Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

The interaction between cannabis use and the Val158Met polymorphism of the COMT gene in psychosis: A transdiagnostic meta – analysis

  • Thomas Stephanus Johannes Vaessen ,

    Contributed equally to this work with: Thomas Stephanus Johannes Vaessen, Lea de Jong

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Lea de Jong ,

    Contributed equally to this work with: Thomas Stephanus Johannes Vaessen, Lea de Jong

    Roles Conceptualization, Data curation, Investigation, Project administration, Writing – original draft, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Annika Theresia Schäfer,

    Roles Conceptualization, Data curation, Writing – original draft

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Thomas Damen,

    Roles Conceptualization, Data curation, Formal analysis

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Aniek Uittenboogaard,

    Roles Conceptualization, Data curation, Writing – original draft

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Pauline Krolinski,

    Roles Conceptualization, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Chinyere Vicky Nwosu,

    Roles Conceptualization, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Florentina Maria Egidius Pinckaers,

    Roles Conceptualization, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Iris Leah Marije Rotee,

    Roles Conceptualization, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Antonius Petrus Wilhelmus Smeets,

    Roles Conceptualization, Writing – review & editing

    Affiliation Student Faculty of Health Medicine and Life Sciences, Maastricht University, Maastricht, The Netherlands

  • Ayşegül Ermiş,

    Roles Data curation, Investigation, Writing – review & editing

    Affiliation Department of Psychiatry, Bakirkoy Mazhar Osman Mental Health and Neurological Diseases Education and Research Hospital, Istanbul, Turkey

  • James L. Kennedy,

    Roles Data curation, Investigation, Writing – review & editing

    Affiliation Neurogenetics Section, Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Ontario, Canada

  • Dorien H. Nieman,

    Roles Data curation, Investigation, Writing – review & editing

    Affiliation Department of Psychiatry, Academic Medical Center, Amsterdam, the Netherlands

  • Arun Tiwari,

    Roles Data curation, Investigation, Writing – review & editing

    Affiliation Neurogenetics Section, Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, Ontario, Canada

  • Jim van Os,

    Roles Conceptualization, Supervision, Writing – review & editing

    Affiliations Department of Psychiatry and Psychology, School for Mental Health and Neuroscience (MHeNS), Maastricht University, Maastricht, The Netherlands, Department Psychiatry, Brain Centre Rudolf Magnus, Utrecht University Medical Centre, Utrecht, The Netherlands, King's College London, King's Health Partners Department of Psychosis Studies; Institute of Psychiatry, London, United Kingdom

  •  [ ... ],
  • Marjan Drukker

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Supervision, Writing – review & editing

    Marjan.Drukker@MaastrichtUniversity.nl

    Affiliation Department of Psychiatry and Psychology, School for Mental Health and Neuroscience (MHeNS), Maastricht University, Maastricht, The Netherlands

  • [ view all ]
  • [ view less ]

Abstract

Background

Neither environmental nor genetic factors are sufficient to predict the transdiagnostic expression of psychosis. Therefore, analysis of gene-environment interactions may be productive.

Objective

A meta-analysis was performed using papers investigating the interaction between cannabis use and catechol-O-methyl transferase (COMT) polymorphism Val158Met (COMTVal158Met).

Data sources

PubMed, Embase, PsychInfo.

Study eligibility criteria

All observational studies assessing the interaction between COMTVal158Met and cannabis with any psychosis or psychotic symptoms measure as an outcome.

Study appraisal and synthesis methods

A meta-analysis was performed using the Meta-analysis of Observational Studies in Epidemiology guidelines and forest plots were generated. Thirteen articles met the selection criteria: 7 clinical studies using a case-only design, 3 clinical studies with a dichotomous outcome, and 3 studies analysing a continuous outcome of psychotic symptoms below the threshold of psychotic disorder. The three study types were analysed separately. Validity of the included studies was assessed using "A Cochrane Risk of Bias Assessment Tool: for Non-Randomized Studies of Interventions".

Results

For case-only studies, a significant interaction was found between cannabis use and COMTVal158Met, with an OR of 1.45 (95% Confidence Interval = 1.05–2.00; Met/Met as the risk genotype). However, there was no evidence for interaction in either the studies including dichotomous outcomes (B = -0.51, 95% Confidence Interval -1.72, 0.70) or the studies including continuous outcomes (B = -0.04 95% Confidence Interval -0.16–0.08).

Limitation

A substantial part of the included studies used the case-only design, which has lower validity and tends to overestimate true effects.

Conclusion

The interaction term between cannabis use and COMTVal158Met was only statistically significant in the case-only studies, but not in studies using other clinical or non-clinical psychosis outcomes. Future additional high quality studies might change current perspectives, yet currently evidence for the interaction remains unconvincing.

Introduction

Interaction between genes and environment may increase the risk to develop outcomes in the psychosis spectrum [1, 2]. Although the causes of the transdiagnostic expression of psychosis remain unknown [3], several risk factors have been identified. First, the transdiagnostic expression of psychosis, however defined, clusters in families [4, 5]. Second, environmental factors such as cannabis consumption also increase the risk of developing psychotic disorder or symptoms [68]. Genes and environment may reinforce each other’s effects; thus, it has been suggested that genetic variation may render an individual more sensitive to the psychotogenic effects of cannabis [9]. One example of gene-cannabis interaction is the hypothesized moderating effect of the catechol-O-methyl transferase (COMT) polymorphism Val158Met (COMTVal158Met) in the association between cannabis use and the emergence of the psychosis phenotype [10, 11]. The prevalence of cannabis consumption among patients with a diagnosis of psychosis is significantly higher than in the general population (42.1% lifetime use vs 22.5% lifetime misuse [12]). However, causal inference is difficult. For example, some patients use cannabis as a form of self-medication or to reduce the side effects of anti-psychotic medication [13]. On the other hand, causality may be plausible. The plant produces several compounds classified as cannabinoids, Δ9-tetrahydrocannabinol (THC) being the major psychoactive component. Animal studies have reported cannabis-associated alterations in dopaminergic neurotransmission both in the prefrontal cortex [14, 15] and in the mesolimbic pathway (reward system) [16].

Recent research suggests that GxE likely involves multiple genetic variants [17]. However, it has been advocated to perform replication studies analysing exactly the same hypothesis, phenotype and methodology to detect type I error [2]. Only the interaction between cannabis and COMTVal158Met has been studied frequently enough to allow meta-analysis, although different study designs were employed. The COMT gene encodes for the enzyme catechol-O-methyl transferase (COMT), which is required for the catabolism of essential monoamines [18]. COMTVal158Met, also known as rs4680 is thought to alter synaptic availability of dopamine in the cortex, leading to memory and attention impairments and altered levels of dopamine signalling in the mesolimbic system, thus possibly moderating the risk of developing hallucinations and delusions [19, 20].

