R E V I E W

Drug and Alcohol Review (May 2010), 29, 304–317

DOI: 10.1111/j.1465-3362.2009.00132.x

REVIEW

An overview of systematic reviews on cannabis and psychosis:

Discussing apparently conflicting results dar_132 304..317

SILVIA MINOZZI, MARINA DAVOLI, ANNA M. BARGAGLI, LAURA AMATO,

SIMONA VECCHI & CARLO A. PERUCCI

Department of Epidemiology, Local Health Unit Roma E, Roma, Italy

Abstract
Issues. Cross-sectional surveys have revealed that cannabis is the most widely used illicit substance in Western countries.
Cannabis intoxication can lead to acute, transient psychotic symptoms and the short-term exacerbation of pre-existing psychotic
symptoms. However, controversy exists about whether cannabis can actually cause long-term psychosis. Approach. We
summarised the findings of systematic reviews on the association between cannabis use and psychosis, searching MEDLINE,
EMBASE and CINAHL up to August 2007.We assessed the methodological quality, selected the better quality reviews and
analysed reasons for discordant results. Key Findings.We included five systematic reviews. Four of the reviews performed a
meta-analysis and showed a consistent association between cannabis use and psychosis; the fifth review considered psychological
problems more broadly, did not perform a meta-analysis and reported an inconsistent association.The reasons for discordance
were: different outcomes (psychosis vs. psychological problems), different inclusion criteria for primary studies and different
methods for summarising the results. Implications. This overview shows a consistent association between cannabis use and
psychotic symptoms, though it is not possible to draw firm conclusions about a causal relationship. Reverse causality and residual
confounding cannot be excluded. An interaction with other environmental and genetic factors is difficult to ascertain.
Conclusion.We conclude that there is insufficient knowledge to determine the level of risk associated with cannabis use in
relation to psychotic symptoms and that more information is needed on both the risks of cannabis use and the benefits of
preventive interventions to support evidence-based approaches in this area. [Minozzi S, Davoli M, Bargagli AM, Amato L,
Vecchi S, Perucci CA. An overview of systematic reviews on cannabis and psychosis: Discussing apparently conflict-
ing results. Drug Alcohol Rev 2010;29;304–317]

Key words: cannabis, marijuana abuse, psychotic disorders, systematic review.

impaired work or school performance, loss of self-

Introduction
Cross-sectional studies of the general population and
students have consistently revealed that cannabis is the
most widely used illicit substance in many regions of
the world, including Europe [1], North America and
Australia [2,3]. Cannabis is also the second most fre-
quently cited drug in reports on individuals entering
substance abuse treatment for the first time [1–5].
Clients seeking treatment for cannabis use can present
with social impairment, such as complaints from family
members, loss of friendship, financial difficulties,
confidence, memory loss, legal problems and psychiatric
distress, such as somatisation, depression, anxiety, irri-
tability, phobic anxiety, paranoid ideation and psychosis
[6–8]. Cannabis intoxication can lead to acute, transient
psychotic symptoms in some individuals [9] and
produce a short-term exacerbation or recurrence of
pre-existing psychotic symptoms [10–12]. However,
controversy remains about whether cannabis use can
actually cause long-term psychosis [1].
There is evidence of an association between cannabis
use and the development of psychotic symptoms, but

Silvia Minozzi MD, Marina Davoli MD, MSc, Anna M. Bargagli BSc, Laura Amato MD, Simona Vecchi BSc, Carlo A. Perucci MD.
Correspondence to Dr Marina Davoli, Department of Epidemiology, Local Health Unit Roma E, Via di Santa Costanza 53, 00198 Rome, Italy.
Tel: +390683060483; Fax: +390683060374; E-mail: davoli@asplazio.it
Conflict of Interest: none
Received:19 September 2008; accepted for publication 2 July 2009.

