The health and social effects of nonmedical cannabis use

6.1 What do we know?

The adverse psychosocial and mental health outcomes that are correlated with long-term cannabis use are most often seen in daily or near-daily users. This section of the document summarizes evidence on the best researched of these health outcomes – namely dependence, educational outcomes, the use of other illicit drugs, cognitive impairment, mental disorders (psychoses, depression and other disorders) and suicidality (risk, ideation, attempts and mortality).

6.1.1 Long-term cannabis use and dependence

Cannabis dependence is a cluster of behavioural, cognitive and physiological phenomena that develop after repeated cannabis use. A diagnosis of dependence requires that three or more of the following criteria are met in the previous year:

(a) strong desire or sense of compulsion to take the substance;

(b) difficulties in controlling substance-taking behaviour in terms of its onset, termination, or levels of use;

(c) a physiological withdrawal state (see F1x.3 and F1x.4) when substance use has ceased or been reduced, as evidenced by: the characteristic withdrawal syndrome for the substance; or use of the same (or a closely related) substance with the intention of relieving or avoiding withdrawal symptoms;

(d) evidence of tolerance, such that increased doses of the psychoactive substances are required in order to achieve effects originally produced by lower doses (clear examples of this are found in alcohol- and opiate-dependent individuals who may take daily doses sufficient to incapacitate or kill nontolerant users);

(e) progressive neglect of alternative pleasures or interests because of psychoactive substance use, increased amount of time necessary to obtain or take the substance or to recover from its effects;

(f) persisting with substance use despite clear evidence of overtly harmful consequences, such as harm to the liver through excessive drinking, depressive mood states consequent to periods of heavy substance use, or drug-related impairment of cognitive functioning; efforts should be made to determine that the user was actually, or could be expected to be, aware of the nature and extent of the harm” (WHO, 1992).

Harmful use of cannabis and cannabis dependence are the most common forms of drug-use disorders in epidemiological surveys in Australia, Canada and the USA. These disorders affect 1-2% of adults in the past year, and 4-8% of adults during their lifetime (Hall & Pacula, 2010; Anthony, 2006). As noted, the risk of dependence has been estimated at 16% in those who initiated cannabis use in adolescence (Anthony, 2006) and 33-50% in daily cannabis users (van der Pol et al., 2013). We do not know how these risk estimates from the early 1990s may have been affected by changes in diagnostic criteria for dependence in DSM-5 or by changes in the potency of cannabis products. However, based on DSM-IV and the large representative USA NESARC study, higher proportions of lifetime users seem to have developed cannabis use disorders (Lev-Ran et al., 2013; Fischer et al., 2015), and nearly 3 of 10 cannabis users in the USA manifested a cannabis-use disorder in 2012-2013 ( Hasin et al., 2015).

Humans develop tolerance to THC (Lichtman & Martin, 2005) and cannabis users who seek help for cannabis-use problems often report withdrawal symptoms such as anxiety, insomnia, appetite disturbance and depression (Budney & Hughes, 2006). These symptoms are of sufficient severity to impair everyday functioning (Allsop et al., 2012) and they are markedly attenuated by doses of an oral cannabis extract (Sativex) that contains THC (Allsop et al., 2014).

Cannabis dependence in and of itself is not the only problem for heavy users. By increasing the duration of regular use, dependence may also increase the risk of any long-term health risks of cannabis that may occur after decades of use, such as cardiovascular and respiratory diseases, and possibly cancers. These risks are discussed in chapter seven of the report.

The mortality of patients with cannabis dependence is also of concern. A study of 46 548 individuals hospitalized in California between 1990 and 2005 with ICD-9 diagnoses of cannabis dependence and cannabis abuse were followed for 16 years. Age-, sex- and race-adjusted standardized mortality rates (SMRs) were generated. Out of the total cohort of people with cannabis-use disorder diagnosis, 1809 deaths across all years were identified (Callaghan et al., 2012). This is an approximately four-fold higher risk of mortality when compared with that of the general population. The underlying reasons for the elevated standardized mortality rates in the cannabis cohort are unknown.

