salomonsen-Sautel et al, 2014; Schuermeyer et al, 2014) and California (Hasin et al, 2017), where dispensaries proliferated. Subsequent memorandums in 2011 and 2013 further clarified these policies by indicating which areas of marijuana law enforcement were still of Federal interest (eg, limiting use among minors; Price 2014).
In states with unrestrictive MMLs (generally states with earlier-passed MMLs), the possibility that medical authorizations were obtained for recreational use is supported by the greater similarity of some medical users to recreational users than to a medically ill population (Harris et al, 2000; Haug et al, 2017; Reinarman et al, 2011; Walsh et al, 2013), and because many medical users used illicit cannabis before their medical authorization (O'Connell and Bou-Matar, 2007). However, two reports on medical marijuana users among adult NSDUH participants provide more representative information. The first (Lin et al, 2016), reporting only on users from MML states, indicated that among past-year cannabis users, 17% used cannabis medically. Medical and recreational users did not differ on race, education, depression, or cannabis use disorders, but medical users were more likely to report poor health and fewer substance use disorders. The second (Compton et al, 2017) reported on all medical cannabis users in all states, as some could use cannabis for medical purposes even if they did not live in a MML state. In this study, 9.8% of all cannabis users were medical users. Their medical use was associated with poor self-reported health, disability, older age, and older age at initiation of cannabis use. Of all medical users, 21.2% lived in states with no MMLs, suggesting that either physicians in non-MML states also recommend medical marijuana use to some patients, or some patients in non-MML states were self-medicating problems with cannabis.
Recreational Marijuana Laws
In 2012, Colorado and Washington became the first states to pass recreational marijuana laws (RMLs). Since then (Figure 2), 6 additional states passed RMLs: Alaska and Oregon (2014), and California, Nevada, Massachusetts, and Maine (2016). Consequently, 21% of the US population now lives in states where recreational use is legal. All eight of the states that passed recreational marijuana laws previously had MML.
RMLs permit legal sale and use of cannabis without the need for medical justification or authorization. Such laws may reduce discriminatory arrests of disadvantaged minorities because of biased enforcement of criminal cannabis statutes (Hall and Lynskey, 2016; Palamar et al, 2014), and satisfy public desire for legal cannabis use. Furthermore, RMLs have the potential to create business opportunities, jobs, and tax revenues (Forbes, 2017; 152,
Relationship of marijuana laws to trends in use and consequences
Marijuana Laws and Trends in Cannabis Potency
Among samples of illegal cannabis seized by law enforcement between 1990 and 2010 (Sevigny et al, 2014) analyzed by state, the mean THC potency was 9.1% in states that eventually passed MMLs and 5.6% in other states. Initial models of state effects on potency found a MML effect on subsequent cannabis potency, but this effect was no longer significant after controlling for a multitude of potential confounders. However, in states with legally operating dispensaries, the MML effect remained significant, increasing THC potency of ∼1%.
As states increasingly pass recreational marijuana laws, the THC potency of illicit cannabis in MML states may no longer be such a salient issue. In Washington (RML passed in 2012), the average THC potency of marijuana flower for one Seattle-based retailer in 2015 was 21.2% (Washington State Marijuana Impact Report, 2016). In Colorado, where THC potency of legally marketed cannabis can range considerably, some strains have THC potencies of 28–32% (CNN, 2016). In 2016, Colorado legislators attempted to limit THC content of marketed cannabis to 16%, but this attempt failed, indicating local support for continued marketing of stronger forms of cannabis. Furthermore, although smoking remains the most common route of administration (Hall et al, 2016), newly popular routes of administration offer even higher THC doses (Hall et al, 2016). These include edibles, vaping (inhaled vapor of heated e-liquids analogous to e-cigarettes), and dabbing (inhaled vapor from heating highly concentrated forms of cannabis or hashish). The health effects of these methods are not yet known, but warrant monitoring (National Institute on Drug Abuse, 2017).
Marijuana Laws and Trends in Prenatal Exposure
No published studies were found on differences in rates of cannabis use by MML or RML state status. Unpublished results from one study (Brown et al, 2017) showed that in MML and non-MML states, rates of cannabis use in pregnant women did not differ significantly.