Given the fact that the same environmental and genetic effects appear to impact psychosis across different clinical and non-clinical levels of the psychosis spectrum [4, 21] we studied GxE in models of transdiagnostic expression of psychosis. In order to verify the validity of this approach, sensitivity analyses were conducted in groups of studies with comparable phenotypic outcomes.

When the primary interest of a study is to assess a possible interaction between genetic (COMTVal158Met) and environmental (cannabis) factors impinging on the development of psychosis, the use of case-only designs is one possibility. This method is used in several studies, such as the study by Costas and colleagues [22]. When a case-only study design to provide evidence for gene-environment interaction is used, the main assumption is that the prevalences of the environmental factor and the genotype are independent of each other in the population (no gene-environment correlation). The basic design is a simple 2x2 table as shown in Table 1, from which ORco (odds ratio in Case-only design) can be calculated. To interpret the ORCO in the context of the case-only design, ORCO is taken as a function of the OR of the exposure alone (ORe), the genotype alone (ORg), and the interaction effect OR (ORg+e) as would be examined in a case-control design. The formulae underlying the ORCO is: ORCO = ORg+e/((ORe * ORg)* Z), where Z describes the odds ratio (OR) between exposure and genotype in the control group [23]. Since the main assumption of the case-only design is that exposure and genotype are independent of each other, Z = 1. At this point, the ORCO obtained from the case-only design is no different than the synergy index obtained from a case-control design. Therefore, ORCO describes departure from only multiplicative effects between the environmental exposure and genotype, similar to the regression coefficient for interaction in a case-control data set.

According to the continuum hypothesis, psychotic symptoms as an outcome of psychosis risk factors should be present not only in subjects diagnosed with psychotic disorder or schizophrenia, but also in subjects from the general population that do not fulfil the clinical criteria [24] and in people considered at ultra-high risk of psychosis [25, 26]. Ultra high risk populations are located on the continuum between the case-only and case-control studies on the one hand and the general population studies on the other. Attenuated psychotic experiences in an ultra high risk (UHR) sample often co-occur with common mental disorder or affective or anxiety symptoms [27]. Under this scenario, studying psychotic symptoms in the general population, in those at ultra-high risk and studying the full blown disorder would hypothetically provide the same evidence for risk factors. This is the reason that in the search the definition of the outcome was broad, including the transdiagnostic expression of psychosis.

Aim

The hypothesis was that evidence from gene-environment interaction between cannabis use and the COMTVal158Met polymorphism would be apparent in studies of the transdiagnostic expression of psychosis. The present meta-analysis aimed to provide an overview of the current literature on this GxE interaction, as well as to provide pooled measures of this interaction. The primary measure of the interaction term between cannabis use and COMTVal158Met impacting the development of psychosis was analysed and evaluated. Because the available studies used three different types of designs and associated phenotypic measures, generalised measures were calculated for each type of study separately: case-only, dichotomous clinical outcomes and continuous outcomes. Only studies using a white ethnic group were included in the present meta-analysis, because the moderating effects of COMTVal158Met likely differ between different ethnic groups [28].

Methods

Data sources

The meta-analysis was conducted according to the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines [29]. In order to identify all suitable studies, the databases PubMed, Psychinfo and Embase were searched up to July 18th of 2017. Various combinations of the main keywords specifically “psychotic disorders”, “schizophrenia”, “psychosis”, “psychotic”, “catechol-o-methyltransferase”, “COMT”, “Val158Met” and “cannabis” were used. The Medical Subject Heading (MeSH) terms further helped with the specification of the search results. Subsequently, the ‘OR’ term was used to combine synonyms in order to yield more results. Second, the ‘AND’ term was used to obtain hits that include at least one term of all three categories (psychosis phenotype, COMT, cannabis). Table 2 illustrates the results of the PubMed search, in particular the search terms used on the left as well as the number of articles found on the right side.

Inclusion and exclusion criteria

All publication years in the databases and articles in English, German and Dutch were included in the search. In addition, no exclusion was conducted based on the type of study. It was decided beforehand that studies would be excluded in case of insufficient data. When information was not available (and was not provided after an e-mail and two reminders to the authors) the study was excluded. In order to ensure methodological quality of the studies, the risk of bias was assessed by two independent reviewers using the ACROBAT-NRSI (A Cochrane Risk of Bias Assessment Tool: for Non-Randomized Studies of Interventions) [30]. Socioeconomic status, age and sex were a priori set as important confounding factors and childhood trauma was identified as a co-exposure.

Data extraction

Selected studies were stratified based on their type of outcome: case-only, dichotomous outcome or continuous outcome, given that pooling of odds ratios and linear regression coefficients in a single meta-analysis is complex. Data on the cannabis-COMTVal158Met interaction were extracted (each study by 2 authors, independently).

Cannabis was a dichotomous variable in all studies. The present study defined Met/Met as the risk genotype (Met/Met = 1; Val/Val and Val/Met = 0), following the majority of the studies. Results of the case-only study defining Val/Val as the risk genotype [31] were recoded to obtain a 2x2 table similar to the other studies; conservatively ignoring the possibility of flip-flop [32]. As opposed to the case-only, all studies analysing dichotomous and continuous outcomes analysed COMTVal158Met coded 0, 1, 2. For the present analysis, all results were recoded so that 0 was Val/Val, 1 was Val/Met and 2 was Met/Met. COMTVal158Met was analysed as a continuous variable, suggesting equal effect sizes when comparing Val/Met to Val/Val and Met/Met to Val/Met.

For case-only studies, the odds ratios of the 2x2 cannabis by COMTVal158Met table (Table 1) were extracted as a primary measure of the GxE interaction. When the authors did not provide the odds ratio, it was calculated from the 2x2 crosstabs. For the dichotomous outcomes, the coefficient (obtained from logistic regression) of the Cannabis X COMTVal158Met interaction and its standard error were extracted (i.e. the exponent of the odds ratio usually reported after logistic regression). For the continuous outcomes, linear regression coefficients of the Cannabis X COMTVal158Met interaction term and its standard error were extracted.

In addition, variables that could potentially modify the results were extracted, including diagnosis of patients (e.g. DSM IV), assessment of cannabis exposure and sex. Hypothetically, those variables can modify the association as defined in the research question.