6.1.2 Long-term cannabis use and cognitive function

The 1990s case-control studies found that regular cannabis users had poorer cognitive performance than non-cannabis-using controls (Hall, Solowij & Lemon, 1994). The challenge was to decide whether cannabis use impaired cognitive performance, or if persons with poorer cognitive functioning were more likely to become regular cannabis users, or both (Hall, Solowij & Lemon, 1994). Better-controlled case-control studies since then (Crane et al., 2013; Solowij & Battisti, 2008; Grant et al., 2003; Schreiner & Dunne, 2012) have consistently found deficits in verbal learning, memory and attention in regular cannabis users (see section 5.1.2). These deficits have usually been correlated with the duration and frequency of cannabis use, the age of initiation and the estimated cumulative dose of THC (Solowij, 2002; Solowij & Pesa, 2012; Solowij et al., 2011). It remains unclear whether cognitive function fully recovers after cessation of cannabis use, with studies producing conflicting results (Solowij, 2002; Solowij & Pesa, 2012).

A longitudinal study from the Dunedin birth cohort suggested that sustained heavy cannabis use over several decades produced substantial declines in cognitive performance that may not be wholly reversible. This study assessed changes in IQ between age 13 (before cannabis was used) and at age 38 in 1037 New Zealanders born in 1972 or 1973 (Meier et al., 2012). Early and persistent cannabis users showed an average decline of eight IQ points compared with peers who had not used cannabis, and cannabis-using peers who had not used cannabis in this sustained way. Rogeberg (2013) argued that the apparent effect of sustained cannabis use on IQ could be due to failure to control for socioeconomic status. Further analysis of the Dunedin data did not support Rogeberg’s hypothesis (Moffitt et al., 2013). A recent study in the USA has provided support for the study of Meier et al. in finding an association between poorer verbal memory and sustained daily use of cannabis throughout adult life (Auer et al., 2005).

As noted in section 4.1, studies of brain structure and function in cannabis users provide some support for these epidemiological findings. MRI studies have reported structural changes in the hippocampus, prefrontal cortex and cerebellum in chronic cannabis users (Yücel et al., 2008) and these were largest in persons who had used cannabis the longest. A recent systematic review (Lorenzetti et al., 2013) found a consistent reduction in hippocampal volume in long-term daily users.

Excluding the possibility of reverse causation as an explanation for these findings has been difficult because younger persons with poorer cognitive performance are more likely to become regular cannabis users. There are also shared risk factors for regular cannabis use and poor cognitive performance. A causal role for regular cannabis use has biological plausibility in that cannabis acutely impairs cognitive performance, and neuroimaging studies have found relationships between the frequency and duration of cannabis use and structural and functional changes in brain regions implicated in memory and cognition.

6.1.3 Long-term psychosocial consequences of adolescent cannabis use

6.1.3.1 Social and educational outcomes

Longitudinal studies since the 1990s have found that cannabis use before the age of 15 years predicts early school-leaving and this persists after adjustment for confounders (e.g. (Ellickson et al., 1998)). A meta-analysis of three Australian and New Zealand longitudinal studies (Horwood et al., 2010) confirmed this finding. Longitudinal studies have also shown that early initiation of heavy cannabis use is associated with lower income, lower college degree completion, a greater need for economic assistance, unemployment, and use of other drugs (Fergusson et al., 2016; Fergusson & Boden, 2008; Brook et al., 2013).

It is plausible that educational outcomes in regular cannabis users are impaired for a combination of reasons: a higher pre-existing risk of educational problems in those who become regular cannabis users, the adverse effects of regular cannabis use on learning in school, increased affiliation of regular cannabis users with other cannabis-using peers who reject school, and the strong desire of younger cannabis users to make a premature transition to adulthood by leaving school (Lynskey & Hall, 2000).

A recent Australian twin study has raised doubts about a causal interpretation of the association between adolescent cannabis use and early school-leaving (Verweij et al., 2013). The study found that the association between early cannabis use and early school-leaving was explained by shared genetic and environmental risk factors. These findings have been supported by two twin studies in the USA (Grant et al., 2012; Bergen et al., 2008), which suggest that the association may be explained by higher levels of recruitment to cannabis use among adolescents who are at higher risk of leaving school earlier.