© 2009 Australasian Professional Society on Alcohol and other Drugs


the nature of this association is widely debated. Four
main hypotheses have been formulated to explain the
observed association [13]:
1. Confounding: the association is spurious; can-
nabis use is often associated with the use of other
drugs or with other factors responsible for the
later onset of psychosis.
2. Interaction: cannabis is a component cause; psy-
chosis is caused only in vulnerable people and in
interaction with other environmental and genetic
factors.
3. Reverse causality: people with psychosis use can-
nabis more often in an attempt to cope with nega-
tive symptoms that stem from psychosis
(depression, anxiety or dysphoria) or to alleviate
the side-effects of antipsychotic medication.
4. Etiological: cannabis is the direct cause of psycho-
sis; it alone is a necessary and sufficient cause.
Systematic reviews critically appraise and summarise
the published evidence concerning a particular
problem and have gained prominence as useful tools for
evidence-based decision making. The number of sys-
tematic reviews has increased at least 500-fold during
the past 10 years, and they have been applied exten-
sively in the field of drug addiction. This increase has
lead to an increased number of reviews addressing very
similar questions, with a concomitant increase in con-
flicting results. Such a discrepancy causes difficulties
for decision makers, clinicians and patients.
The aim of this overview was to summarise the major
findings of published systematic reviews on the associa-
tion between cannabis use and psychosis, and to
analyse the possible reasons for the discordant results
among them.
Methods
Search strategy
We performed a systematic search of MEDLINE,
EMBASE and CINAHL from 2000 to August 2007.
For each database, separate detailed search strategies
were developed based on controlled vocabulary and
free text terms, combining the following search terms
and the boolean operators or/and: [Substance-related
disorders or cannabis or marihuana or marijuana] and
[psychosis or psychotic disorders or schizophrenia or
psychotic*]. A methodological search filter for review
articles was added. We also checked the references of
existing reviews for potentially relevant papers. There
was no language restriction.
Inclusion criteria
We included systematic reviews that clearly state the
objective, define the inclusion criteria for primary
Systematic reviews on cannabis and psychosis 305
studies, report the raw data of the primary studies
included, and include observational studies that assess
the relationship between cannabis use and psychosis.
Data extraction and quality assessment
Two authors independently extracted data and assessed
the quality of the included reviews using the Overview
Quality Assessment Questionnaire (OQAQ) scale [14];
any disagreement was resolved by discussion among the
authors.
The OQAQ scale is composed of nine items designed
to assess specific aspects of the methodological quality
of systematic reviews and one overall assessment item
that requires assessors to assign a quality score on a
7-point scale, where 1 represents the presence of exten-
sive flaws and 7 represents minimal flaws.
Analysis of discordant reviews
In order to analyse the reasons for discordant results
among the reviews, we used the following criteria pro-
posed by Jadad [15]:
• Differences in methodological quality.
• Differences in clinical question: population, expo-
sure, outcomes.
• Differences in study selection: search strategy,
inclusion criteria.
• Differences in data extraction: methods to
measure exposure and outcomes.
• Differences in the assessment of study quality:
methods and interpretation of quality assessments.
• Differences in methods to incorporate quality
assessments into the review.
• Differences in the ability to combine studies: sta-
tistical methods, criteria to judge the ability to
combine studies.
We also retrieved the full text of primary studies
included in the reviews in order to understand better
the reasons for inclusion or exclusion and to determine
the appropriateness of any statistical synthesis of the
results.
Results
We identified 41 reports. Twenty-one reports were
excluded on the basis of the title and abstract because
they were not pertinent to the objective of our review,
and the full text of 20 articles was retrieved for more
detailed evaluation. Fifteen of the 20 articles were
excluded because they did not fulfil the inclusion cri-
teria. Thus, five systematic reviews were included in the
present review [16–20].
© 2009 Australasian Professional Society on Alcohol and other Drugs
306 S. Minozzi et al.
Characteristics of included reviews
Four of the five included reviews [16,17,19,20]
assessed the association between cannabis exposure
and the occurrence of schizophrenia (any type),
schizophrenia-like disorders, psychosis not otherwise
specified, or psychotic symptoms; the fifth review
assessed the association between any illicit drug use and
the occurrence of any psychological or social harm, but
it also included studies assessing the association
between cannabis use and psychosis considered as ‘psy-
chological or social harm’ [18]. One of the reviews [20]
also assessed the relationship between cannabis use and
affective disorders (depression, suicidal ideation or
attempt, anxiety). Three of the reviews [16,18,20]
included only longitudinal cohort studies, whereas the
other two [17,19] included both cross-sectional and
longitudinal cohort studies. Table 1 shows the study
design of each study included and excluded in each of
the five reviews.
Overall, eight longitudinal cohort studies [21–32]
assessing the association between cannabis exposure
and psychosis development were identified and
included in the reviews. Macleod [18] included 16 lon-
gitudinal studies, only four of which were on the asso-
ciation between cannabis use and psychosis and
included in the other reviews. The other studies in that
review assessed the effect of cannabis use on affective
disorders, adult role, job success and quality of social
relationships.
Methodological quality
The methodological quality of three of the reviews was
poor (overall quality score 2/7 [16,17] and 3/7 [19]),
whereas the other two were of good quality (overall
quality score 5/7 [18] and 6/7 [20]) (Table 2).The main
weaknesses concerned the comprehensiveness of the
bibliographic search (cannot tell for 2/5 reviews), the
avoidance of selection bias (cannot tell for 4/5 reviews),
the assessment of the methodological quality of
primary studies (not done in 3/5 reviews). In one review
[18], conclusions were not supported by the data
because the results of the primary studies were not
clearly described. Only one review [20] reported the
number of excluded studies and the reasons for exclu-
sion. The information provided about the studies
included in the reviews was not detailed enough to
ascertain important information on the primary studies
for all but one of the reviews [20], which provided a
detailed table of the study characteristics. In particular,
the reviews had no clear information regarding:
• Study design: some studies were described as
case–control studies but appear in the description
© 2009 Australasian Professional Society on Alcohol and other Drugs
as cross-sectional studies, which invalidated any
result on possible causal association;
• Exposure assessment: how cannabis use was
assessed and if frequency, length and amount of
use were considered;
• The assessment of confounding: how the use of
other drugs was assessed and if frequency, length
and amount of other drug use were considered;
• The outcome definition: whether primary studies
considered ‘schizophrenia’ using the Diagnostic
and Statistical Manual of Mental Disorders
(DSM-IV) or International Classification of
Disease (ICD-10) criteria, or ‘schizophreniform
disorders’ or ‘psychotic symptoms’. One review
[18] used the term ‘psychological problems’ as the
outcome, and it is not clear which outcomes used
in the primary studies were considered under this
broader term;
• Effect measure estimate: why and how odds ratios
(OR) were calculated in cohort studies, how they
took losses to follow up into consideration; if the
overall estimates reported in the reviews were
computed using consistent measures of exposure
(‘heavy users’ or ‘any users’) and outcome; and
how confounding was incorporated into the
pooled estimate.
Findings
Four of the reviews [16,17,19,20] performed a meta-
analysis and were concordant in finding an association
between cannabis use and the occurrence of psychotic
disorders (Table 3). Nevertheless, the way of combin-
ing results varied among the individual reviews:
Arsenault et al. [16] included a duplication of Swedish
conscript data in the meta-analysis and combined
results for ‘ever use’ and dependence; Henquet et al.
[17] and Semple et al. [19] combined cross-sectional
and longitudinal data in the meta-analysis and did not
distinguish between ‘ever use’ and dependence; Semple
et al. [19] combined the crude odd ratios; and Moore
et al. [20] combined only longitudinal studies, pre-
sented separate results for ‘ever use’ and ‘most frequent
use’ and combined the adjusted OR. None of the
reviews found significant statistical heterogeneity. The
review without a meta-analysis [18] found an inconsis-
tent association between cannabis use and psychologi-
cal problems.The authors reported that the adjustment
for confounding factors lead to substantial attenuation
of the association but did not report the original data.
To analyse possible reasons for the discordant results,
we restricted our analysis to the two higher quality
reviews [18,20]. The comparability of the two reviews
using Jadad’s criteria is reported in Table 4. The main
difference between the two reviews was outcome defi-
 Table 1. Primary studies included and excluded in systematic reviews

Arseneault Henquet Macleod et al. Semple et al. Moore et al.