Marijuana Laws and Trends in Childhood Exposure
A study of 2005–2011 calls to poison centers for unintentional pediatric cannabis exposures showed little increase in states with no MML, an increase of 11.5% in states passing MMLs in 2005–2011, and an increase of 30.3% in states passing MMLs before 2005 (Wang et al, 2014). In this study, among children unintentionally exposed to cannabis, those in states with MMLs passed before 2005 were more likely to be evaluated in a health-care facility, experience major or moderate effects, and be admitted to critical care units (Wang et al, 2014). Another study found that calls to Colorado poison control centers for unintentional pediatric cannabis exposures between 2009 and 2015 (when medical dispensaries proliferated and recreational use was legalized) increased 34% annually, significantly greater than the 19% increase in cannabis-related calls received by poison control centers in other states (Wang et al, 2016a). A further breakdown of Colorado poison center calls (Wang et al, 2017) showed significant increases in those aged 0–8 and 9–17 years after liberalization of medical marijuana laws in 2010, and an even further increase in these ages after enactment of legalized recreational use in 2014. Thus, liberalization of marijuana laws appears related to increases in childhood nonfatal cannabis exposures.
Marijuana Laws and Trends in Adolescent Cannabis Use
Because of the potential harms of early adolescent cannabis use, concerns were raised that MMLs could increase adolescent use by conveying a message about acceptability or lack of negative health consequences, even if MML implementation was delayed or narrowly defined (Hasin et al, 2015b). A study using cross-sectional national data showed that adolescent cannabis use was more prevalent in MML states (Wall et al, 2011). However, cross-sectional associations do not necessarily indicate a causal relationship, as states with MMLs could have had higher rates of use before MML passage (ie, reverse causation). Although randomized assignment of MMLs to states would provide clarity about the direction of effect, such assignment is clearly not possible. Therefore, studies used difference-in-difference (DiD) tests to examine differences in rates in states before and after MML passage compared with contemporaneous differences during the same years in non-MML states (Angrist and Krueger, 1999; Angrist and Pischke, 2009; Hunt and Miles, 2015; Imbens and Wooldridge, 2009). These studies also controlled for many state- and individual-level confounders to sharpen conclusions on causality. Studies used MTF data from 1991 to 2014 (Hasin et al, 2015b), NSDUH data from 2002 to 2012 (Wen et al, 2015), and several other large databases (Sarvet et al, under review-a). Of the 17 large surveys using DiD methods spanning different states, periods, and specifications, 16 indicated no MML effects on adolescent use (Anderson et al, 2015; Choi, 2014; Hasin et al, 2015b; Keyes et al, 2016; Martins et al, 2016; Pacula et al, 2015; Sarvet et al, under review-a; Smart, 2015; Wen et al, 2015). Despite differences in methodology, the findings were very consistent: post-MML adolescent cannabis use did not increase compared with pre-MML levels and to national trends in non-MML states during the corresponding years. In MTF data, perceived harmfulness of marijuana use decreased and marijuana use increased following legalization of recreational marijuana use in Washington State in 2012 whereas, in contrast Colorado did not exhibit any differential change in perceived harmfulness or past-month adolescent marijuana use following legalization (Cerda et al, 2017).
Marijuana Laws and Trends in Adult Cannabis Use and Cannabis Use Disorders
Fewer studies are available on the relationship of MMLs to adult cannabis outcomes. A cross-sectional study using national 2004–2005 data showed an association between living in a MML state with adult cannabis use and cannabis disorders (Cerda et al, 2012). However, this design also did not address causality. DiD tests were therefore used to examine adult outcomes. A 15–20% post-MML increase occurred in marijuana possession arrests in major cities (Chu, 2014), and a 20% post-MML increase occurred in first-time adult marijuana admissions (Chu, 2014). Using NSDUH 2004–2013 data (Table 2) at a point when 10 states had passed MMLs, DiD tests indicated significant post-MML increases for cannabis use, daily or near-daily use, and (with 1- and 2-year time lags) CUD (Wen et al, 2015), an effect that remained for cannabis use among adults aged 26 years in a reanalysis (Martins et al, 2016). Using DiD tests and three NIAAA surveys from 1991–1992 to 2012–2013 (Table 2) to compare 15 MML states with other states, post-MML effects were found on adult cannabis use and CUD (Hasin et al, 2017). Between 1991–1992 and 2001–2002, rates of adult use and CUD generally decreased, but in contrast, rates of adult use and CUD increased in early-MML states. Between 2001–2002 and 2012–2013, rates generally increased, with greater increases in late-MML than non-MML states. During this later time, California and Colorado showed marked increases, consistent with the proliferation of dispensaries in these two states after the 2009 Department of Justice Memo. Thus, although the research base is not extensive, existing studies are consistent in showing post-MML increases in adult cannabis-related outcomes.