Statistical analysis

All analyses were performed using Stata version 13 [33]. The Stata command metan generated forest plots for each group of studies using random effects (DerSimonianLaird method).

The metan command also provided the between study variance (tau-square) and the Higgins I-square which is a measure of heterogeneity. When sufficient data was available (case-only studies), modifiers were analysed using the metareg command (diagnosis using DSM IV or revised DSM IV, cannabis use never versus ever or less stringent criteria, >70% male sex vs <70% male sex). The metainf command was used to check single study effects. Finally, publication bias was tested using the metafunnel command to generate a funnel plot, the metatrim command to identify the possibility of unpublished negative findings and to control for that (trim and fill) [34] as well as the metabias command to obtain Egger's test for small study effects.

Because only three studies analysing continuous outcomes were included (4 samples), a meta-regression model including the modifying effect of the variable study (using 2 dummies) was analysed rather than the three modifiers, separately. Study was not a modifier (F = 5.20, df = 2,1, p = 0.30), but visual inspection showed that results from the ultra-high risk population [35] were different.

Sensitivity analyses

The present meta-analysis aimed to study the full psychosis spectrum, i.e. at the level of full blown disorder as well as at the level of psychotic symptoms below the clinical threshold. When studies with samples at different levels of the psychosis spectrum were included in the same meta-analysis, sensitivity analyses were performed including the majority of studies assessing the same population type. For example, the study analysing an ultra-high risk population was excluded in the continuous outcomes sensitivity analysis, the studies with phenotypes at the level of psychotic experiences in the general population remaining in the analysis.

Results

A total of 41 articles were selected based on the inclusion of the three search terms (Table 2, Fig 1). After searching PubMed, searching Psychinfo and Embase didn't provide any extra studies. After the initial search, backward citation tracking was utilised in order to ensure that all relevant studies were identified. However, no further research articles were found. After applying the aforementioned exclusion criteria, 13 articles were included in the final selection (7 case-only studies [22, 28, 31, 3639], 2 studies analysing dichotomous outcomes [10, 40] and 3 analysing continuous outcomes [35, 41, 42]). A third study analysing dichotomous outcomes [43] was excluded from the meta-analysis because the data needed for extraction were not provided and the authors did not respond to requests for additional data. Authors of all articles checked for Hardy Weinberg equilibrium in order to ensure that the genotyping was done correctly.

Case-only studies

Table 3 presents background information of the 7 case-only studies (n = 1954). One study presented results using two different populations, which are both included in the meta-analysis [22].

Validity of the case-only studies is presented in Table 4. As explained earlier, the design of a case-only study does not include a control group. The difficulty to find a good control group is one reason for the existence of case-only studies (20). Distribution of confounders in the non-existing control group is per definition the same in the patients. In addition, the question whether controls were sampled from the same population as the patients is not applicable (selection of participants). Three studies [36, 37, 39] did not describe how cannabis use was assessed and, therefore, scored “moderate” on “measurement of intervention” (i.e. the exposures cannabis and COMTVal158Met as defined in the NRSI criteria list [30]). None of the studies used statistical methods to account for missing data, whilst most had incomplete data. A more detailed appraisal of the validity is available upon request.

In case-only studies, the interaction between cannabis use and COMTVal158Met was statistically significant (OR: 1.45, 95% CI: 1.05–2.0; Fig 2). The I-square of 52% implies heterogeneity. Tau-square was 0.084.When omitting all studies one by one, the OR varied between 1.24 and 1.67 and in 3 instances, the OR was no longer statistically significant (metainf, results available upon request).

thumbnail
Fig 2. Forest plot case-only studies (figure shows odds ratios; random effects).

https://doi.org/10.1371/journal.pone.0192658.g002

The Egger test indicated little evidence for publication bias (bias = 1.07, p = 0.086) and the funnel plot in Fig 3 shows some evidence for omitted small negative studies, indicating publication bias. The trim-and-fill method identified 2 missing studies and correcting for this resulted in a small reduction in effect (OR = 1.40, 95% CI 1.04–1.88). Considering that the number of included studies is limited, firm conclusions on publication bias are not possible.

Results from meta-regression showed that neither method of diagnosis (p = 0.73) nor cannabis assessment (p = 0.36), nor high vs low percentages of male sex (p = 0.68) were modifiers.

Dichotomous outcomes

Table 5 presents study characteristics of the three studies analysing dichotomous outcomes (n = 3433). Validity of these studies is presented in Table 3. Only one study [40] controlled for socioeconomic status. One study [43] assessed current cannabis use rather than cannabis use at adolescence (bias of measurement of interventions: moderate).

The forest plot showed no interaction between cannabis use and COMTVal158Met in the dichotomous outcomes (B: -0.51, 95% CI: -1.72–0.70; Fig 4). The study not included in the meta-analysis [43] reported no two-way interaction, but test statistics were not provided. In females, the authors reported the highest psychosis risk in cannabis users who were Val/Val homozygous, but the two-way interaction term was not statistically significant (p = 0.15). This study [43] assessed current cannabis use (at least once a week during a minimum of 2 weeks in the preceding month), which is of limited value as a proxy for cannabis exposure during onset of psychosis.

thumbnail
Fig 4. Forest plot dichotomous outcomes (figure shows regression coefficient of the interaction term, random effects).

https://doi.org/10.1371/journal.pone.0192658.g004

A sensitivity analysis excluding the study analysing psychotic experiences [40] was not performed because then data of only one study were available for analysis [10]. Results of this study were statistically significant with strongest association between cannabis and psychosis in Val/Val subjects, but the confidence interval was wide.

As the data set included only two studies (two rows in the data), testing modifiers and publication bias was not possible.

Continuous outcomes

The three studies analysing continuous outcomes included a total of 1823 subjects (Table 6)[35, 41, 42]. One study presented results in a discovery and replication sample; these were treated as two separate studies in the meta-analysis [41]. Two studies made use of the Community Assessment of Psychic Experiences (CAPE) for the assessment of psychotic experiences. In the present analysis, the total CAPE sum score was used. However, in one study [42] the depressive dimension of the CAPE score was not measured and, therefore, the sum score was slightly different [42]. The third study was performed in an ultra-high risk population and used the Comprehensive Assessment of At Risk Mental States (CAARMS; CAARMS positive symptoms sum score selected for the present meta-analysis), a semi-structured interview specifically designed to identify subjects with an elevated risk to develop a first psychosis [35].

Validity of the two general population studies and the ultra-high risk study population (see below) was rather similar (Table 3). One study [41] did not have substantial missing data after genotyping.