In an earlier Australian, study early-onset users had significantly higher rates of later substance use, juvenile offending, mental health problems, unemployment and school dropout. The links between early-onset cannabis use and later outcomes were largely explained by two routes that linked cannabis use to later adjustment. First, those electing to use cannabis were a high-risk population characterized by social disadvantage, childhood adversity, early-onset behavioural difficulties and adverse peer affiliations. Secondly, early-onset cannabis use was associated with subsequent affiliations with delinquent and substance-using peers, moving away from home and dropping out of education, with these factors in turn being associated with increased psychosocial risk (Fergusson et al., 1997). A substantial proportion of those who become cannabis users continued to smoke tobacco and use alcohol in a harmful or hazardous way and they were more likely to use a range of other illicit drugs (Hasin et al., 2015).

6.1.3.2 Other illicit drug use

Epidemiological studies in Australia, New Zealand and the USA in the 1970s and 1980s found that regular cannabis users were more likely to use heroin and cocaine, and that the younger they were when they first used cannabis the more likely they were to use the other drugs (Kandel, 2002). Three explanations were offered for these patterns: (a) that cannabis users had more opportunities to use other illicit drugs because these were supplied by the same black market as cannabis; (b) that early cannabis users were more likely to use other illicit drugs for reasons that were unrelated to their cannabis use (e.g. their propensity to take risks, behave impulsively, or engage in sensation-seeking); and (c) that the pharmacological effects of cannabis increased a young person’s interest in using other illicit drugs (Hall & Pacula, 2010).

Patterns of drug involvement similar to those in the USA have been reported in a number of countries by epidemiological research (Swift et al., 2012), although the order in which drugs are used varies with the prevalence of different illicit drugs among adults (Degenhardt et al., 2010). Research has also supported the first two hypotheses in that young people in the USA who have used cannabis report more opportunities to use cocaine at an earlier age (Wagner & Anthony, 2002). Additionally, socially deviant young people (who are also more likely to use cocaine and heroin) start using cannabis at an earlier age than their peers (Fergusson, Boden & Horwood, 2008).

Simulations suggest that shared risk factors could explain these relationships between cannabis and other illicit drug use (Morral, McCaffrey & Paddock, 2002). The shared risk factor hypothesis has been tested in longitudinal studies by assessing whether cannabis users are more likely to report heroin and cocaine use after statistically controlling for confounding factors (Lessem et al., 2006; Fergusson, Boden & Horwood, 2006). Adjustment for confounders has reduced but not eliminated the relationship (Hall & Lynskey, 2005).

Studies of twins who are discordant for cannabis use (i.e. one used cannabis and the other did not) have been used to test whether shared genetic vulnerability explains the higher rates of illicit drug use among heavy cannabis users. Lynskey and colleagues (2003) found that the twin who had used cannabis prior to age 17 was more likely to have used other illicit drugs than the co-twin who had not. This relationship persisted after controlling for non-shared environmental factors. Similar results have been reported in discordant twin studies in the USA (Grant et al., 2010) and Netherlands (Lynskey, Vink & Boomsma, 2006).

Preclinical studies of early adolescent exposure to THC in rodents are supportive of these findings. Adult rats pre-treated with THC during adolescence and then allowed to mature to adults without THC are more likely to use heroin than rats not exposed to cannabis during adolescence. The endogenous opioid system was also disturbed in the brain of adults exposed to THC during adolescence (Ellgren, Spano & Hurd, 2007; Ellgren, 2008; Tomasiewicz et al., 2012).

6.1.3.3 Tobacco and alcohol use

In the early 1990s, cigarette smoking in many developed countries generally started before cannabis use, and regular tobacco smoking was a predictor of regular cannabis use and was regarded as a gateway to cannabis use. Over the past 20 years the relationship between cannabis and tobacco use has changed in some developed countries with a low prevalence of cigarette smoking and a high prevalence of cannabis use. In Australia and the USA, as a result of public health campaigns to prevent tobacco smoking among young people, young people increasingly start cannabis smoking before they smoke tobacco (Johnston et al., 2010). In these countries cannabis use increases the risk of becoming a tobacco smoker, a pattern described as a “reverse gateway” (Patton et al., 2005). Both gateway patterns probably reflect a shared route of administration (smoking) (Agrawal & Lynskey, 2009), the fact that cannabis smokers affiliate with tobacco smokers, and the effects of mixing tobacco and cannabis in joints.