Primary studies et al. 2004 [16] et al. 2005 [17] 2004 [18] 2005 [19] 2007 [20]

Date of bibliographic search Study design, outcomes Not reported Not reported June 2003 January 2004 September 2006

Andréasson et al. 1987 [21], Longitudinal population Yes Yes Yes Yes Yes
Zammit et al. 2002 [22], based (conscripts)
Zammit 2004 [23] schizophrenia
(Swedish Conscripts
study)

Arsenault et al. 2002 [24], Birth cohort Yes Yes Yes Yes Yes
Caspi et al. 2005 [25] Psychotic symptoms
(Dunedin study) Schizophreniform disorders
van Os et al. 2002 [26] Longitudinal population Yes Yes No Yes Yes
(NEMESIS study) based Psychotic symptoms
Fergusson et al. 2003 [27], Birth cohort Yes Yes Yes Yes (not included in Yes
2005 [28] (CHDS study) Psychotic symptoms meta-analysis
Henquet et al. 2005 [29] Longitudinal population No Yes No No Yes
(EDSP study) based
Psychotic symptoms
Weiser et al. 2002 [30] Adolescent longitudinal No Yes Included as low quality No No. Excluded because
population based. study. Not considered no data specifically
Schizophrenia spectrum for summary results on cannabis use
personality disorders

Tien & Anthony Longitudinal Population No No No Yes (not included in Yes

1990 [31] (ECA study) based Psychotic symptoms meta-analysis)

Wiles et al. 2006 [32] Longitudinal Population No No No No Yes

(NPMS) based Psychotic symptoms
Stefanis et al. 2004 [33] Cross-sectional No Yes No No No. Excluded because
Psychotic symptoms cross sectional
Rolfe et al. 1993 [34] Cross sectional No No No Yes No. Excluded because
schizophrenia cross sectional
Agosti et al. 2002 [35] Cross sectional No No No Yes No. Excluded because
Non affective psychosis cross sectional
Degenhardt & Hall Cross sectional. No No No Yes No. Excluded because
2001a [36] Psychotic symptoms cross sectional
Systematic reviews on cannabis and psychosis

(continued)
307

© 2009 Australasian Professional Society on Alcohol and other Drugs

 308

Table 1. (Continued)

Arseneault Henquet Macleod et al. Semple et al. Moore et al.

Primary studies et al. 2004 [16] et al. 2005 [17] 2004 [18] 2005 [19] 2007 [20]
S. Minozzi et al.

Date of bibliographic search Study design, outcomes Not reported Not reported June 2003 January 2004 September 2006

Degenhardt et al. 2001b [37] Cross sectional. No No No Yes No. Excluded because
Psychotic symptoms cross sectional
Farrel et al. 2002 [38] Cross sectional. No No No Yes No. Excluded because
Psychosis population are prison
sample
Grech et al. 1998 [39] Cross sectional. No No No Yes No. Excluded because
Psychosis participants had
already psychotic
symptoms

© 2009 Australasian Professional Society on Alcohol and other Drugs

Miller et al. 2002 [40] Cross sectional. No No No Yes No. Excluded because
Psychotic symptoms cross sectional
Phillips & Johnson 2002 [41] Longitudinal study; very No No No Yes No
high-risk population
Psychotic symptoms
Resnick et al. 1997 [42], Longitudinal Population No No Yes No
Dornbusch et al. 1999 [43] based Violent behavior
(National Longitudinal
Study on adolescent health)
Guy et al. 1993 [44], Stein Longitudinal Population No No Yes No No. Excluded because
et al. 1993 [45] (Boston based Job involvement no data specifically
Schools project) on cannabis use
Brook et al. 1998 [46] Longitudinal Population No No Yes No No. Included for
(Children in the based Affective outcomes affective outcomes
community project)
Brook et al. 1996 [47], 1999 Longitudinal Population No No Yes No No. Excluded because
[48] (Children in the based Violent behaviour, not examined
community project) adult role psychotic or affective
outcomes
Brook et al. 1999 [49] (East Longitudinal Population No No Yes No No
Harlem study) based. Behavioural
outcomes
Brunswick & Messeri 1986 Longitudinal Population No No Yes No No. Excluded because
[50], 1999 [51] (Central based. Physical health, not examined
Harlem Study) behaviour outcomes, social psychotic or affective
attainment outcomes
 Newcomb et al. 1993 [52], Longitudinal Population No No Yes No No. Included for
1999 [53] (LA schools based. Affective outcomes affective outcomes
study)
Kandel et al. 1995 [54] (New Longitudinal Population No No Yes No No. Excluded because
York schools study) based. Income outcome not examined
psychotic or affective
outcomes
Friedman et al. 1996 [55] Longitudinal Population No No Yes No No
(National Collaborative based. Delinquency,
Perinatal Project, NCPP) antisocial behaviour
Kaestner 1994 [56], Windle Longitudinal Population No No Yes No No
1997 [57] National based. Income outcome
Longitudinal survey of
youth)
White et al. 1999 [58] Longitudinal Population No No Yes No No
(Pittsburgh Youth Study) based. Violent behaviour
Ellickson et al. 1998 [59], Longitudinal Population No No Yes No No
2000 [60] (Project Alert) based. Violent behaviour
Bray et al. 2000 [61] (South Longitudinal Population No No Yes No No
eastern public schools based. Educational
study attainment
Bates & Labouvie 1997 [62] Longitudinal Population No No Yes No No. Excluded because
(Woodlawn study) based. Alcohol use not examined
psychotic or affective
outcomes
Juon & Ensminger 1997 [63] Longitudinal Population No No Yes No No. included for
(Woodlawn study) based. Suicidal ideation or affective outcome
attempts
Systematic reviews on cannabis and psychosis
309

© 2009 Australasian Professional Society on Alcohol and other Drugs

310 S. Minozzi et al.

Table 2. Quality of conduct (OQAQ checklist)

Arseneault et al. Henquet et al. Macleod et al. Semple et al. Moore et al.
Item 2004 [16] 2005 [17] 2004 [18] 2005 [19] 2007 [20]

Search methods described Partially No Yes Yes Yes

Search comprehensive Can’t tell Can’t tell Yes Yes Yes
Inclusion criteria reported Yes Yes Yes Yes Yes
Selection bias avoided Can’t tell Can’t tell Can’t tell Can’t tell Yes
Criteria for quality assessment reported Partially No Yes No Partially
Criteria used appropriate No No Yes No Yes
Methods to combine findings reported Yes Yes Yes Yes Yes
Findings appropriately combined No No Yes No Yes
Conclusions supported by the data Yes Yes No Yes Yes
Overall score of scientific quality (*) 2 2 5 3 6

*Score from 1 (extensive flaws) to 7 (minimal flaws); OQAQ: Overview Quality Assessment Questionnaire.