In only one of the four samples a significant cannabis X COMTVal158Met interaction was found. Val/Val was associated with a stronger association between cannabis and positive symptoms (negative regression coefficient in the meta-analysis) [35]. Combining all continuous outcomes resulted in an interaction term close to zero (B = -0.04, 95% CI: -0.16–0.08; Fig 5).

thumbnail
Fig 5. Forest plot continuous outcomes (figure shows regression coefficient of the interaction term, random effects).

https://doi.org/10.1371/journal.pone.0192658.g005

When omitting studies (metainf) B was between -0.02 and -0.76, but never statistically significant. Trim and fill did not identify missed studies and Egger’s test for small study effects was not statistically significant (B = -1.86, p = 0.21).

Because only four samples analysing continuous outcomes were included (3 rows because one study included two samples), a meta-regression model including the modifying effect of the variable study (using 1 dummy) was analysed rather than the three modifiers, separately. Although visual inspection of the forest plot showed that the study in the ultra-high risk sample with the significant cannabis X COMTVal158Met interaction was different [35], the variable study (p = 0.30) nor the variable ultra-high risk (p = 0.08) were significant modifiers.

Sensitivity analysis

A sensitivity analysis excluding the study in the ultra-high risk population showed similar results, but the confidence interval was smaller (B = -0.02, 95% BI -0.06–0.02).

Discussion

A significant cannabis X COMTVal158Met interaction was found in the case-only studies (OR = 1.43, 95% CI = 1.01; 2.04; Met/Met as the risk genotype), but not in the dichotomous outcomes (B = 0.51; CI -0.70; 1.72) or the continuous outcomes (B = -0.04; CI -0.16; 0.08).

Methodological issues

The strength of the present study is the power. Using 13 studies in 3 different meta-analyses, we were able to analyse the cannabis X COMTVal158Met interaction including high numbers of individuals (n = 1954, 3433, and 1823, respectively). In order to show interaction effects a very high power is needed. The present null finding is not a consequence of power because the effect size was relevant only in the case-only studies and decreased with the increase of the validity of the included studies in the other two meta-analyses. Thus, it was possible to find an unequivocal null finding. However, the present study has some limitations. First, despite the praised efficiency to find interaction in case-only studies [23], this design has lower methodological validity than the other studies included in the present meta-analysis. Case-only studies have been reported to overestimate the true effects; in particular when the exposure-genotype independence assumption is violated [44]. It is possible that COMTVal158Met causes behavioural changes in cannabis use, causing an imbalance between COMTVal158Met and cannabis that is not a result of the hypothesized interaction. Although the other study types did not provide evidence for gene-environment correlation (see below), the impact of a small violation of the assumption forces us to interpret the case-only results with caution. In addition, this design cannot analyse the main effects of cannabis nor genotype alone, but only the interaction. Conversely, only in the case-only data could we assess the impact of modifiers and publication bias, because this was the largest group of studies.

Second, unfortunately, the number of studies was too small to perform a meta-analysis on cannabis X COMTVal158Met in non-white ethnic groups. Thus, the present null-finding is only valid for the white ethnic group.

Furthermore, the included studies were rather heterogeneous as is often the case in meta-analyses. Not only full-blown psychosis, but also attenuated psychotic symptoms were studied. Although this can be seen as a strength when findings are replicated across various populations, it makes interpretation more difficult when findings are different. However, because only in the design with the lowest quality (case-only) significant cannabis X COMTVal158Met interaction was found, we feel that the present null-finding is valid. The study in the ultra-high risk population also showed this interaction, but this interaction was reversed.

Finally, most included studies were cross-sectional. Case-control studies when analysing DSM-diagnoses as well as cohort studies when analysing continuous outcomes are scarce. All studies except one analysed cannabis use during adolescence, introducing a longitudinal element in the cross-sectional data. However, when cross-sectional results do not show an interaction, longitudinal data are also unlikely to show interaction.

Flip-flop

There is a debate on the existence of flip-flop and how to analyse data when flip-flop is observed [32]. Flip-flop describes the phenomenon that in different populations a different allele of the single nucleotide polymorphism (SNP) is associated with the disease. This is the case when not the studied SNP, but a SNP close to that location mediates the risk for a disease. In other words, due to rare cross-over events during conception the disease can be associated with either the one or the other allele of the SNP in different populations. In the present paper, we conservatively assumed no flip-flop [32, 45].

While the pooled effect in the case-only studies identified Met/Met as the risk genotype, in one case-only study Val/Val was the risk genotype [31] n.s.). In addition, in two of the four continuous outcome samples ([42] n.s., [35]) and two dichotomous outcome studies ([10] n.s. [43] females excluded) Val/Val was identified as the risk, while the other results were null findings.

When results of all these studies were reversed as suggested by the flip-flop theory, the OR for interaction in case-only studies was 1.55 (1.14–2.10), while the regression coefficients for interaction were 0.51 (-0.70–1.72) and 0.01 (-0.12–0.13) in studies with dichotomous and continuous outcomes, respectively. Thus, when assuming flip-flop results are rather similar to the original results.

Ultra-high risk

Given that ultra-high risk samples majorly consist of individuals with ‘attenuated psychotic symptoms’, that conceptually and psychometrically resemble the concept of ‘psychotic experiences’ in general population studies [46, 47], the ultra-high risk sample and the general population samples were combined in the same meta-analysis. Although a sensitivity analysis excluding the ultra-high risk paper yielded very similar results (B = -0.02; CI -0.06; 0.02), results in the ultra-high risk group were rather different (B = -1.96; CI -3.16; -0.76; meta-regression coefficient of the difference between general population and ultra-high risk = n.s.). One explanation is that the outcome parameter in the ultra-high risk study was total score on the positive items of the CAARMS whereas in the other continuous outcome studies, CAPE total score (that also includes negative and depressive symptom items) was used. However, the three CAPE scores in practice are very strongly correlated with each other (in the order of 0.7–0.8)[48], suggesting they tap into the same underlying dimension and making it unlikely that use of total CAPE score explains differential findings with other outcomes. In addition, the ultra-high risk study interviewed subjects using the CAARMS, while the general population studies used the CAPE self-report.

Discussion of main findings

There may be two reasons why results of the case-only meta-analysis are different from the other results. First, the dichotomous and continuous outcomes meta-analyses analysed studies with a superior study design compared to the case-only studies. Second, case-only designs may overestimate the true effects [44] as was discussed above. Despite the large numbers of subjects included in the case-only meta-analysis, the lower limit of the confidence interval was close to no effect (OR = 1). Correction for publication bias in the case-only studies decreased the effect, further supporting the null-finding in the other study types. Moreover, the low replication rate of gene–environment interactions in general is evidence for publication bias in this research area [49], but despite this the overall result points at a null finding.

