Cannabis is one of the most widely used psychoactive substances worldwide, with 6–7% of the population in Europe and 15·3% of the population in the USA using it each year. There is a global trend towards decriminalisation and legalisation, with 11 US states, Canada, and Uruguay now permitting the sale and recreational use of cannabis in addition to its medicinal use. Given the projected increase in rates of cannabis use,
first described an association between cannabis use and psychotic symptoms, such as paranoia and hallucinations, more than 150 years ago. Subsequently, the main psychoactive constituent of cannabis, Δ9-tetrahydrocannabinol (THC), was shown to induce a significant increase in psychotic (also referred to as positive) symptoms as well as negative symptoms, such as poor rapport, and general psychiatric symptoms, such as depression, relative to placebo.
Central nervous system effects of haloperidol on THC in healthy male volunteers.
highlighting the need to determine the consistency and magnitude of these effects. Furthermore, potential modifiers of these effects, such as dose, previous cannabis use, route of administration, age, sex, tobacco use, and type of THC, have not been systematically evaluated.
Research in context
Evidence before this study
Studies in healthy people indicate that the cannabis constituent Δ9-tetrahydrocannabinol (THC) can induce positive and negative symptoms but findings have been inconsistent. Thus, the magnitude, consistency, and moderators of the induction of schizophreniform and other symptoms by THC remain unclear, including the role of other cannabis constituents such as cannabidiol (CBD). MEDLINE (from Jan 1, 1946, to May 21, 2019), Embase (from Jan 1, 1974, to May 21, 2019), and PsycINFO (from Jan 1, 1806, to May 21, 2019) were searched using the following keywords: (“THC” OR “tetrahydrocannabinol” OR “9THC” OR “9tetrahydrocannabinol” OR “delta9THC” OR “d9THC” OR “delta9tetrahydrocannabinol” OR “dronabinol” OR “marinol” OR “bedrobinol” OR “anandamide” OR “methanandamide” OR “WIN,55,212-2” OR “ACPA” OR “CP55940” OR “bedrocan” OR “spice” OR “JWH-018” OR “AM251” OR “SR161716A” OR “rimonabant” OR “cannabidiol” OR “CBD” OR “cannabinoid”) AND (“BPRS” OR “brief psychiatric rating scale” OR “PANSS” OR “positive and negative syndrome scale”).
Added value of this study
In this meta-analysis of 15 studies, we determined that the acute administration of THC induces positive, negative, and other symptoms associated with schizophrenia and other mental disorders in healthy adults with large effect sizes. Evidence of CBD’s modifying effect is inconclusive. We also found lower induction of psychotic symptoms by THC in studies with more tobacco smokers, and that cannabis use did not moderate the induction of symptoms by THC. These findings extend the literature by systematically showing that THC induces psychotic and other psychiatric symptoms across a range of forms, routes of administration, doses, and settings.
Implications of all the available evidence
Our finding that THC induces positive and other psychiatric symptoms highlights the risks associated with the use of cannabis products, which should be factored into risk–benefit discussions between patients and medical practitioners. This work will inform regulators, public health initiatives, and policy makers considering the medical use of cannabis products or their legalisation for recreational use. Our findings also have implications for mental health policy in terms of education on risks and harm minimisation strategies for products containing THC, and for research into effects in people who might be vulnerable to mental illness.
There is increasing interest in the effects of cannabidiol (CBD), another constituent of cannabis.
Potency of Δ9THC and other cannabinoids in cannabis in England in 2005: implications for psychoactivity and pharmacology.
clarification of the moderating effects of CBD is needed.
We aimed to investigate the psychotomimetic effects of THC and CBD alone and in combination on healthy volunteers to determine the magnitude and consistency of the psychiatric effects of THC and CBD, to investigate the moderating effects of CBD on THC-induced symptoms, and to evaluate the moderating effects of demographic and clinical factors on the induction of symptoms.
