- Top of page
- Material and methods
- Declaration of interest
It is well established that attenuated psychotic symptoms occur in some individuals from the general population [1-3]. In the absence of illness or the need for treatment, these milder forms of psychotic symptoms are referred to as psychotic experiences (PEs) . It has been suggested that clinical and subclinical expression of psychosis share genetic and/or environmental factors in their aetiology . Therefore, the study of the risk factors for PEs would ultimately contribute to the understanding of the aetiology of psychotic disorders.
In this context, both cannabis use [5-7] and childhood adversity [8-10] have been associated with an increased risk of developing psychosis in clinical and non-clinical samples. However, not everyone exposed to childhood adversity develops psychotic symptoms later in life. Similarly, only a minority of cannabis users develop psychotic symptoms suggesting the implication of other factors in this association .
In this regard, several studies have shown that the joint exposure to these two environmental factors, cannabis use and childhood adversity, may increase the likelihood of developing psychotic symptoms to a greater extent than the risk expected for each factor working independently [12-15].
These results are neurobiologically plausible, as both stressful experiences and delta-9-tetrahydrocannabinol (THC), the main psycho-active constituent of cannabis, have been found to increase dopaminergic signalling in the mesolimbic system , which is hypothesized to result in an increased risk of delusions and hallucinations . However, a recent study of a large sample drawn from the general population failed to replicate the interaction effect reported between cannabis and childhood trauma on the risk of developing psychotic symptoms . Individual differences in neurobiological susceptibility to the impact of childhood abuse and cannabis use might help to explain this failure to replicate. Indeed, recent evidence suggest that differential sensitivity to environmental stress occasioned by the Val158Met polymorphism of the catechol-O-methyltransferase (COMT) gene, probably in interaction with other factors, might be underlying psychosis risk [19-21].
The COMT gene encodes the enzyme catechol-O-methyltransferase, which plays an important role in the degradation of dopamine in the brain, and contains a functional polymorphism (COMT-Val158Met) that results in two common variants of the enzyme (Val and Met) . The Val variant is associated with increased COMT activity, which results in a combination of reduced dopamine neurotransmission in the prefrontal cortex and increased levels of dopamine in mesolimbic areas . Individuals carrying the Met/Met genotype have the lowest COMT activity and heterozygotes are considered to be of intermediate activity, as the two alleles are codominant .
In this regard, gene–environment interaction studies have shown that the Val158Met polymorphism of the catechol-O-methyltransferase (COMT) gene moderates i) the association between cannabis use and psychosis [25-27], although some studies failed to replicate the original findings from Caspi and colleagues [For review see:  and ] and ii) the association between childhood trauma and schizotypal traits . However, to our knowledge, no study to date has investigated whether the impact of the joint effect of exposure to childhood adversity and cannabis use on the subsequent development of PEs might be influenced by the COMT-Val158Met polymorphism.
Aims of the study
This study aimed to investigate whether the impact of the childhood adversity and cannabis effects on the development of psychotic experiences varies according to COMT-Val158Met polymorphism genotypes.
- Top of page
- Material and methods
- Declaration of interest
In the current sample, 40.7% of the individuals reported that often or almost always experienced at least one positive PE. For the negative dimension, 47.6% of the sample often or almost always experienced at least one negative PE. Of note, prevalences for some items addressing more severe psychotic experiences were lower. For example, 4.8% of the sample often or almost always felt that they were ‘under the control of some force or power other than themselves’; similarly, 1.8% of the sample often or almost always ‘heard voices talking to each other’ [CAPE; ].
With regard to childhood abuse, 25.5% of the individuals were exposed to at least one abusive event during childhood. Nevertheless, regarding to specific and severe forms of childhood abuse and neglect, only the 9.2% and 10.3% of the sample reported being exposed to sexual abuse and physical neglect respectively.
For cannabis use, 29.1% of the sample used cannabis monthly, weekly or daily.
All the variables included in the model were available for 419 individuals from the total sample. In this final sample, the genotype frequencies for the Val158Met polymorphism of the COMT gene were as follows: Val/Val: 30.3% (n = 127); Val/Met: 48.0% (n = 201); and Met/Met: 21.7% (n = 91). These frequencies did not differ from others described in Caucasian individuals . The Hardy–Weinberg equilibrium was verified for the present population (χ2 = 0.47; df = 2; P = 0.49).
A main effect of childhood abuse was found in both positive (β = 0.09; SE = 0.04; 95% CI .01–0.17; P = 0.047) and negative PEs (β = 0.11; SE = 0.05; 95% CI .01–0.21; P = 0.038). Cannabis use showed a main effect on negative PEs (β = 0.88; SE = 0.44; 95% CI .01–1.75; P = 0.047) but not on positive PEs. However, these main effects did not remain significant after correcting for multiple testing. No main effect was found for the Val158Met polymorphism of the COMT gene on either dimension of PEs.
None of the two-way interactions tested (childhood abuse*cannabis use; childhood abuse*COMT gene or cannabis use*COMT gene) were significant.
However, a significant three-way interaction among childhood abuse, cannabis use and the COMT gene was found in positive PEs [β = −0.30; SE = 0.11; 95% CI (−0.51)–(−0.09); P = 0.006] (Table 1; Fig. 1). This result was significant even after correction for multiple testing. It accounted for 2% of the variance of positive PEs (η2 = 0.2).