By contrast, some results of the dichotomous and continuous outcome results that were not included in the data set for the meta-analysis did suggest interaction between cannabis and COMTVal158Met. For example, when continuous outcomes were standardised, the cannabis X COMTVal158Met interaction in the sum score of positive symptoms was more than 4 times as large as in the total sum score and statistically significant (30). Second, two continuous outcome studies included child maltreatment as a third interaction variable in a three-way interaction. When unravelling the regression models including the two-way interaction and the three-way interaction, the two-way interaction in the Vinkers discovery sample was 0.39 (p = 0.006) and 0.1 in non-abused and abused children respectively, while the two-way interaction in the Alemany sample was -0.452 (n.s.) and -0.755 (significance unknown) in non-maltreated and maltreated children [42]. In the Vinkers replication sample both two-way and three-way interaction terms were close to zero [41].

Third, Caspi analysed various dichotomous outcomes [10]. Diagnosis of schizophreniform disorder (B = 1.26, p = 0.025) was included in the present meta-analysis. The cannabis X COMTVal158Met interaction was not statistically significant when analysing self-report of psychotic symptoms and evidence of hallucinatory experiences (B = 0.88, p = 0.49 and B = 0.73, p = 0.21 respectively).There was no interaction when studying delusional beliefs and informant reports of psychotic symptoms.

Additionally, Pelayo—Teran [31] showed a related interaction. Cannabis prevented the protective effects of the met variant of COMTVal158Met. Age of onset is earlier in both patients with the val variant and in patients with the met variant using cannabis. Finally, one study reported a cannabis X COMTVal158Met interaction in females only, but this interaction was not statistically significant (31).

Similar to the case-only meta-analysis, most above-mentioned results do suggest evidence for gene-environment interaction. However, results are isolated and inconsistent. Moreover, if similar analyses in subgroups of other studies resulted in a null finding, publication bias is likely [49]. Thus, there is little evidence for an interaction between cannabis and COMTVal158Met in the psychosis phenotype.

Support in the literature

Although main effects of COMTVal158Met have been reported earlier [50], recent work showed that the association between COMTVal158Met and schizophrenia may be inconsistent [51, 52]. However, the presence of a main effect is not a condition to warrant studying interaction. The number of individual studies on the cannabis X COMTVal158Met interaction as well as the ongoing debate warranted a meta-analysis. In addition, while GWAS showed small effects of SNPs, both individual and cumulative, the importance of gene-gene and gene-environment interactions has been emphasized [53]. For example, the COMTVal158Met val/val variant may increase vulnerability to cannabis [53] or cannabis may be associated with epigenetic modulation of the COMTVal158Met gene [54].

Besides the epigenetic modulation various other mechanisms have been suggested, but there is no agreement. Two recent publications provide a summary of the current state of knowledge [52, 55]. In short, COMTVal158Met is involved in dopamine regulation in the brain. One of the cannaboid receptors (CB1) reacts not only to endogenous cannabinoids, but also to THC, thus potentially establishing a link between THC and dopamine levels. The COMTVal158Met val/val variant may significantly worsen the effects of THC on for example cognition through its impact on dopaminergic neurotransmission [52].

An interaction between cannabis and COMTVal158Met is plausible when examining experiments in humans and laboratory animals. First, several rodent studies have addressed this subject. O’Tuathaigh and colleagues [56] showed that THC administration in COMTVal158Met knockout mice was more strongly associated with indicators for psychosis related phenotypes in humans than in wild type mice. Furthermore, Batalla and colleagues [57] studied neuroanatomical changes after cannabis use in a neuroimaging study. They showed that these changes were modulated by the COMTVal158Met gene [57]. A laboratory study in humans indicated that the behavioural response to THC is moderated by COMTVal158Met [58]. On the other hand, COMTVal158Met did not impact on the association between direct THC administration and the CAPE total score [59]. The above-mentioned results on mechanisms are scarce and the research on the mechanisms why cannabis causes psychosis is still in its infancy [2].When mechanisms of the main effect are unclear, mechanisms of the gene–environment interaction are even more difficult to study. The lack of undisputed mechanisms further supports the null finding.

As stated in the introduction, gene-environment interaction in psychosis and psychotic disorders does not only include cannabis X COMTVal158Met. Besides the association between cannabis and psychosis, multiple other environmental factors may play a role. Examples include childhood trauma [60] and urbanicity [61]. Initially, researchers tried to identify a single or a limited number of locations on the DNA as risk loci for various diseases (association studies), including the psychotic phenotype. The COMTVal158Met gene was among the few that were widely studied and, therefore, this was the only genetic variation that could be included in the current meta-analysis. Currently, a large number of genetic loci with small effect sizes are associated with the psychotic phenotype. For this reason, the polygenic risk sore for schizophrenia was constructed [62]. Instead of analysing each locus individually, currently the focus is on the polygenic risk score, to avoid multiple testing and to increase power. Thus, the cannabis X COMTVal158Met interaction is only one of multiple gene—environment interactions that are plausible to co-exist. Although including the schizophrenia polygenic risk score in gene-environment studies is advocated [63], these types of studies are scarce. A recent study reported that the polygenic risk score defining genetic risk for schizophrenia was a modifier in the association between cannabis and brain maturation in males [64]. In contrast, a pilot study did not show evidence for interaction between polygenic risk score and childhood trauma in psychosis [65]. Thus, although there is no evidence for an interaction between cannabis and COMTVal158Met, interaction with other genetic risk factors has not been studied. More research using the polygenic risk score of schizophrenia or other more sophisticated genetic assessments are needed before this can be analysed in a meta-analysis.

Gene environment correlation

It has been argued that gene-environment interaction can only be studied when there is no gene-environment correlation [1]. Not all studies included in the meta-analyses reported on gene-environment correlation. In the case-only studies, this analysis is impossible because it overlaps with the assumption used for this type of study. In the other six studies, there was no gene-environment correlation in 2 studies [10, 35, 40, 42]. The other authors did not report gene-environment correlation [41, 43]. Because more than half of the samples did not have gene-environment correlation in their data, we feel that gene-environment correlation in this meta-analysis is unlikely.

Conclusion and suggestions for further research

In conclusion, the present meta-analysis did not show evidence for an interaction between cannabis and COMTVal158Met when studying psychotic symptoms or psychotic disorder. For future studies, multiple other factors should be taken into account. The analysis of gene-environment interplay may provide useful information about the development and treatment of psychotic disorders.