Of 517 studies screened, 15 studies met the inclusion criteria for meta-analysis of the acute administration of THC in healthy individuals (figure 1). Four studies on CBD’s effect on THC were identified, which was insufficient for a meta-analysis; therefore, only a systematic review was done. Table 1 provides summary details of the studies included, with further details provided in the appendix (p 8). 331 healthy controls received both THC and placebo conditions (see table 1 for summary of placebos used). Regarding study quality, 13 of 15 studies had scores of 7 or more on the Newcastle Ottawa Scale, indicating low risk of bias.
Impairment of inhibitory control processing related to acute psychotomimetic effects of cannabis.
we were unable to contact the authors. As the study descriptions do not indicate sample overlap, we included them in the main analysis. However, in case there was overlap, we repeated the main analysis excluding the smaller study.
Preliminary evidence of cannabinoid effects on brain-derived neurotrophic factor (BDNF) levels in humans.
), involving 196 participants. THC significantly increased total symptom severity compared with placebo, with a large effect size (SMC 1·10 [95% CI 0·92–1·28], pfigure 2). The result remained significant in all iterations of the leave-one-out analysis (SMC ranged from 1·03 [95% CI 0·92–1·36] to 1·15 [0·95–1·35]; appendix p 21).
No between-sample inconsistency was detected (I2=0%, Cochran’s Q=9·27, p=0·41). Egger’s test did not identify evidence of publication bias (p=0·14). However, trim-and-fill analysis estimated two missing studies on the left-hand side. The SMC was reduced but remained significant after imputation of the two missing studies (SMC 1·02 [95% CI 0·78–1·25], pappendix p 14).
There were no significant linear relationships between the magnitude of placebo–THC differences and age (n=10, β=0·02 [95% CI −0·07 to 0·11], p=0·68), sex (n=10, β=–0·01 [–0·02 to 0·00], p=0·10), tobacco smoking (n=6, β=–0·02 [–0·06 to 0·02], p=0·30), THC dose (n=6, β=–0·05 [–0·26 to 0·16], p=0·65; including studies of intravenous THC only because of insufficient data for analysis for other routes of administration), or study quality (n=10, β=–0·07 [–0·40 to 0·26], p=0·69). Moreover, the induction of total symptoms was not modified by the use of intravenous or inhaled THC (intravenous vs inhaled: Z=–0·90, p=0·37), frequent cannabis use (Z=35, p=0·73), current cannabis use (Z=0·07, p=0·95) or study author (Z=1·06, p=0·29). An insufficient number of studies used BPRS, synthetic THC, or oral THC to enable a moderator analysis of these variables.
Positive symptoms were assessed in 14 studies (15 independent samples) involving 324 participants. THC increased positive symptom severity compared with placebo (SMC 0·91 [95% CI 0·68–1·14], pfigure 3). The result remained significant in all iterations of the leave-one-out analysis (SMC ranged from 0·85 [95% CI 0·63–1·07] to 0·96 [0·75–1·18]; appendix p 22).
There was medium between-sample inconsistency (I2=65·70%, Cochran’s Q=43·73, pappendix p 14). The SMC was reduced but remained significant after imputation of the missing study (SMC 0·87 [95% CI 0·63–1·11], p
Intravenous THC induced more severe positive symptoms than did inhaled THC (Z=2·34, p=0·014; appendix p 15), and studies completed by the D’Souza group were also associated with more severe positive symptoms than studies by other authors (Z=2·89, p=0·0038; appendix p 15). There was an insufficient number of studies to evaluate the effect of oral THC. There was a negative association between tobacco smoking and positive symptoms induced by THC (n=10, β=–0·01 [95% CI −0·02 to 0·00], p=0·019; appendix p 16). Studies with higher quality were associated with a greater effect on positive symptoms (n=15, β=0·26 [95% CI 0·06–0·47], p=0·011; appendix p 16).
By contrast, there were no significant linear relationships between the magnitude of THC–placebo differences and age (n=15, β=0·09 [95% CI −0·01 to 0·19], p=0·069), sex (n=15, β=–0·01 [–0·02 to 0·01], p=0·27), or dose of THC (n=10, β=–0·01 [–0·21 to 0·18], p=0·91; only reported for studies using intravenous administration). Similarly, frequent cannabis use (Z=0·87, p=0·38), current cannabis use (Z=–1·10, p=0·27), and type of THC (synthetic vs purified; Z=–0·73, p=0·47) did not significantly moderate the induction of positive symptoms. An insufficient number of studies used BPRS to enable a moderator analysis of symptom scale used.