In connection with the 2011 European School Survey Project on Alcohol and Other Drugs (ESPAD), a special study was undertaken on the prevalence of polydrug use among students from European countries that participated in the 2011 ESPAD survey (Hibell et al., 2012). Polydrug use was defined as the use of more than one of the following substances: tobacco (more than five cigarettes per day in the past 30 days), alcohol (consumption on 10 or more occasions in the past 30 days), cannabis (any use in the past 30 days), other illicit drugs (any lifetime use) and tranquillizers/sedatives without a prescription (any lifetime use). The overall prevalence of polydrug use (2-plus substances) in the total sample was very close to 9% in both survey years. The combination tobacco-cannabis was found in 9.7% of the polydrug group and the combination alcohol-cannabis was found in 5.7 %. The most common combination was tobacco–alcohol which was found in 12.4 % of the group (Hibell et al., 2012).

6.1.4 Psychosis and schizophrenia

In discussing relationships between cannabis use, psychosis and schizophrenia, it is necessary to define psychosis and schizophrenia clearly. Schizophrenia is a mental and behavioural disorder classified in the ICD-10. Schizophrenia is characterized by distortions in thinking, perception, emotions, language, sense of self and behaviour. Common experiences include hearing voices and delusions (WHO, 1992). Regular cannabis use has been reported to be more common among persons with schizophrenia (Myles, Myles & Large, 2015). The regular use of cannabis with a higher THC content and a lower CBD concentration may increase the risk for schizophrenia and lower the age of onset of the disease (Di Forti et al., 2014, 2015)

A 15-year follow-up study of schizophrenia among 50 465 Swedish male conscripts found that those conscripts who had tried cannabis by age the age of 18 years were 2.4 times more likely to be diagnosed with schizophrenia over the next 15 years than those who had not (Andréasson et al., 1987). After statistical adjustment for a personal history of psychiatric disorder by age 18 and a number of psychosocial confounders, those who had used cannabis 10 or more times by age 18 were 2.3 times more likely to be diagnosed with schizophrenia than those who had not used cannabis.

Zammit et al. (2002) reported a 27-year follow-up of the above-mentioned Swedish cohort. They also found a dose-response relationship between frequency of cannabis use at the age of 18 years and the risk of schizophrenia during the whole follow-up period (although the strength of the relationship declined with age). This effect persisted after statistically controlling for confounding factors. The researchers estimated that 13% of cases of schizophrenia would have been averted if no one in the cohort had used cannabis.

The Swedish cohort findings have been supported in smaller longitudinal studies in the Netherlands (van Os et al., 2002), Germany (Henquet et al., 2004) and New Zealand (Arseneault et al., 2002; Fergusson, Horwood & Swain-Campbell, 2003; Stefanis et al., 2014). All of these studies found a relationship between cannabis use and psychotic disorders or psychotic symptoms and these relationships persisted after adjustment for confounders. A meta-analysis of these longitudinal studies (Moore et al., 2007) reported that psychotic symptoms or psychotic disorders were higher in regular cannabis users than in non-users (OR 2.09 [95% CI: 1.54, 2.84]).

Reverse causation is a possible explanation of these findings if persons with schizophrenia use cannabis to relieve the symptoms of their illness. This possibility has been addressed to some extent in some of these longitudinal studies by excluding cases who reported psychotic symptoms at baseline, or by statistically adjusting for pre-existing psychotic symptoms. However, several large studies show that cannabis use preceded onset of psychosis (Andréasson et al., 1987; DiForti et al., 2009; Fergusson et al., 2003).