In the future, researchers should invest in higher quality research designs rather than performing another case-only study. In addition, as has been advocated previously [2, 49], the present results show once again that replication is always highly needed; a single positive study result can hardly be seen as evidence. When replications are performed, we would urge researchers to use a standard set of instruments. That would make pooling in a meta-analysis easier. Currently, for the diagnosis of psychotic disorder, DSM IV or 5 or the Structured Clinical Interview (SCID) for DSM-IV are exchangeable. When assessing psychotic symptoms or attenuated psychotic symptoms in the general population and in ultra-high risk populations the preferred instruments are the CAPE and the CAARMS, respectively. Therefore, we would advise to include those instruments in future research.

Supporting information

S3 File. Excel data dichotomous outcome studies.

https://doi.org/10.1371/journal.pone.0192658.s003

(XLSX)

S4 File. Excel data continuous outcome studies.

https://doi.org/10.1371/journal.pone.0192658.s004

(XLSX)

Acknowledgments

We thank the authors of the original articles who provided extra data, so that we could include their paper in the analyses. In particular S. Alemany, L. Fañanas and C.H. Vinkers.

References

  1. 1. Van Os J, Rutten BP, Poulton R. Gene-environment interactions in schizophrenia: review of epidemiological findings and future directions. Schizophr Bull. 2008;34(6):1066–82. Epub 2008/09/16. doi: sbn117 [pii] pmid:18791076; PubMed Central PMCID: PMC2632485.
  2. 2. Van Winkel R, Kuepper R. Epidemiological, neurobiological, and genetic clues to the mechanisms linking cannabis use to risk for nonaffective psychosis. Annu Rev Clin Psychol. 2014;10:767–91. pmid:24471373.
  3. 3. Messias EL, Chen C-Y, Eaton WW. Epidemiology of schizophrenia: review of findings and myths. Psychiatric Clinics of North America. 2007;30(3):323–38. pmid:17720026
  4. 4. Linscott RJ, Van Os J. An updated and conservative systematic review and meta-analysis of epidemiological evidence on psychotic experiences in children and adults: on the pathway from proneness to persistence to dimensional expression across mental disorders. Psychol Med. 2013;43(6):1133–49. pmid:22850401.
  5. 5. Jeppesen P, Larsen JT, Clemmensen L, Munkholm A, Rimvall MK, Rask CU, et al. The CCC2000 Birth Cohort Study of Register-Based Family History of Mental Disorders and Psychotic Experiences in Offspring. Schizophr Bull. 2015;41(5):1084–94. pmid:25452427; PubMed Central PMCID: PMCPMC4535626.
  6. 6. Avramopoulos D, Pearce BD, McGrath J, Wolyniec P, Wang R, Eckart N, et al. Infection and inflammation in schizophrenia and bipolar disorder: a genome wide study for interactions with genetic variation. PloS one. 2015;10(3):e0116696. pmid:25781172
  7. 7. Kuepper R, Van Os J, Lieb R, Wittchen HU, Hofler M, Henquet C. Continued cannabis use and risk of incidence and persistence of psychotic symptoms: 10 year follow-up cohort study. BMJ. 2011;342:d738. pmid:21363868; PubMed Central PMCID: PMCPMC3047001.
  8. 8. Rognli EB, Bramness JG, Skurtveit S, Bukten A. Substance use and sociodemographic background as risk factors for lifetime psychotic experiences in a non-clinical sample. J Subst Abuse Treat. 2017;74:42–7. pmid:28132699.
  9. 9. Bayer TA, Falkai P, Maier W. Genetic and non-genetic vulnerability factors in schizophrenia: the basis of the “two hit hypothesis”. Journal of psychiatric research. 1999;33(6):543–8. pmid:10628531
  10. 10. Caspi A, Moffitt TE, Cannon M, McClay J, Murray R, Harrington H, et al. Moderation of the effect of adolescent-onset cannabis use on adult psychosis by a functional polymorphism in the catechol-O-methyltransferase gene: longitudinal evidence of a gene X environment interaction. Biological psychiatry. 2005;57(10):1117–27. pmid:15866551
  11. 11. Russo EB, Jiang H-E, Li X, Sutton A, Carboni A, Del Bianco F, et al. Phytochemical and genetic analyses of ancient cannabis from Central Asia. Journal of experimental botany. 2008;59(15):4171–82. pmid:19036842
  12. 12. Green B, Young R, Kavanagh D. Cannabis use and misuse prevalence among people with psychosis. The British Journal of Psychiatry. 2005;187(4):306–13.
  13. 13. Mane A, Fernandez-Exposito M, Berge D, Gomez-Perez L, Sabate A, Toll A, et al. Relationship between cannabis and psychosis: Reasons for use and associated clinical variables. Psychiatry Res. 2015;229(1–2):70–4. pmid:26235479.
  14. 14. Chen J, Paredes W, Lowinson JH, Gardner EL. Delta 9-tetrahydrocannabinol enhances presynaptic dopamine efflux in medial prefrontal cortex. Eur J Pharmacol. 1990;190(1–2): 259–62. pmid:1963849.
  15. 15. Verrico CD, Jentsch JD, Roth RH. Persistent and anatomically selective reduction in prefrontal cortical dopamine metabolism after repeated, intermittent cannabinoid administration to rats. Synapse. 2003;49(1):61–6. pmid:12710016.
  16. 16. Fadda P, Scherma M, Spano MS, Salis P, Melis V, Fattore L, et al. Cannabinoid self-administration increases dopamine release in the nucleus accumbens. Neuroreport. 2006;17(15):1629–32. pmid:17001282.
  17. 17. NIMH. Schizophrenia fact sheet: National Institute of Mental Health 2015. Available from: https://infocenter.nimh.nih.gov/pubstatic/NIH%2015-3517/NIH%2015-3517.pdf; Accessed July 2017
  18. 18. Williams HJ, Owen MJ, O'Donovan MC. Is COMT a susceptibility gene for schizophrenia? Schizophrenia Bulletin. 2007;33(3):635–41. pmid:17412710
  19. 19. Goldberg TE, Egan MF, Gscheidle T, Coppola R, Weickert T, Kolachana BS, et al. Executive subprocesses in working memory: relationship to catechol-O-methyltransferase Val158Met genotype and schizophrenia. Arch Gen Psychiatry. 2003;60(9):889–96. pmid:12963670.