Recreational Marijuana Laws
Aside from Cerda et al (2017) and Wang et al (2017) who suggest that RMLs may have further effects on child exposures or adolescent use in states that already have MMLs, little is known about the effects of recreational marijuana laws on cannabis use, or cannabis-related consequences. To our knowledge, no studies have yet been published on the relationship of RML to adult outcomes, an area that is clearly in need of studies as data become available.
Epidemiologic evidence on mmls and cannabis used as a substitute for opioids or alcohol
Cannabis to Treat Pain and as a Substitute for Opioids
Chronic pain is common in U.S. adults (Hardt et al, 2008; Institute of Medicine, 2011; Nahin, 2015; Tsang et al, 2008). Opioids are important to treat acute pain, but are also widely prescribed for chronic pain (Levy et al, 2015; Volkow and McLellan, 2016) with inconsistent clinical benefits (Chou et al, 2014) and serious risks (Chou et al, 2014; Volkow and McLellan, 2016), eg, physical dependence, addiction, transition to heroin, and overdose (Okie, 2010; Paulozzi et al, 2012; Rudd et al, 2016). Therefore, despite problems associated with medical marijuana (cognitive/motor impairments (Volkow et al, 2016), side effects (Whiting et al, 2015), no standard product formulations (Thomas and Pollard, 2016)), its advantages (analgesia (Whiting et al, 2015), lack of fatal overdose (Hall, 2017), or transition to heroin) have led to professional calls to substitute medical marijuana for opioids (Choo et al, 2016), although this debate continues (Saxon and Browne, 2014). Many medical marijuana patients use it for pain relief, some as a partial or complete substitute for opioids (Boehnke et al, 2016; Davis et al, 2016a; Lucas and Walsh, 2017; Lucas et al, 2016; Nunberg et al, 2011; Piper et al, 2017; Reinarman et al, 2011), and others continuing to use or abuse prescription opioids. In an online convenience sample, participants in both MML and non-MML states used cannabis for pain (Corroon et al, 2017). Thus, studies of adult substitution/complementarity of cannabis and opioids are needed, including whether this occurs more in MML and RML states.
Several ecological studies of MML and rates of opioid outcomes have been conducted, based on the premise that if MMLs provide marijuana to those who need it, opioid use/misuse will be reduced. The studies found that MMLs led to lower rates of opioid prescriptions in Medicaid (Bradford and Bradford, 2017) and Medicare Part D (Bradford and Bradford, 2016a). MMLs also led to lower rates of opioid overdoses (Bachhuber et al, 2014; Pardo, 2016), although according to Bachhuber et al (2014) significance was lost after controlling for state-specific linear time trends that adjusted for differential factors changing linearly over the study period (eg, hard-to-measure attitudes or cultural changes). MMLs were also associated with decreased hospitalization for OUD (Powell et al, 2015; Shi, 2017) and opioids detected in fatally injured drivers (Kim et al, 2016) (but of all ages studied, only in those aged 21–40 years). In the single individual-level study (NSDUH 2004–2012), MML was unrelated to non-medical opioid or heroin use (Wen et al, 2015). Studies are thus largely but not fully consistent. The field concurs that more individual-level studies of MML and opioids are needed, controlling for important individual and state-level covariates (Bradford and Bradford, 2016b, 2017; Finney et al, 2015; Hall and Lynskey, 2016). Opioid outcomes should include medical use, ie, prescriptions; benzodiazepine co-prescriptions (shown repeatedly (Garg et al, 2017; Hawkins et al, 2013; Hwang et al, 2016; Sun et al, 2017) to be very risky for overdose, and common in some groups, eg, veterans (Hawkins et al, 2015)); frequent non-medical use; and consequences of medical or nonmedical use: OUD, DUI, overdose. Studies should include state-level demographic characteristics, MML provisions, and opioid policies. Studies should also determine whether MML effects remain after controlling for important individual demographic and clinical characteristics, eg, pain or pain-related medical conditions, and other substance and psychiatric disorders (Hayes and Brown, 2014).