Table 3. Results of meta-analysis

Author Meta-analysis of included studies n studies included

Arseneault et al. 2004
[16]
Henquet et al. 2005
[17]
Macleod et al. 2004
[18]
Semple et al. 2005 [19]
Moore et al. 2007 [20]
Adjusted OR: 2.34 (95% CI: 1.69, 2.95) Five studies, n of
Longitudinal studies. participants 101 497
Ever use and dependence considered altogether
The same cohort from the Swedish conscript’s study was included
twice in the meta-analysis (Andreasson et al. and Zammit et al.
results)
Adjusted OR: 2.1 (95% CI: 1.7, 2.5) Seven studies, n of
Cross-sectional and longitudinal studies participants 112 218
Ever use and dependence considered altogether
Meta-analysis not performed Sixteen studies, n of
Inconsistent association between cannabis use and psychological participants not
problems (data not reported). The adjustment for confounding reported
factors lead to substantial attenuation of the association
Crude OR: 2.93 (95% CI: 2.36, 3.64) Seven studies , n of
Cross-sectional and longitudinal studies participants 51 688
Ever use and dependence considered altogether
Adjusted OR: ever use OR 1.41 (95% CI: 1.20, 1.65) Seven studies, n of
Heavy users: OR 2.09 (95% CI: 1.54, 2.84) participants not
Longitudinal studies reported

CI, confidence interval; OR, odds ratio.

nition, which could also explain differences in the
included studies. Macleod et al. [18] considered any
psychological problem without any further explana-
tion or definition, whereas Moore et al. [20] consid-
ered only psychotic disorders and affective disorders
defined by the DSM-IV. Furthermore, six studies
included by Macleod as studies considering psycho-
social problems as the outcome measure were also
included by Moore, but three were defined as studies
assessing affective disorders and three as studies
assessing psychosis as the outcome measures. Again,
in the list of studies excluded by Moore, there were
six that were included by Macleod as high-quality
© 2009 Australasian Professional Society on Alcohol and other Drugs
studies; the reason for excluding these studies was a
lack of considering psychotic or affective outcomes
(four studies) or no specific data on cannabis use (two
studies). The other major difference was in reporting
the outcomes of primary studies: Moore reported an
adjusted OR with a 95% confidence interval for each
study, whereas Macleod gave a generic qualitative
description that was more prone to interpretation.
Macleod stated that the association was often substan-
tially reduced, adjusting for confounding factors, but
did not report the data, whereas Moore clearly
reported how much the association was reduced for
each study.
 Systematic reviews on cannabis and psychosis 311

Table 4. Source of discordant results among the two higher quality reviews (criteria proposed by Jadad)

Criteria Macleod et al. 2004 [18] Moore et al. 2007 [20]

Clinical question
Population General population aged 25 or younger General population; excluded people with
mental illness or substance related
problems
Exposure Any illicit drug (results reported for cannabis) Cannabis use
Outcomes Any psychological or social harm Psychosis (schizophrenia, schizophreniform
or other psychotic disorders, psychotic
symptoms.
Affective, mood or bipolar disorders,
depression, suicidal ideation or attempts,
anxiety, neurosis and mania
Study selection and inclusion
Selection criteria Population based prospective studies assessing Population based prospective studies
the association between any drug use among assessing the association between cannabis
young people and any psychological or social use and psychosis or affective disorders
harm
Search strategy Medline, Embase, CINAHL, PsychLIT, Web of Medline, Embase, CINAHL, PsycINFO, ISI
Science, specialised databases, contact with Web, specialised databases, contact with
authors. No language restriction. Up to June authors. No language restriction. Up to
2003 September 2006
Included studies 48 longitudinal studies, three of which included 7 longitudinal studies for psychosis, three of
in Moore for the outcome psychosis and which included in Macleod. 15 cohort
three for the outcome affective disorders studies for affective disorders, three of
which included in Macleod
Data extraction
Methods to measure No raw data reported; only descriptive OR with 95% confidence interval adjusted
outcomes for confounding
End points Psychological problems without further Risk of psychosis
description Risk of affective disorders
Assessment of study quality
Method to assess Two reviewers independently Two reviewers independently
quality
Quality criteria used Sample size and representativeness, age at Sampling strategy, response rate, missing
recruitment, duration and completeness of data, attrition, attempts to address reverse
follow up, validity and reliability of exposure causation, intoxication effects and
and outcomes measures, degree of adjustment confounding factors.
for potential confounding factors
Methods to incorporate Considered only the 16 studies judged of better Not stated
quality assessment in methodological quality
review
Assessment of the ability to combine results
Statistical methods Not used Meta-analysis of adjusted OR using the Der
Simonian and Laird random effects model
Clinical criteria to Quantitative synthesis considered misleading Studies on psychosis judged homogeneous
judge the ability to because heterogeneity in primary studies for enough to perform a meta-analysis.
combine results method to assess exposure and outcomes Studies on affective disorders judged too
heterogeneous to perform a meta-analysis

OR, odds ratio.