  20. 20. Kapur S. Psychosis as a state of aberrant salience: a framework linking biology, phenomenology, and pharmacology in schizophrenia. Am J Psychiatry. 2003;160(1):13–23. pmid:12505794.
  21. 21. Van Os J, Linscott RJ, Myin-Germeys I, Delespaul P, Krabbendam L. A systematic review and meta-analysis of the psychosis continuum: evidence for a psychosis proneness-persistence-impairment model of psychotic disorder. Psychol Med. 2009;39(2):179–95. pmid:18606047.
  22. 22. Costas J, Sanjuán J, Ramos-Ríos R, Paz E, Agra S, Tolosa A, et al. Interaction between COMT haplotypes and cannabis in schizophrenia: a case-only study in two samples from Spain. Schizophrenia research. 2011;127(1):22–7.
  23. 23. Yang Q, Khoury MJ, Flanders WD. Sample size requirements in case-only designs to detect gene-environment interaction. American Journal of Epidemiology. 1997;146(9):713–20. pmid:9366618
  24. 24. Verdoux H, Van Os J. Psychotic symptoms in non-clinical populations and the continuum of psychosis. Schizophrenia research. 2002;54(1–2):59–65. pmid:11853979.
  25. 25. Miller TJ, McGlashan TH, Rosen JL, Cadenhead K, Cannon T, Ventura J, et al. Prodromal assessment with the structured interview for prodromal syndromes and the scale of prodromal symptoms: predictive validity, interrater reliability, and training to reliability. Schizophr Bull. 2003;29(4):703–15. pmid:14989408.
  26. 26. Yung AR, Yuen HP, McGorry PD, Phillips LJ, Kelly D, Dell'Olio M, et al. Mapping the onset of psychosis: the Comprehensive Assessment of At-Risk Mental States. Aust N Z J Psychiatry. 2005;39(11–12):964–71. pmid:16343296.
  27. 27. Rutigliano G, Valmaggia L, Landi P, Frascarelli M, Cappucciati M, Sear V, et al. Persistence or recurrence of non-psychotic comorbid mental disorders associated with 6-year poor functional outcomes in patients at ultra high risk for psychosis. J Affect Disord. 2016;203:101–10. pmid:27285723.
  28. 28. Kantrowitz JT, Nolan KA, Sen S, Simen AA, Lachman HM, Bowers MB Jr. Adolescent cannabis use, psychosis and catechol-O-methyltransferase genotype in African Americans and Caucasians. Psychiatric quarterly. 2009;80(4):213–8. pmid:19633959
  29. 29. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Jama. 2000;283(15):2008–12. pmid:10789670
  30. 30. Sterne J, Higgins J, Reeves B, on behalf of the development group for ACROBATNRSI. A Cochrane Risk Of Bias Assessment Tool: for Non-Randomized Studies of Interventions (ACROBATNRSI),. Version 1.0.0, 24 September 2014:[Available from: https://sites.google.com/site/riskofbiastool/welcome/home; Accessed July 2017
  31. 31. Pelayo-Terán JM, Pérez-Iglesias R, Mata I, Carrasco-Marín E, Vázquez-Barquero JL, Crespo-Facorro B. Catechol-O-Methyltransferase (COMT) Val158Met variations and cannabis use in first-episode non-affective psychosis: clinical-onset implications. Psychiatry research. 2010;179(3):291–6. pmid:20493536
  32. 32. Lin PI, Vance JM, Pericak-Vance MA, Martin ER. No gene is an island: the flip-flop phenomenon. Am J Hum Genet. 2007;80(3):531–8. pmid:17273975; PubMed Central PMCID: PMCPMC1821115.
  33. 33. Statacorp. Stata Statistical Software: Release 13 College Station, TX: StataCorp LP; 2013.
  34. 34. Duval S, Tweedie R. Trim and fill: A simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics. 2000;56(2):455–63. pmid:10877304.
  35. 35. Nieman DH, Dragt S, Van Duin ED, Denneman N, Overbeek JM, de Haan L, et al. COMT Val(158)Met genotype and cannabis use in people with an At Risk Mental State for psychosis: Exploring Gene x Environment interactions. Schizophr Res. 2016;174(1–3):24–8. pmid:27052366.
  36. 36. Zammit S, Spurlock G, Williams H, Norton N, Williams N, O’Donovan MC, et al. Genotype effects of CHRNA7, CNR1 and COMT in schizophrenia: interactions with tobacco and cannabis use. The British Journal of Psychiatry. 2007;191(5):402–7.
  37. 37. De Sousa KR, Tiwari AK, Giuffra DE, Mackenzie B, Zai CC, Kennedy JL. Age at onset of schizophrenia: cannabis, COMT gene, and their interactions. Schizophr Res. 2013;151(1–3):289–90. pmid:24268936.
  38. 38. Estrada G, Fatjó‐Vilas M, Munoz M, Pulido G, Minano M, Toledo E, et al. Cannabis use and age at onset of psychosis: further evidence of interaction with COMT Val158Met polymorphism. Acta psychiatrica Scandinavica. 2011;123(6):485–92. pmid:21231925
  39. 39. Ermis A, Erkiran M, Dasdemir S, Turkcan AS, Ceylan ME, Bireller ES, et al. The relationship between catechol-O-methyltransferase gene Val158Met (COMT) polymorphism and premorbid cannabis use in Turkish male patients with schizophrenia. In Vivo. 2015;29(1):129–32. pmid:25600541.
  40. 40. Zammit S, Owen MJ, Evans J, Heron J, Lewis G. Cannabis, COMT and psychotic experiences. The British Journal of Psychiatry. 2011;199(5):380–5. pmid:21947654
  41. 41. Vinkers CH, Van Gastel WA, Schubart CD, Van Eijk KR, Luykx JJ, Van Winkel R, et al. The effect of childhood maltreatment and cannabis use on adult psychotic symptoms is modified by the COMT Val 158 Met polymorphism. Schizophrenia research. 2013;150(1):303–11. pmid:23954148
  42. 42. Alemany S, Arias B, Fatjó‐Vilas M, Villa H, Moya J, Ibanez M, et al. Psychosis‐inducing effects of cannabis are related to both childhood abuse and COMT genotypes. Acta psychiatrica Scandinavica. 2014;129(1):54–62. pmid:23445265
  43. 43. Gutierrez B, Rivera M, Obel L, McKenney K, Martinez-Leal R, Molina E, et al. Variability in the COMT gene and modification of the risk of schizophrenia conferred by cannabis consumption. Rev Psiquiatr Salud Ment. 2009;2(2):89–94. pmid:23034243.