A second possibility is the common cause hypothesis – i.e. that the association is explained by other factors (e.g. genetic risk, childhood abuse) that increase the risk that young people will use cannabis and develop schizophrenia. This possibility has been addressed in some studies by comparing the rate of schizophrenia in persons who abuse different drugs. In a nationwide cohort of 30 547 patients receiving treatment for substance use disorders in Chile, there was an increased risk of a diagnosis of schizophrenia among cannabis users compared with patients who were users of other drugs (RR = 2.08, 1.6–2.7) and a dose-response association between cannabis use and risk of a schizophrenia diagnosis (Jorquera et al., 2015).

The common cause hypothesis has been harder to exclude because the association between cannabis use and psychosis is attenuated after statistical adjustment for potential confounders in many studies, and no study has been able to assess all plausible confounders. Genetic epidemiological studies have assessed the degree to which shared genetic risk factors may explain the association between cannabis use and psychoses. These have included studies of sib-pairs (McGrath et al., 2010), studies of the strength of the relationship between cannabis and psychosis in persons who differ in genetic relationship (Giordano et al., 2014), and correlations between polygenic risk scores for schizophrenia and cannabis use in large twin samples (Power et al., 2014). These studies suggest that shared genetic factors may explain some but not all of the association between cannabis and psychosis.

Researchers who favour a causal explanation point to its biological plausibility (e.g. (Di Forti et al., 2009)). This is indicated by double-blind studies which show that THC produces dose-related increases in positive and negative symptoms of psychosis in persons who do and do not have psychoses (D'Souza, 2004; Morrison, 2009; Murray et al., 2013). Psychotic syndromes have also been reported in patients who have been treated with the cannabinoid extract Sativex (Therapeutic Goods Administration, 2013). Compared with matched controls, those with psychotic disorders, and their siblings, are more sensitive to the psychotogenic effects of acute THC administration (D’Souza et al, 2005; Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014). A recent case-control study by Di Forti et al (2009) suggested that regular use of cannabis with high levels of THC and low levels of CBD increased the risk of developing schizophrenia by 3-5 times.

Researchers who remain sceptical about a causal role for cannabis (e.g. Gage, Zammit & Hickman, 2013) point to the absence of an increase in the incidence of schizophrenia as cannabis use increased among young adults. The evidence is mixed. An Australian modelling study did not find any marked increase in incidence after steep increases in cannabis use during the 1980s and 1990s (Degenhardt, Hall & Lynskey, 2003). However, a similar modelling study in the United Kingdom (Hickman et al., 2007) argued that it was too early to say. Two case register studies in Britain (Boydell et al., 2006) and Switzerland (Ajdacic-Gross et al., 2007) reported an increased incidence of psychoses in recent birth cohorts, but a United Kingdom study of general practice patients did not (Advisory Council on the Misuse of Drugs, 2008).

The available evidence points to a modest contributory causal role for cannabis in schizophrenia. There is a consistent dose-response relationship in a number of prospective studies between cannabis use in adolescence and the risk of developing psychotic symptoms or schizophrenia. Self-medication is implausible, and a causal relationship is biologically plausible (see Evins, in Haney & Evins, 2016). Researchers who are not convinced by the evidence argue that these studies have not excluded the possibility that the relationship is explained by residual confounding (see Haney, in Haney & Evins, 2016).

6.1.5 Other mental disorders

Depression is a common mental health problem and one of the most important contributors to the global burden of disease (Ustün et al., 2004; Moussavi et al., 2007). Findings of high prevalence of comorbid cannabis use and depression have been replicated in many large-scale cross-sectional studies and in mental health surveys. Persons with cannabis-use disorders have higher rates of depressive disorders (Swift, Hall & Teesson, 2001). In longitudinal studies, the relationship between regular cannabis use and depression has been much weaker than that for cannabis and psychosis (Degenhardt & Hall, 2012; Manrique-Garcia et al., 2012; Fergusson & Horwood, 1997). Meta-analyses of these studies (Moore et al., 2007) found modest associations between regular or heavy cannabis use and depressive disorders (Moore at al., 2007: OR = 1.49 [95% CI: 1.15, 1.94]; Lev-Ran et al., 2014: OR = 1.62 [95% CI 1.21-2.16]). Many of these studies did not adequately control for confounders, or excluded the possibility that depressed young people were more likely to use cannabis (Horwood et al., 2012) and in some studies associations disappear when better control is introduced (Feingold et al., 2015).