4–5 years in two studies [28,32] and more than 9 years

Characteristics of longitudinal primary studies
We analysed the primary studies included in the meta-
analysis by Moore et al. [20] to assess if they were
homogeneous enough to allow the pooling of their
results (Table 5).
The main difference among the studies was in the
length of follow up: 1–1.5 years in two studies [33,35],
in three studies [24,26,30] (range 1–27 years).
The other parameters that were considered did not
differ among the studies:
• The presence of psychotic symptoms at baseline
were assessed using structured and validated
instruments;
© 2009 Australasian Professional Society on Alcohol and other Drugs
 312
Table 5. Characteristic of longitudinal studies on psychosis outcomes included in the reviews

Participants and Follow-up Confounding factor Dose-response

Study baseline screening Exposure Outcome attrition considered effect

Andréasson et al. n: 50 053 Ever use Admission for 27 years Psychiatric diagnosis at Evidence of a
S. Minozzi et al.

1987 [21], Adult population based Frequency: 1 time, ICD-10 Attrition: data baseline, IQ, dose-response
Zammit et al. conscripts 2–4 times, 5–10 diagnosis of not reported interpersonal effect
2002 [22], Sweden times, 11–50 times, schizophrenia or relationship,
Zammit 2004 [23] Baseline: psychiatric >50 times schizoaffective disturbed behaviour
(Swedish interview for ICD-8 disorders in childhood, family
Conscripts study) diagnosis history of psychiatric
illness, alcohol
misuse, other drug
use
Tien & Anthony n: 2 295 DIS interview for DIS interview for any 1 year Age, gender, school Not studied.
1990 [31] (ECA adult population based lifetime ever use and self-reported Attrition: 20% attendance, Stronger effect
study) cohort daily use psychotic experience educational level, for daily use
USA employment and than ever use

© 2009 Australasian Professional Society on Alcohol and other Drugs

Baseline: DIS interview for marital status,
psychotic symptoms baseline mental
health problem, other
drug and alcohol use
Arsenault et al. 2002 n: 759 Ever use in the past DIS for 11 years sex, socioeconomic Not studied
[24], Caspi et al. Birth cohort year schizophreniform status, other drug
2005 [25] New Zealand Frequency of use in disorders (DSM-IV use, Psychotic
(DUNEDIN Baseline. 11 years the past year criteria) and symptoms at baseline
study) DIS-C for psychotic Use before age 15 psychotic
symptoms symptoms
van Os et al. 2002 n: 4 045 L section of M-CIDI BPRS for any 3 years Age, sex, ethnic group, Evidence of a
[26] (NEMESIS adult population based Any use psychotic Attrition : 30% level of education, dose-response
study) cohort. Frequency: symptoms previous risk factor effect
Netherlands <1 month, for psychosis, other
Baseline: G section of 3–4 months, drug use.
M-CIDI for psychotic 1–2 weeks, Subjects with
symptoms 3–4 weeks, daily psychotic symptoms
at baseline excluded
Weiser et al. 2002 n:50 413 Use of drug, frequency, Israeli National 4–15 years Intellectual functioning, Not studied
[30] Adolescent population addiction. Psychiatric Attrition: data social functioning,
based cohort (male) Type of drug not Hospitalization not reported non psychotic
Israeli specified, but cross Registry ; Diagnosis psychiatric disorders
Baseline: 16–17 years sectional surveys of schizophrenia- at baseline
Draft board compulsory made in the same spectrum
assessment (intelligence, time period indicated personality
personality, behavioural that the majority of disorders
traits) adolescents used (ICD-9)
cannabis
 Fergusson et al. 2003 n: 1 055 L section of M-CIDI SCL-90 for 9 years Sex, socio-economic Evidence of a
[27], 2005 [28] Birth cohort Any use psychotic Attrition: 17% variables, family dose-response
(CHDS study) New Zealand Frequency of use symptoms functioning, child effect
Baseline: 16 years. past year: <1 month, abuse, child
SCL-90 for psychotic >1 month, >1 week, neuroticism, IQ, self
symptoms daily esteem, novelty
seeking, prior
psychotic symptoms
or mental disorders,
other drug abuse
Henquet et al. 2005 n: 2 437 L section of M-CIDI G section of 4 years Age, sex, Evidence of a
[29] (EDSP study) adult population based Ever use (>5 times) M-CIDI for Attrition: 16% socio-economic dose-response
cohort Frequency: psychotic status, urbanicity, effect
Germany <1 month, symptoms childhood trauma,
Baseline: SCL-90 R for 3–4 months, (at least two other drug and
psychopathological 1–2 weeks, psychotic alcohol use,
experience 3–4 weeks, daily symptoms) predisposition to
psychosis
Wiles et al. 2006 [32] n: 3 045 Any use in the past year PSQ for psychotic 18 months Age, sex, marital status, Not studied
(NPMS study) adult population based but not dependent symptoms Attrition: 32% urbanicity, IQ, Suggestion for
cohort Dependency educational level, increasing
UK employment, effect for
Baseline: PSQ for psychotic smoking, alcohol use, dependent
symptoms psychotic symptoms people
at baseline, social
relation

BPRS, Brief Psychiatric Rating Scale; DIS, diagnostic interview schedule; DIS-C, Diagnostic Interview Schedule for children; ICD, International Classification of
Diseases; M-CIDI, Munich-Composite International Diagnostic Interview; PSQ, Psychosis Screening Questionnaire; SCL-90, Symptoms checklist 90.
Systematic reviews on cannabis and psychosis
313