Negative symptoms were assessed in 12 studies (13 independent samples) involving 267 participants. THC increased the severity of negative symptoms compared with placebo, with a large effect size (SMC 0·78 [95% CI 0·59–0·97], pfigure 4). The result remained significant in all iterations of the leave-one-out analysis (SMC ranged from 0·72 [95% CI 0·55–0·90] to 0·83 [0·66–1·00]; appendix p 22). THC induced a greater effect on positive symptoms than on negative symptoms (Z=2·06, p=0·039), although this finding did not remain significant when refitting the model with a lower between-symptom correlation coefficient (r=0·1, Z=1·53, p=0·13; appendix p 20).
There was medium between-sample inconsistency (I2=40·57%, Cochran’s Q=24·24, p=0·019). Egger’s test implied significant publication bias (p=0·0069). Trim-and-fill analysis did not identify any missing studies (appendix p 17).
As with positive symptoms, intravenous THC induced greater negative symptoms than did inhaled THC (Z=2·43, p=0·015; appendix p 17). An insufficient number of studies used oral THC to evaluate its modifying effects. Higher mean age of the sample predicted greater negative symptoms induced by THC (n=13, β=0·08 [95% CI 0·01–0·15], p=0·022; appendix p 18).
There were no significant linear relationships between the magnitude of THC–placebo differences and sex (n=13, β=–0·00 [95% CI −0·01 to 0·01], p=0·89), tobacco smoking (n=8, β=–0·00 [–0·01 to 0·01], p=0·41), THC dose (n=9, β=0·03 [–0·12 to 0·18], p=0·73; only assessed in studies of intravenous THC), or study quality (n=13, β=–0·00 [–0·21 to 0·20], p=0·99). Similarly, frequent cannabis use (Z=–0·23, p=0·82), current cannabis use (Z=–0·94, p=0·35), type of THC (synthetic vs purified; Z=–1·35, p=0·18), and study author (Z=0·062, p=0·95) did not significantly moderate the induction of negative symptoms. An insufficient number of studies used BPRS to enable a moderator analysis of symptom scale used.
General symptoms were assessed in eight studies (nine independent samples) involving 162 participants. THC significantly increased general symptoms compared with placebo with a large effect size (SMC 1·01 [95% CI 0·77–1·25], pfigure 5). The result remained significant in all iterations of the leave-one-out analysis (SMC ranged from 0·90 [95% CI 0·70–1·11] to 1·08 [0·81–1·35]; appendix p 22). No significant differences were found between the effect on general symptoms and positive (Z=0·44, p=0·66) or negative symptoms (Z=1·90, p=0·058), although the latter became significant when refitting the model with a higher between-symptom correlation coefficient (r=0·7, Z=2·01, p=0·044; appendix p 20).
There was medium between-sample inconsistency, with an I2 value of 28·90% (Cochran’s Q=20·67, p=0·0081). Egger’s test implied significant publication bias (p=0·0002). Trim-and-fill analysis estimated three missing studies on the left side (appendix p 18). The SMC was reduced but remained significant after imputation of the missing study (SMC 0·85 [95% CI 0·53–1·17], p
There were no significant linear relationships between the magnitude of THC–placebo differences in general symptoms and age (n=9, β=–0·00 [95% CI −0·13 to 0·13], p=0·95), sex (n=9, β=–0·00 [–0·02 to 0·01], p=0·72), tobacco smoking (n=6, β=–0·01 [–0·04 to 0·03], p=0·67), THC dose (n=7, β=–0·08 [–0·33 to 0·17], p=0·52; only assessed in studies of intravenous THC), or study quality (n=9, β=–0·02 [–0·48 to 0·45], p=0·95). Similarly, intravenous and inhaled THC (Z=–0·31, p=0·76), frequent cannabis use (Z=–0·068, p=0·95), current cannabis use (Z=–0·84, p=0·38), and study author (Z=1·06, p=0·29) did not significantly moderate the induction of general symptoms. An insufficient number of studies used BPRS, oral THC, or synthetic THC to enable moderator analyses of these variables.