  44. 44. Albert PS, Ratnasinghe D, Tangrea J, Wacholder S. Limitations of the case-only design for identifying gene-environment interactions. American Journal of Epidemiology. 2001;154(8):687–93. pmid:11590080
  45. 45. Laird NM, Lange C. The Fundamentals of Modern Statistical Genetics. Gail M, Krickeberg K, S J.M., Tsiatis A, Wong W, editors. New York Dordrecht Heidelberg London: Springer; 2011.
  46. 46. Simon AE, Velthorst E, Nieman DH, Linszen D, Umbricht D, de Haan L. Ultra high-risk state for psychosis and non-transition: a systematic review. Schizophr Res. 2011;132(1):8–17. pmid:21784618.
  47. 47. Kaymaz N, Drukker M, Lieb R, Wittchen HU, Werbeloff N, Weiser M, et al. Do subthreshold psychotic experiences predict clinical outcomes in unselected non-help-seeking population-based samples? A systematic review and meta-analysis, enriched with new results. Psychol Med. 2012:1–15. Epub 2012/01/21. doi: S0033291711002911 [pii] pmid:22260930.
  48. 48. Konings M, Bak M, Hanssen M, Van Os J, Krabbendam L. Validity and reliability of the CAPE: a self-report instrument for the measurement of psychotic experiences in the general population. Acta psychiatrica Scandinavica. 2006;114(1):55–61. pmid:16774662.
  49. 49. Duncan LE, Keller MC. A critical review of the first 10 years of candidate gene-by-environment interaction research in psychiatry. Am J Psychiatry. 2011;168(10):1041–9. pmid:21890791; PubMed Central PMCID: PMCPMC3222234.
  50. 50. Glatt SJ, Faraone SV, Tsuang MT. Association between a functional catechol O-methyltransferase gene polymorphism and schizophrenia: meta-analysis of case-control and family-based studies. Am J Psychiatry. 2003;160(3):469–76. pmid:12611827.
  51. 51. Goes FS, McGrath J, Avramopoulos D, Wolyniec P, Pirooznia M, Ruczinski I, et al. Genome-wide association study of schizophrenia in Ashkenazi Jews. Am J Med Genet B Neuropsychiatr Genet. 2015;168(8):649–59. pmid:26198764.
  52. 52. Fatjó-Vilas M, Prats C, Fañanás L. Chapter e4. COMT Genotypes, Cannabis Use, and Psychosis: Gene-Environment Interaction Evidence from Human Populations, and Its Methodological Concerns. In: Preedy V, editor. Handbook of Cannabis and Related Pathologies: Biology, Pharmacology, Diagnosis, and Treatment. 1: Academic Press; 2017. p. e29–241.
  53. 53. Harrison PJ. Recent genetic findings in schizophrenia and their therapeutic relevance. J Psychopharmacol. 2015;29(2):85–96. pmid:25315827; PubMed Central PMCID: PMCPMC4361495.
  54. 54. Van der Knaap LJ, Schaefer JM, Franken IH, Verhulst FC, Van Oort FV, Riese H. Catechol-O-methyltransferase gene methylation and substance use in adolescents: the TRAILS study. Genes Brain Behav. 2014;13(7):618–25. pmid:24902721.
  55. 55. Cornerstone research collective—medicinal cannabis. Genetics Modify Response to Cannabis: A Look at the COMT Gene Eagle Rock, CA: Cornerstone; 2017 [updated April 27th, 2017]. Available from: https://cornerstonecollective.com/genetics-modify-response-cannabis-look-comt-gene/; Accessed January
  56. 56. O'Tuathaigh CM, Hryniewiecka M, Behan A, Tighe O, Coughlan C, Desbonnet L, et al. Chronic adolescent exposure to Δ-9-tetrahydrocannabinol in COMT mutant mice: impact on psychosis-related and other phenotypes. Neuropsychopharmacology. 2010;35(11):2262–73. pmid:20631688
  57. 57. Batalla A, Soriano‐Mas C, López‐Solà M, Torrens M, Crippa JA, Bhattacharyya S, et al. Modulation of brain structure by catechol‐O‐methyltransferase Val158Met polymorphism in chronic cannabis users. Addiction biology. 2014;19(4):722–32. pmid:23311613
  58. 58. Henquet C, Rosa A, Delespaul P, Papiol S, Faňanás L, Van Os J, et al. COMT Val158Met moderation of cannabis‐induced psychosis: a momentary assessment study of ‘switching on’hallucinations in the flow of daily life. Acta psychiatrica Scandinavica. 2009;119(2):156–60. pmid:18808401
  59. 59. Tunbridge EM, Dunn G, Murray RM, Evans N, Lister R, Stumpenhorst K, et al. Genetic moderation of the effects of cannabis: Catechol-O-methyltransferase (COMT) affects the impact of Δ9-tetrahydrocannabinol (THC) on working memory performance but not on the occurrence of psychotic experiences. Journal of Psychopharmacology. 2015:0269881115609073.
  60. 60. Varese F, Smeets F, Drukker M, Lieverse R, Lataster T, Viechtbauer W, et al. Childhood Adversities Increase the Risk of Psychosis: A Meta-analysis of Patient-Control, Prospective- and Cross-sectional Cohort Studies. Schizophr Bull. 2012;38(4):661–71. Epub 2012/03/31. doi: sbs050 [pii] pmid:22461484.
  61. 61. Marcelis M, Navarro-Mateu F, Murray R, Selten JP, Van Os J. Urbanization and psychosis: a study of 1942–1978 birth cohorts in The Netherlands. Psychol Med. 1998;28(4):871–9. pmid:9723142.
  62. 62. Purcell SM, Wray NR, Stone JL, Visscher PM, O'Donovan MC, Sullivan PF, et al. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature. 2009;460(7256):748–52. Epub 2009/07/03. doi: nature08185 [pii] pmid:19571811.
  63. 63. Iyegbe C, Campbell D, Butler A, Ajnakina O, Sham P. The emerging molecular architecture of schizophrenia, polygenic risk scores and the clinical implications for GxE research. Soc Psychiatry Psychiatr Epidemiol. 2014;49(2):169–82. pmid:24435092.
  64. 64. French L, Gray C, Leonard G, Perron M, Pike GB, Richer L, et al. Early cannabis use, polygenic risk score for schizophrenia and brain maturation in adolescence. JAMA psychiatry. 2015;72(10):1002–11. pmid:26308966
  65. 65. Trotta A, Iyegbe C, Di Forti M, Sham PC, Campbell DD, Cherny SS, et al. Interplay between Schizophrenia Polygenic Risk Score and Childhood Adversity in First-Presentation Psychotic Disorder: A Pilot Study. PLoS One. 2016;11(9):e0163319. pmid:27648571; PubMed Central PMCID: PMCPMC5029892.