© 2009 Australasian Professional Society on Alcohol and other Drugs

314 S. Minozzi et al.
• Exposure was assessed considering both ‘ever use’
and frequency, and the way to assess exposure was
quite similar; all studies distinguished between spo-
radic and frequent use, also if in a different way;
• The outcome assessed was similar: psychotic
symptoms for six studies [26,28,30,31,33,35] and
schizophrenia or schizophreniform disorders for
two studies [24,26];
• The potential confounders considered were sub-
stantially similar: age, sex, psychotic symptoms at
baseline, use of alcohol and/or other drugs and
socio-economical variables.
Discussion
Four reviews [16,17,19,20] found a consistent associa-
tion between cannabis use and the onset of psychotic
symptoms. Three reviews, which considered adjusted
OR, reported that the association remains, although it is
reduced, after adjusting for confounding factors. On
the other hand, the review that did not perform a meta-
analysis [18] found an inconsistent association between
cannabis use and psychological problems.
Three reviews had many methodological limitations
[16,17,19]. Only two reviews [18,20] met almost all
relevant quality standards but provided inconsistent
conclusions. One review [20] reported and quantified a
clear association between cannabis use and the onset of
psychotic symptoms, whereas the other review [18]
reported an inconsistent association between cannabis
use and psychological problems. These two reviews are
only partially overlapping because they considered dif-
ferent outcomes and different inclusion criteria for
primary studies. The reviews included different studies
and used a different definition of outcomes to reach
discordant conclusions and judgements about the
safety of cannabis use and the implication for public
health.
Pooling longitudinal study results seems to be a rea-
sonable approach because they were quite homoge-
neous for all of the relevant characteristics: population,
exposure and outcome assessment, and confounding
factors. The only big difference was in the length of
follow up, which could cause non-comparability of the
results if a long latency is expected or if a cumulative
dangerous effect of cannabis is supposed, but it is
unknown if these two situations happen. However, the
availability of an overall cumulative estimate of the
association between cannabis use and the occurrence of
psychotic symptoms does not alone prove causality of
the association and can be easily misinterpreted [64].
All of the reviews draw the conclusion that cannabis
is neither a necessary nor a sufficient cause of psychosis
but that it could be a component cause that interacts
with genetic and environmental factors in vulnerable
© 2009 Australasian Professional Society on Alcohol and other Drugs
individuals. Arseneault et al. [16] pointed out that most
studies were unable to determine whether prodromal
manifestations of psychosis precede the onset of can-
nabis use. Henquet et al. [17] concluded that the con-
founding hypothesis can be ruled out because all of the
studies adjusted for confounding (preceding psychotic
symptoms, socio-demographic characteristics, other
drug use) and the association persisted, although
reduced, and that the reverse causality hypothesis can
be ruled out because the association remained signifi-
cant in studies that excluded subjects with preceding
psychotic illness. Semple et al. [19] concluded that the
question of whether cannabis is a precipitating factor
in vulnerable individuals or a causative agent is still
unanswered.
Macleod et al. [18] concluded that the causal nature
of this association is far from clear because of flaws in
the primary studies: a dose–response relationship was
difficult to assess because only binary exposure catego-
ries were examined in many of the studies, and the
reverse causation hypothesis cannot be excluded
because unreported or sub-clinical problems might
have preceded cannabis use, even in studies that
adjusted for psychological symptoms at baseline. More-
over, cannabis use and psychological problems seem
to share common antecedents, and the relationship
between cannabis use and psychosocial problems could
simply reflect this association. Adjusting for confound-
ing factors is useful, but its power to abolish the con-
founding component depends on the completeness and
precision of the measures used. Moore et al.’s review
[20], the best from a methodological point of view,
concluded that the association between cannabis use
and psychosis seems consistent and remains, even if
attenuated by roughly 45%, after adjusting for compre-
hensive lists of confounding factors. The authors
reported that associations are unlikely to reflect reverse
causality because all studies excluded people with psy-
chosis at baseline or were adjusted for. Nevertheless,
the possibility that the observed association between
cannabis exposure and psychosis was a result of unac-
counted for confounding factors cannot be ruled out,
and these uncertainties are unlikely to be resolved in
the near future. The authors concluded that there is
enough evidence to inform people that using cannabis
could increase their risk of developing a psychotic
illness in the near future.
Other authors [65] have discussed the issue of a lack
of specificity as poor criteria for causality, given that
one of the included studies [30] shows an association
only among cannabis addicted individuals and not
among non-addicted individuals; the study also found
an association between tobacco smoking and the occur-
rence of psychosis independent of the use of cannabis.
Therefore, the considered studies are not able to dis-