The effect of CBD on psychopathology compared with placebo was evaluated in two within-person studies and one between-person study (figure 1), with one further study that used the CAPE scale identified by our additional searches (appendix p 19). In the systematic review, there were no significant differences between CBD and placebo in any of the subscales reported (appendix p 11).
Similarly, two within-person and two independent group design studies assessed the effects of CBD on the induction of symptoms by THC (figure 1; table 2; appendix p 12). The first study demonstrated a significant reduction in positive symptoms,
Opposite effects of Δ-9-tetrahydrocannabinol and cannabidiol on human brain function and psychopathology.
albeit in a modest sample. A further study found no significant effect of CBD in the main analysis, but an exploratory analysis demonstrated a significant reduction in symptoms when restricted to participants who had an increase of 3 or more points on the psychotic scale with THC alone.
We demonstrate that acute administration of THC induces significant increases in positive, negative, general, and total symptoms with large effect sizes in adults with no history of psychotic or other major psychiatric disorders. Notably, effect sizes were greater for positive symptoms than for negative symptoms but not for general symptoms, indicating that THC induces positive symptoms to a greater extent than negative symptoms. This result is consistent with findings that symptom severity is greater for positive than negative symptoms in cannabis users.
Affect processing and positive syndrome schizotypy in cannabis users.
Our findings extend these previous findings to show that this is also the case in experimental settings. Although the effect of THC on symptoms remained significant for different routes of administration, the effects of intravenous administration were more pronounced than those of inhalation. This finding indicates that the route of administration modifies THC’s effects, although this association might be confounded by dose or rate of administration. We were unable to test this formally because of a lack of power. It would be useful for future studies to investigate this. Although positive symptoms were also more pronounced in studies by the D’Souza group, all of these studies used intravenous THC, which, as intravenous administration is associated with larger effects, could underlie this association. In addition, lower rates of tobacco use and higher study quality were associated with greater positive symptoms, whereas higher mean age was associated with greater induction of negative symptoms. Notably, positive symptoms were not moderated by dose or previous cannabis use. This result contrasts with findings from several primary studies that report dose–response relationships and evidence of a blunted psychotomimetic effect among regular cannabis users.
Blunted psychotomimetic and amnestic effects of Δ-9-tetrahydrocannabinol in frequent users of cannabis.
The lack of relationship in our analysis might reflect limited power in this analysis and suggests further work is needed to investigate these factors. There was an insufficient number of studies to meta-analyse the effect of CBD alone or the moderating effects of CBD on THC-induced symptoms. Our systematic review found that there is no evidence for CBD having a significant effect on positive, negative, general, or total symptoms. Similarly, although a single, small study (n=6) reported a significant reduction in THC-induced positive symptoms by CBD,
Effects and interaction of delta-9-tetrahydrocannabidiol and cannabidiol on psychopathology, neurocognition, and endocannabinoids in serum of healthy volunteers: influence on psychopathology.
A strength of our analysis is that it focused on experimental studies with placebo control conditions, which avoids the risk of reverse causality and residual confounding factors associated with observational studies of psychotic symptoms in cannabis users.
Dopaminergic function in cannabis users and its relationship to cannabis-induced psychotic symptoms.
However, a number of study limitations should be considered in evaluating our findings. Many of the meta-regression analyses comprised fewer than ten studies and so were underpowered to detect small-to-moderate effects. Thus, we cannot exclude a modifying effect of some variables on our findings, in particular tobacco use and THC dose on the induction of total, negative, or general symptoms by THC, or age or gender on general symptoms, although our analyses suggest that any potential effects are not large. There was a preponderance of male-dominated samples in the studies. Although no effect of sex was identified, future studies should include more females to ensure generalisability. We identified potential publication bias in positive, negative, and total symptom domains. This bias might be due to selective reporting of symptom scales with significant findings. Nevertheless, effect sizes for positive symptoms were positively associated with study quality, suggesting that our findings might be underestimating effect size as a result of the inclusion of lower quality studies that have smaller effect sizes, and findings remained largely unchanged after adjusting for putatively missing studies. Finally, we used summary symptom measures that combine scores across several symptoms, which precludes the analysis of individual symptoms. Future work should focus on the effect of cannabinoids on specific symptoms of interest, such as hallucinations and delusions.