Adults and Alcohol
Some medical marijuana clients substitute cannabis for alcohol (Lucas and Walsh, 2017; Lucas et al, 2016; Nunberg et al, 2011; Reiman, 2009; Reinarman et al, 2011). In studies that used data from 1990 to 2010, passage of MMLs was followed by decreased alcohol-related traffic fatalities (Anderson et al, 2013) and binge drinking (Anderson et al, 2013), suggesting that cannabis was substituted for alcohol in these states. However, in NSDUH adults surveyed between 2004 and 2012, passage of MMLs was followed by increased binge drinking (Anderson et al, 2013; Guttmannova et al, 2016; Wen et al, 2015). An unpublished economics report showed that passage of MML was followed by increases in any drinking and alcohol-related treatment admissions, but only if the MMLs permitted dispensaries (Pacula et al, 2013). A state-level study using data from 1985 to 2014 showed that passage of MMLs was followed by decreased overall US traffic fatalities (Santaella-Tenorio et al, 2017), particularly in ages 25–44 years, but only in 7 states, and therefore the results were interpreted as indicating heterogeneous effects. Thus, findings on MMLs, alcohol, and cannabis have been inconsistent (Guttmannova et al, 2016), perhaps because of analyses of different outcomes and state-level control variables in the different studies (Guttmannova et al, 2016), leading to the need for further exploration in designs that could clarify the sources of inconsistent results.
Speculation about RML effects largely assume that substituting cannabis for alcohol will be better for public health (Anderson and Rees, 2014; Carnevale et al, 2017). Some experts have assumed that RMLs will lead to substitution of cannabis for alcohol (Anderson and Rees, 2014; Kilmer, 2017), whereas others have been less sure (Edwards, 1974; Hall and Lynskey, 2016; Hawken et al, 2013; Pacula and Sevigny, 2014). Whether cannabis will actually be substituted for alcohol after passage of RMLs, and whether the effects of any such substitution on public health will be positive, negative, or neutral is currently unknown.
Epidemiologic evidence on marijuana laws and cannabis used as a substitute for psychiatric medication
Personal and anecdotal testimonies suggest that cannabis (medical or otherwise) is effective in treating symptoms of depression and anxiety (Broadly, 2016; Grass City, 2010). Consistent with this, surveys of medical marijuana patients show that many of them use marijuana to treat these symptoms (Bohnert et al, 2014; Bonn-Miller et al, 2014; Harris et al, 2000; Nunberg et al, 2011; Piper et al, 2017; Reinarman et al, 2011; Walsh et al, 2017). Although theoretical reasons suggest that synthetic oral cannabinoids may be helpful for some aspects of PTSD (Haney and Evins, 2016), scientific reviews (National Academies of Sciences, 2017; Walsh et al, 2017; Whiting et al, 2015) of studies to date find no evidence for the efficacy of cannabinoids in the treatment of depression or anxiety disorders. In addition, prospective studies show adverse cannabis effects on the course of depression (Bahorik et al, 2017) and PTSD (Wilkinson et al, 2015). When medical marijuana clients are asked about actual symptom relief, less than half report such relief (Bonn-Miller et al, 2014); other medical marijuana clients (Swift et al, 2005) report return of anxiety symptoms on cessation of use, suggesting the symptoms might be due to cannabis withdrawal (Walsh et al, 2017). Because many cannabis withdrawal criteria are depression/anxiety symptoms (American Psychiatric Association, 2013) (Table 1), regular users may seek cannabis to obtain short-term symptom relief, unaware that this could perpetuate a longer-term withdrawal problem. Nevertheless, medical marijuana is authorized for PTSD in 21 states, and in many others, permitted conditions are vague enough that use for depression or anxiety may also be authorized. In nationally representative 2004–2005 data, participants with DSM-IV depressive or anxiety disorders were more likely to self-medicate their symptoms with cannabis if they lived in MML than non-MML states (Sarvet et al, under review-b). However, a more recent survey of a convenience sample suggested that self-medication of depression or anxiety with cannabis is equally common in MML and non-MML states (Corroon et al, 2017). Without recent data from a representative sample, whether the more recent convenience survey reflects a change or a biased result due to the sampling method remains unknown.
In data on all fee-for-service Medicaid prescriptions from 2007 to 2014, antidepressant and anti-anxiety prescriptions were lower in states with MMLs than in other states (Bradford and Bradford, 2017). In 2010–2013 data on prescriptions filled by Medicare Part D enrollees, antidepressant prescriptions fell significantly in MML states once a medical marijuana law was implemented (Bradford and Bradford, 2016a). If cannabis/cannabinoid products were effective treatments for depression or anxiety disorders, substituting cannabis for FDA-approved medication would be a positive MML result (eg, by reducing medical costs (Bradford and Bradford, 2016b)). As evidence on efficacy suggests otherwise (National Academies of Sciences, 2017; Walsh et al, 2017; Whiting et al, 2015), and as marijuana use may be due to confusion between cannabis withdrawal and depressive/anxiety disorders (Swift et al, 2005), such shifts in clinical care would be an adverse MML outcome. However, to understand the effects of MML and RML on use of antidepressants and antianxiety medication, large-scale studies using individual-level data (Bradford and Bradford, 2017) are needed that can include relevant state socioeconomic and policy variables, and individual-level demographic and clinical covariates.