tinguish between the effect of use from the effect of
dependence, and the effect of tobacco use from the
effect of cannabis use.
Many narrative reviews published on this topic con-
sidered the same original studies, and all of them agree
with the conclusion that the use of cannabis is a com-
ponent cause that can lead to adult psychosis. Genes
are likely to moderate the association between cannabis
use and later psychosis by increasing susceptibility to
psychotic outcomes, but no study has verified an inter-
action effect between candidate genes and cannabis
use. The question of whether cannabis is the precipitat-
ing or causative factor in the development of psychosis
is unanswered.
In conclusion, the findings provided by these reviews
show a potential risk of psychotic symptoms associated
with the use of cannabis, but these results are by no
means definitive as far as the causality of the association
is concerned.
The issue of whether further longitudinal studies
could provide more evidence is still open to debate;
some have argued that Mendelian randomisation could
be of help [17], but others have suggested that such an
approach would require the presence of genetic variants
that are strongly associated with cannabis use, which is
not yet the case [66].
We can conclude that there is insufficient knowledge
to determine the level of risk associated with cannabis
use in relation to psychotic symptoms and that more
information is needed on both the risks of cannabis use
and the benefits of preventive interventions to support
evidence-based approaches in this area.
References
[1] EMCDDA. Annual report on the state of the drugs
problem in Europe 2006. Lisbon: European Monitoring
Centre for Drugs and Drug Addiction, 2006.
[2] Copeland J, Swift W, Roffman R, Stephens R. A random-
ized controlled trial of brief cognitive-behavioral interven-
tions for cannabis use disorders. J Subst Abuse Treat
2001;20:45–52.
[3] Donnelly N, Hall W. Patterns of cannabis use in Australia.
National Drug Strategy Monograph Series No. 27. Can-
berra: Australian Government Publication Service, 1994.
[4] Australian Institute of Health and Welfare (AIHW). Alcohol
and other drug treatment services in Australia: findings
from the National Minimum Data Set 2001–02. Bulletin
no. 10. AIHW cat. no. AUS40. Canberra, 2003:1–2. Aus-
tralian Institute of Health and Welfare.
[5] Davoli M, Pasqualini F, Belleudi V, Bargagli AM, Perucci
CA. Changing pattern of drug abuse among patients enter-
ing treatment in Lazio, Italy, between 1996 and 2003; tran-
sition from heroin to cocaine use. Eur Addict Res 2007;13
(4):185–91.
[6] Budney AJ, Novy PL, Hughes JR. Marijuana withdrawal
among adults seeking treatments for marijuana depen-
dence. Addiction 1999;94:1311–22.
Systematic reviews on cannabis and psychosis 315
[7] Budney AJ, Higgins ST, Radonovich KJ, Novy PL. Adding
voucher-based incentives coping skills and motivational
enhancement improves outcomes during treatment for
marijuana dependence. J Consult Clin Psychol 2000;
68:1051–61.
[8] Stephens RS, Roffman RA, Simpson EE. Adult marijuana
users seeking treatment. J Consult Clin Psychol 1993;
61:1100–4.
[9] D’Souza C, Cho HS, Perry EB, et al. A cannabinoid model
psychosis, dopamine-cannabinoid interaction and implica-
tion for schizophrenia. In: Castle DJ, Murray R, eds. Mari-
juana and madness. Cambridge: Cambridge University
Press, 2004:142–65.
[10] Thornicroft G. Cannabis and psychosis, is there epidemio-
logical evidence of an association? Br J Psychiatry
1990;15:25–33.
[11] Masthers DC, Ghodse AH. Cannabis and psychotic illness.
Br J Psychiatry 1992;161:648–53.
[12] Hall W, Degenhardt L. Is there a specific ‘cannabis psycho-
sis’? In: Castle DJ, Murray R, eds. Marijuana and madness.
Cambridge: Cambridge University Press, 2004:89–100.
[13] Smith F, Bolier L, Cuijpers P. Cannabis use and the risk of
later schizophrenia. Addiction 2004;99:425–39.
[14] Oxman AD, Guyatt GH. Validation of an index of the
quality of review articles. J Clin Epidemiol 1991;44:1271–
8.
[15] Jadad AR, Cook DJ, Browman GP. A guide to interpret
discordant systematic reviews. Can Med Assoc J 1997;
156:1411–16.
[16] Arseneault L, Cannon M, Witton J, Murray RM. Causal
association between cannabis and psychosis: examination of
the evidence. Br J Psychiatry 2004;184:110–17.
[17] Henquet C, Murray R, Linszen D, van Os J. The environ-
ment and schizophrenia: the role of cannabis use. Schizophr
Bull 2005;31 (3):608–12.
[18] Macleod J, Oakes R, Copello A, et al. Psychological and
social sequelae of cannabis and other illicit drug use by
young people: a systematic review of longitudinal, general
population studies. Lancet 2004;363:1579–88.
[19] Semple DM, McIntosh AM, Lawrie SM. Cannabis as a risk
factor for psychosis: systematic review. J Psychopharmacol
2005;19 (2):187–94.
[20] Moore THM, Zammit S, Lingford-Hughes A, et al. Can-
nabis use and risk of psychotic or affective mental health
outcomes: a systematic review. Lancet 2007;370:319–28.
[21] Andréasson S, Allebeck P, Engström A, Rydberg U. Can-
nabis and schizophrenia: a longitudinal study of Swedish
conscripts. Lancet 1987;2:1483–6.
[22] Zammit S, Allebeck P, Andreasson S, Lundberg I, Lewis G.
Self reported cannabis use as a risk factor for schizophrenia
in Swedish conscripts of 1069: historical cohort study. BMJ
2002;325:1199–201.
[23] Zammit S. An investigation into the use of cannabis and
tobacco as risk factors for schizophrenia. PhD thesis,
Cardiff University, 2004.
[24] Arsenault L, Cannon M, Poulton R, Murray R, Caspi A,
Moffitt TE. Cannabis use in adolescence and risk of adult
psychosis: longitudinal prospective study. BMJ 2002;325:
1212–13.
[25] Caspi A, Moffitt TE, Cannon M, et al. Moderation of the
effect of adolescent onset cannabis use on adult psychosis
by a functional polymorphism in the catechol-O-
methyltransferasis gene: longitudinal evidence of a gene
environment interaction. Biol Psychiatry 2005;57:1117–
27.
© 2009 Australasian Professional Society on Alcohol and other Drugs
316 S. Minozzi et al.
[26] van Os J, Bak M, Hanssen M, Bijl RV, de Graaf R, Verdoux
H. Cannabis use and psychosis: a longitudinal population
bases study. Am J Epidemiol 2002;156:319–27.
[27] Fergusson DM, Horwood LJ, Swain-Campbell NR. Can-
nabis dependence and psychotic symptoms in young
people. Psychol Med 2003;33:15–21.
[28] Fergusson DM, Horwood LJ, Ridder EM. Tests of causal