The studies we analysed used doses of THC ranging from 1·25 mg to 10 mg, leading to peak THC blood concentrations of 4·56–5·1 ng/mL when orally administered
Comparison of cannabinoid concentrations in oral fluid and whole blood between occasional and regular cannabis smokers prior to and after smoking a cannabis joint.
Thus, our findings have implications for the 188 million people who use cannabis and other THC-containing cannabinoids worldwide each year, and for the therapeutic use of cannabis and its derivatives. They indicate that use of THC-containing products could induce a range of psychiatric symptoms, including psychotic symptoms such as hallucinations and paranoia.
Effects and interaction of delta-9-tetrahydrocannabidiol and cannabidiol on psychopathology, neurocognition, and endocannabinoids in serum of healthy volunteers: influence on psychopathology.
Thus, currently, the experimental evidence base is not strong for increasing CBD content in cannabis to counter the effects of THC.
Our finding that the induction of psychotic symptoms was lower in people with higher tobacco use could suggest that tobacco use is a protective factor, but further work is needed to test causality and this finding should not be taken as a recommendation to use tobacco to counter the effects of THC. Tobacco smoking is associated with lower brain CB1 receptor levels,
Decreased cannabinoid CB1 receptors in male tobacco smokers examined with positron emission tomography.
which could mean smokers are less sensitive to the effects of THC. The association between lower induction of psychotic symptoms by THC and higher tobacco use might also relate to the upregulation of UDP-glucuronosyltransferase by nicotine,
Exogenous cannabinoids as substrates, inhibitors, and inducers of human drug metabolizing enzymes: a systematic review.
Finally, although this study investigated the acute effects of THC, the magnitude and consistency of effects across symptom domains add to the evidence implicating the endocannabinoid system and cannabis use in the pathophysiology of schizophrenia and other psychotic disorders.
The neural and molecular basis of working memory function in psychosis: a multimodal PET-fMRI study.
In conclusion, these findings demonstrate that the acute administration of THC induces positive, negative, and general psychiatric symptoms with large effect sizes. By contrast, CBD does not induce psychiatric symptoms, and there is inconclusive evidence that it moderates the induction of psychiatric symptoms by THC. These effects are larger with intravenous administration than with inhaled administration, and tobacco smokers have less severe positive symptoms. These findings highlight the acute risks of cannabis use, which are highly relevant as medical, societal, and political interest in cannabinoids continues to grow.
GH, KB, FB, and ODH formulated the study design and literature search. GH and KB collected data. DK, SG, RR, and DCD assisted with data collection. GH, KB, FB, CEG, RM, and ODH contributed to data analysis and data interpretation. GH and ODH wrote the manuscript. GH, KB, and CEG prepared the figures and tables. GH, KB, FB, CEG, RM, DK, SG, RR, DCD, and ODH edited the manuscript.
RR is supported by grants from the Dana Foundation David Mahoney Program, Neurocrine Biosciences, and Clinical and Translational Science Award grant number UL1 TR001863 from the National Centre for Advancing Translational Science, which are components of the US National Institutes of Health (NIH), NIH roadmap for Medical Research. DCD reports grants from the NIH, VA R&D, the Heffter Foundation, the Wallace Foundation, and Takeda (outside the submitted work), and he serves on the Physicians Advisory Board of the Medical Marijuana Program for the State of Connecticut. ODH reports grants from Angellini, AstraZeneca, Autifony, Biogen, Eli Lilly, Heptares, Jansenn, Lundbeck, Lyden-Delta, Otsuka, Sunovion, Rand, Recordati, and Roche (outside the submitted work). All other authors declare no competing interests.