Implications and future research directions
The implications of changing laws, attitudes, and prevalence of cannabis use have implications for clinicians, policy makers, and the public. All should be aware that despite a lack of risk for fatal overdose or transition to heroin, both of which are serious risks for prescription opioids, cannabis is not a harmless substance, and use can involve impairments, addiction, and risks for serious consequences. In NESARC 2001–2002 data, 6.4% of those with current DSM-IV cannabis abuse and 18.1% of those with current DSM-IV cannabis dependence received any kind of intervention for drug use problems (Stinson et al, 2006) and of those with lifetime DSM-IV cannabis abuse or dependence, 9.8 and 34.7% received any type of intervention. Thus, drug treatment among those with cannabis use disorders was rare, and whether the treatment focused on cannabis use disorders is unknown. NESARC-III 2012–2013 data showed that among those with DSM-5 current and lifetime cannabis use disorders, 7.2 and 13.7% received any type of intervention specifically for cannabis problems (Hasin et al, 2016). Furthermore, despite the clear increases in adult cannabis use and related problems in the general population, the proportion of patients in substance abuse treatment whose primary substance was cannabis was 16% in 2003 and 17% in 2013 (Substance Abuse and Mental Health Services Administration and Center for Behavioral Health Statistics and Quality, 2015), showing no increase. Thus, cannabis use disorders remain seriously undertreated in the US general population.
Mental health clinicians, especially those treating younger patients or patients with affective or anxiety symptoms, should consider screening their patients for cannabis use patterns and cannabis use disorder criteria to determine whether a disorder is present, or explore with patients whether a cannabis use/withdrawal cycle may be perpetuating depressive or anxiety symptoms. Although patients may not initially be receptive to the idea that cannabis is causing or contributing to their symptoms or problems rather than alleviating them, continued discussion and some monitoring may help in this regard. Health policy makers, for example, commissioners of state or city mental health or substance abuse services should also be aware of the national trends toward increasing use and consequences, and encourage system-wide awareness of current information on cannabis use and its consequences across all professional levels of service providers. State legislators, when considering passage or modifications of medical or recreational marijuana laws, should consider increases in the occurrence of potential adverse consequences of increasingly widespread use, and design state policies for distributing, advertising, and taxing cannabis with these risks in mind. Finally, the public should be made aware of the risks as well. Public education campaigns based on exaggerated scare tactics are unlikely to be successful. However, public education efforts can be effective in such areas as reduced rates of drunk-driving fatal accidents (Niederdeppe et al, 2017a). Well-designed, evidence-based programs may change public attitudes in a positive, more health-promoting direction on many health policy issues involving substance use (Bachhuber et al, 2015; Niederdeppe et al, 2017b).
This review has identified a number of areas needing further research. The following are open areas of research that could be investigated using epidemiologic designs.
Studies on the relationship of medical and recreational marijuana laws to indicators of driving under the influence of cannabis, or of other substances.
Studies of postnatal outcomes among women using marijuana but not other substances while pregnant that may be more possible now than in previous years if women who would not consider drinking alcohol or smoking cigarettes see marijuana as a harmless way to treat pregnancy-related symptoms (The New York Times, 2017).
Long-term comparisons of cognitive functioning in cannabis users and nonusers, with observation begun in childhood, before cannabis exposure.
Studies of the relationship between cannabis withdrawal symptoms, diagnoses of depressive and anxiety disorders, self-medication with cannabis, and utilization of psychiatric medication.
Studies examining the effectiveness of various interventions aimed at increasing public awareness of cannabis risks and their likelihood.
Continued monitoring of time trends in cannabis use and consequences, overall and by demographic characteristics.
Continued studies of the relationship of MML to cannabis use and consequences, and to the use of and consequences of other substances, including opioids.
Studies of the relationship of RML to cannabis use and consequences, and to the use of and consequences of other substances, including opioids.
In the past 10 years, much has been learned about cannabis, its use, and its consequences. However, in this time of rapidly changing marijuana laws and attitudes, much remains to be learned in order to advance public health and to guide personal and societal decisions regarding the use of cannabis.
Funding and disclosure
The author declares no conflict of interest.
Lili Sar-Graycar and Brittney Davis are thanked for their editorial assistance. This work was supported by a grant from the National Institute on Drug Abuse (R01DA034244) and the New York State Psychiatric Institute.