Much the same has been true of studies of cannabis-use disorders among persons diagnosed with bipolar disorders (e.g. (Lai & Sitharthan, 2012; Lev-Ran et al., 2013; Silberberg, Castle & Koethe, 2012; Agrawal, Nurnberger & Lynskey, 2011)). In one longitudinal study, cannabis use at baseline predicted an increased risk of manic symptoms in a three-year follow-up (Henquet et al., 2006). However, these studies have not adequately controlled for confounding variables or ruled out reverse causation with cannabis being used to lift depressed mood and reduce manic excitement (Silberberg, Castle & Koethe, 2012).

Persons with cannabis-use disorders also have higher rates of anxiety, conduct disorders, eating disorder and personality disorders (Goodman & George, 2015). The reasons for these common patterns of comorbidity have not been as well investigated in prospective studies as those between cannabis-use disorders and psychosis and depression. It remains to be discovered whether these disorders increase the risks of using cannabis (as is plausible for conduct and personality disorders), whether their outcomes are worsened by cannabis-use disorders, and to what degree these disorders share common risk factors with cannabis-use disorders (Hall, Degenhardt, & Teesson, 2009).

The high prevalence of comorbidity between drug-use disorders and other mental disorders does not mean that one causes the other, but comorbidity between mental and substance-use disorders is highly prevalent across countries. In general, people with a substance-use disorder had higher comorbid rates of mental disorders than vice versa, and people with drug-use disorders had the highest rates of comorbid mental disorders. In general, while there are associations between regular cannabis use or cannabis-use disorders and most mental disorders, causality has not been established. Reverse causation and shared risk factors cannot be ruled out as explanations of these relationships.

6.1.6 Suicide risk, ideation and attempts

Bagge and Borges (Bagge & Borges, 2015) conducted a case-crossover study of 363 persons who had recently attempted suicide and were treated in a trauma hospital for a suicide attempt within the previous 24 hours in the state of Mississippi, USA. The researchers compared rates of cannabis use in the 24 hours leading up to the suicide (case period) to that in the 24 hours of the day before the suicide (control period). They found that 10.2% of suicide attempters had used cannabis in the case period while 13.2% used cannabis in the control period.

The USA’s Drug Abuse Warning Network (DAWN) estimated rates of cannabis use among drug-related visits to hospital emergency departments for suicide in 2011 (SAMHSA, 2013). Cannabis was coded as positive if hospital staff perceived it to be the cause or a contributor to the emergency visit. Cannabis was involved in an estimated 6.5% of drug-related suicide attempts, and in 46% of attempts the person also used alcohol. In the 23% of drug-related suicide attempts with toxicology reports, 16.8% tested positive for cannabis, although this cannabis use could have occurred days or even up to one week earlier. In general, 9.5% of all toxicology reports for deaths by suicide (Borges, Bagge & Orozco, 2016) show the presence of cannabis. There is preliminary evidence of higher detection of cannabis among suicide decedents that do not involve overdose (CDC, 2006) and higher detections among male suicide decedents using non-overdose methods than among females (Darke, Duflou & Torok, 2009; Shields et al., 2006). Homicide victims appear to have higher detection rates of cannabis at the time of death than suicide victims do (Darke, Duflou & Torok, 2009; Sheehan et al., 2013).

Overall, studies on cannabis use and suicide ideation and attempts have produce mixed results. A case-control study of 302 serious suicide attempts in New Zealand and general hospital community controls (Beautrais, Joyce & Mulder, 1999) found an association between harmful use of cannabis and suicide attempt. The association was substantially reduced after statistical adjustment for confounding. A small case-control study in the USA did not find an association (Petronis et al., 1990). Results from longitudinal studies are more numerous and have varied as to whether the associations persisted after adjustment for confounders, with newer and larger studies reporting positive associations. Fergusson and colleagues (Fergusson, Lynskey & Horwood, 1996; Fergusson & Horwood, 1997) found that regular cannabis use at the age of 15 years predicted ideation and attempts at 16-17 years in New Zealand, but these associations disappeared after controlling for confounders (Fergusson & Horwood, 1997). A 30-year follow-up of the cohort (van Ours et al., 2013) found a dose-response relation between cannabis use and suicidal ideation that persisted after controlling for confounding variables. The New Zealand Dunedin birth cohort (McGee, Williams & Nada-Raja, 2005) also reported an association between cannabis use at 15 years of age and suicidal ideation at 18-21 years of age, but this was no longer statistically significant after adjustment for confounders. A pooled analysis of Australian and New Zealand cohort studies found a dose-response relation between the frequency of cannabis use before the age of 17 years and suicide attempts at 17-25 years (Silins et al., 2014).