linkages between cannabis use and psychotic symptoms.
Addiction 2005;100:354–66.
[29] Henquet C, Krabbendam L, Spauwen J, et al. Prospective
cohort study of cannabis use, predisposition for psychosis,
and psychotic symptoms in young people. BMJ 2005;
330:11–14.
[30] Weiser M, Knobler HY, Noy S, Noy S, Kaplan Z. Clinical
characteristic of adolescents later hospitalized for schizo-
phrenia. Am J Med Genet 2002;114:949–55.
[31] Tien AY, Anthony JC. Epidemiological analysis of alcohol
and drug use as risk factors for psychotic experiences.
J Nerv Ment Dis 1990;178:473–80.
[32] Wiles NJ, Zammit S, Bebbington P, Singleton N, Meltzer
H, Lewis G. Self reported psychotic symptoms in the
general population: results from the longitudinal study of
the British National Psychiatric. Br J Psychiatry 2006;
188:519–26.
[33] Stefanis NC, Delespaul P, Henquet C, Bakoula C, Stefanis
CN, van Os J. Early adolescent cannabis exposure and
positive and negative dimensions of psychosis. Addiction
2004;99:1333–41.
[34] Rolfe M, Tang CM, Sabally S, Todd JE, Sam EB, Hatib
N’Jie AB. Psychosis and cannabis use in the Gambia. A
case-control study. Br J Psychiatry 1993;163:798–801.
[35] Agosti V, Nunes E, Levin F. Rates of psychiatric como-
rbidity among U.S. residents with lifetime cannabis
dependence. Am J Drug Alcohol Abuse 2002;28:643–
52.
[36] Degenhardt L, Hall W. The association between psychosis
and problematic drug use among Australian adults: finding
from the National Survey of Mental Health and Well-Being.
Psychol Med 2001;31:659–68.
[37] Degenhardt L, Hall W, Lynskey M. Alcohol, cannabis and
tobacco use among Australians: a comparison of their asso-
ciation with other drug use and use disorders, affective and
anxiety disorders, and psychosis. Addiction 2001;96:1603–
14.
[38] Farrell M, Boys A, Bebbington P, et al. Psychosis and drug
dependence: results from a national survey of prisoners.
Br J Psychiatry 2002;181:393–8.
[39] Grech A, Takey N, Murray R. Comparison of cannabis use
in psychotic patients and control in London and Malta.
Schizophr Res 1998;29:22.
[40] Miller P, Lawrie SM, Hodges A, Clafferty R, Cosway R,
Johnstone EC. Genetic liability, illicit drug use, life stress
and psychotic symptoms: preliminary findings from the
Edinburgh study of people at high risk of schizophrenia. Soc
Psychiatry Psychiatr Epidemiol 2002;36:338–42.
[41] Phillips P, Johnson S. Cannabis use is not associated with
the development of psychosis in ‘ultra’ high risk group. Aust
N Z J Psychiatry 2002;36:800–6.
[42] Resnick MD, Bearman PS, Blum RW. Protecting
adolescents from harm: findings from the National Longi-
tudinal Study on adolescent health. JAMA 1997;278:823–
32.
[43] Dornbusch SM, Lin IC, Munroe PT, et al. Adolescent
polydrug use and violence in the United States. Int J
Adolesc Med Health 1999;11 (3–4):197–219.
© 2009 Australasian Professional Society on Alcohol and other Drugs
[44] Guy SM, Smith GM, Bentler PM. Adolescent socialization
and use of illicit substances. Impact on adult health. Psychol
Health 1993;8:463–87.
[45] Stein JA, Bentler PM, Smith GM, Guy SM, Sybille M.
Consequences of adolescent drug use on young adult job
behaviour and job satisfaction. J Appl Psychol 1993;
78:463–74.
[46] Brook JS, Cohen O, Brook DW. Longitudinal
study of co-occurring psychiatric disorders and substance
use. J Am Acad Child Adolesc Psychiatry 1998;37:322–
30.
[47] Brook JS, Whiteman M, Finch SJ, Cohen P. Young adult
drug use and delinquency: childhood antecedents and ado-
lescent mediators. J Am Acad Child Adolesc Psychiatry
1996;35:1584–92.
[48] Brook JS, Balka EB, Whiteman M. The risk for late adoles-
cence of early adolescent marijuana use. Am J Public Health
1999;89:1549–54.
[49] Brook JS, Richter L, Whiteman M, Cohen P. Consequences
of adolescent marijuana use: incompatibility with the
assumption of adult roles. Genet Soc Gen Psychol Monogr
1999;125:193–207.
[50] Brunswick AF, Messeri P. Drugs, life style and health: a
longitudinal study of urban black youth. Am J Public Health
1986;76 (1):52–7.
[51] Brunswick AF, Messeri PA. Life stage, substance use and
health decline in a community cohort of urban African
Americans. J Addict Dis 1999;18:53–71.
[52] Newcomb MD, Scheier LM, Bentler PM. Effects on ado-
lescent drug use on adult mental health: a prospective study
of a community sample. Exp Clin Psychopharmacol
1993;1:241.
[53] Newcomb MD, Vargas-Carmona J, Galaif ER. Drug prob-
lems and psychological distress among a community sample
of adults: predictors, consequences or confound. J Comm
Psychol 1999;27:405–29.
[54] Kandel D, Chen K, Grill A. The impact of drug use on
earning. A life span perspective. Soc Forces 1995;74:243–
70.
[55] Friedman AS, Kramer S, Kreisher C, Granicks S. The
relationship of substance abuse to illegal and violent behav-
iour in a community sample of young adult African Ameri-
can men and women (gender differences). J Subst Abuse
1996;8:379–402.
[56] Kaestner R. New estimate of the effect of marijuana and
cocaine use on wages. Ind Labor Relat Rev 1994;47:454–
70.
[57] Windle M. Mate similarity, heavy substance use and family
history of problem drinking among young adult women.
J Stud Alcohol 1997;58:573–80.
[58] White HR, Loeber R, Stouthamer-Loeber M, Farrington
DP. Developmental association between substance use and
violence. Dev Psychopathol 1999;11:785–803.
[59] Ellickson P, Bui K, Bell R, McGuigan KA. Does early drug
use increase the risk of dropping out of high school? J Drug
Issues 1998;28:357–80.
[60] Ellickson P, McGuigan KA. Early predictors of
adolescent violence. Am J Public Health 2000;90:566–
72.
[61] Bray JW, Zarkin GA, Ringwalt C, Qi J. The relationship
between marijuana initiation and dropping out from high
school. Health Econ 2000;9:9–18.
[62] Bates ME, Labouvie EW. Adolescent risk factors and the
prediction of persistent alcohol and drug use into adult-
hood. Alcohol Clin Exp Res 1997;21:944–50.

[63] Juon HS, Ensminger ME. Childhood, adolescent and young
adult predictors of suicidal behaviours: a prospective study.
J Child Psychol Psychiatry 1997;38:553–63.
[64] No authors. Rehashing the evidence on psychosis and can-
nabis. Editorial. Lancet 2007;370 (9598):1541.
Systematic reviews on cannabis and psychosis 317
[65] Zullino DF, Rathelot T, Khazaal Y. Correspondence.
Lancet 2007;370 (9598):1540.
[66] Zammit S, Moore T, Lewis G. Correspondence. Lancet
2007;370 (9598):1539–40.
© 2009 Australasian Professional Society on Alcohol and other Drugs