Longitudinal studies in the USA and other countries have found associations between cannabis use and suicidality over varying follow-up periods. In some studies the associations vary with age and the measure of cannabis use (e.g. Newcomb, Vargas-Carmona & Galaif, 1999; Newcomb, Scheier & Bentler, 1993). Others have found associations with suicidal ideation but not with suicide attempts (Juon & Ensminger, 1997). In some studies the association has persisted after controlling for confounding variables (e.g. Bovasso, 2001; Borowsky, Ireland & Resnick, 2001; Clarke et al, 2014; Pedersen, 2008), whereas in other studies it has not persisted, or has persisted only among subgroups (e.g. Wilcox & Anthony, 2004; Zhang & Wu, 2014; Wichstrom, 2000).

6.1.7 Suicide mortality

There have been very few studies of associations between regular cannabis use and death by suicide. A follow-up study of Swedish conscripts (Andréasson & Allebeck, 1990) reported that those who had used cannabis more than 50 times by the age of 18 years were at increased risk of dying by suicide. The same association was observed in a 33-year follow-up (Price, 2009) but it was no longer significant after adjusting for baseline alcohol, tobacco and other drug use, and psychiatric disorders.

A case-control study conducted among 108 individuals who committed suicide and 108 who died in accidents, matched for age and gender, in Cali, Colombia, found an increased odds ratio (OR=2.85(95% CI= 1.31–6.24) among those with cannabis-use disorders (Palacio et al., 2007). A large case-control study of 1463 suicides and 7392 natural deaths (Kung, 2003; Kung, 2005) found an association between any cannabis use and suicide risk after adjusting for depression, alcohol and mental health services. So did a four-year follow-up of a large group of patients with cannabis-use disorders in Denmark, which found an increased risk (Males OR=2.28 (95% CI=1.54– 3.37); Females OR=4.82 (95% CI= 2.47–9.39)) of suicide among those with cannabis-use disorders (Arendt, 2013).

6.1.8 Areas that require more research

There have been recent reports that higher proportions of lifetime users seem to have developed cannabis-use disorders.

As a result, updated longitudinal research (including dose-response, potency, frequency of use, and age of onset and reason for use) is needed to identify if and why more users seem to develop cannabis-use disorders.

Better epidemiological and longitudinal studies are needed to determine the association between cannabis use and the risk of different types of mental disorders and suicidal ideation, attempts and death. These studies should include a wide age range, diverse social and geographical populations, and should better measure cannabis use in order to assess dose-response relations.

• Baseline measures should be chosen to rule out reverse causation and, if possible, third causes such as concurrent use of other substances, mental disorders and genetic vulnerability.
• Improved genetic studies are required to investigate the overlap between genotype and phenotype for schizophrenia and cannabis-use disorders.
• Persons with cannabis-use disorders have higher rates of depressive disorders. Many of these studies have not adequately controlled for confounders or excluded the possibility that depressed young people are more likely to use cannabis.
• Persons with cannabis-use disorders also have higher rates of anxiety, conduct disorders, eating disorder and personality disorders. The reasons for these common patterns of comorbidity have not been well investigated in prospective studies. Such investigation is needed.

While the evidence tends to suggest that cannabis use is associated with suicide ideation and suicidal behaviour, the lack of homogeneity in the measurement of cannabis exposure across studies and, in some instances, the lack of systematic control for known risk factors are clear limitations in current knowledge (Borges, Bagge & Orozco, 2016).

• Specifically for suicidal behaviour, efforts should be made to disentangle any effects of regular cannabis use from the short-term effects of use on suicidal ideation and behaviour.