Figure 1. Graphic representation of the interaction effect among childhood abuse, cannabis use and the Val158Met polymorphism of the COMT gene on positive psychotic experiences (PEs) corrected for age, gender, schizotypal personality and trait anxiety. Cannabis use and the Val158Met polymorphism of the COMT gene have a negligible effect on positive PEs when individuals are not exposed to childhood abuse or exposed to low rates of such events (red and blue lines). The use of cannabis in individuals exposed to childhood abuse has opposite effects depending on their genotype (purple and green lines). Positive PEs score increases as a function of the number of copies of the Met allele of the COMT gene in those individuals exposed to childhood abuse who do not use cannabis (green line). Thus, Met carriers seem to be especially vulnerable to the effect of childhood abuse on their later development of PEs and cannabis use may have a protective effect. However, in individuals exposed to childhood abuse who use cannabis, a positive PEs score increases as a function of the Val allele copies of the COMT gene (purple line).
Download figure to PowerPoint
Table 1. 1) Main effects, 2) two-way interaction effects and 3) three-way interaction effects of childhood abuse, cannabis use and the COMT Val158Met polymorphism are presented for positive psychotic experiences (PEs) and negative PEs. All the models were corrected by age, gender, schizotypal personality and trait anxiety. Adjusted R2 values (Adj-R2) are presented for each step for positive and negative PEs. Significant results are indicated in bold
| ||Positive PEs||Negative PEs|
|β||SE|| P ||β||SE|| P |
|1) Main Effects|
|Childhood abuse|| 0.088 || 0.044 || 0.047 a || 0.107 || 0.051 || 0.038 a |
|Cannabis use||0.378||0.384||0.325|| 0.883 || 0.443 || 0.047 a |
| COMT ||0.148||0.241||0.541||−0.157||0.278||0.573|
|2) Two-way interaction effects|
|Childhood abuseaCannabis use||0.058||0.089||0.516||−0.063||0.103||0.539|
|3) Three-way interaction effects|
|Childhood abuseaCannabis useaCOMT||−0.303|| 0.110 || 0.006 a ||−0.156||0.129||0.228|
In individuals exposed to childhood abuse who used cannabis, positive PEs score increased as a function of the Val allele dose of the COMT gene. However, among individuals exposed to childhood abuse who did not use cannabis, the positive PEs score increased as a function of the Met allele copies of the COMT gene. When individuals were exposed to low rates of childhood abuse, cannabis use and the Val158Met polymorphism of the COMT gene had a negligible effect on the presence of positive PEs scores.
The log-likelihood ratio test indicated that addition of the three-way interaction term in the third step resulted in a statistically significant improvement in model fit compared to the main effects (χ2 = 12.7; df = 2; P = 0.013).
Additional logistic regression analyses revealed that neither childhood abuse (OR = 1.01; 95% CI .96–1.07; P = 0.671) nor the COMT-Val158Met polymorphism (OR = 1.19; 95% CI .71–1.98; P = 0.513) was associated with cannabis use.
- Top of page
- Material and methods
- Declaration of interest
Rates for PEs and childhood trauma in the current sample were consistent with previous reports in European and North American samples [4, 37, 42] [further details can be found elsewhere [1, 30]]. Also, the rate of individuals using cannabis (monthly, weekly or daily) was 29.1%, which is similar to the rates reported in other European countries .
As previously shown in this sample, childhood abuse was associated with both positive and negative PEs . These findings support the role of childhood abuse in the development of PEs in the general population, as reported in previous research [8-10]. Furthermore, the fact that cannabis use did not show a main effect on positive PEs in the current study may be related to the inclusion of childhood abuse in the model [in univariate analyses cannabis was significantly associated with positive PEs (β = 1.20; SE = 0.44; P = 0.007)]. As previous studies have suggested, explorations of the association between cannabis and psychosis need to consider the effects of childhood trauma as an important potential effect modifier [12, 14]. Nevertheless, both childhood abuse and cannabis use were associated independently with negative PEs. The association between cannabis and negative PEs has been reported previously [7, 44].
The term gene–environment correlation refers to the fact that exposure to an environmental risk factor is not random but is influenced by the individual's genotype. Similarly, environment–environment correlation occurs when the exposure to a given environmental factor is influenced by the previous exposure to another environmental factor [45, 46]. With regard to these mechanisms, additional analyses enabled us to rule out the possibility that childhood abuse increased the likelihood of using cannabis (environment–environment correlation). A gene–environment correlation can also be discarded, as COMT genotypes were not associated with cannabis use.
In accordance with recent evidence, we did not find an interaction between the effect of childhood abuse and cannabis use on PEs . However, we believe that this may be related to the inclusion of COMT genotypes in the analyses, as a significant gene–environment–environment interaction effect was detected. This finding is consistent with previous studies indicating that environmental exposures, in interaction with genetic factors, may induce psychological or physiological alterations that can be traced to a final common pathway of altered dopamine neurotransmission. This pathway facilitates the onset and persistence of psychotic symptoms .
Therefore, our main findings suggest that the psychosis-inducing effects of childhood abuse and cannabis use are moderated by the Val158Met polymorphism of the COMT gene, which supports a gene–environment–environment interaction effect.
This three-way interaction effect indicated that positive PEs showed almost no variation for individuals exposed to low rates of childhood abuse, regardless of their cannabis use frequency or their genotype for the Val158Met polymorphism of the COMT gene. However, among individuals exposed to childhood abuse, cannabis use only increased the likelihood of reporting positive PEs if individuals were carriers of the Val allele of the COMT gene. Furthermore, Met carriers exposed to childhood abuse were more likely to report positive PEs without cannabis use. Thus, our findings suggest that use of cannabis after exposure to childhood abuse may have opposite effects on the development of positive PEs depending on the COMT genotypes.
Although the effect size of this finding is modest (2% of the variance of positive PEs) and requires replication, these results may partially account for previous discrepancies found when examining the possible moderator role of COMT genotypes in the association between cannabis and clinical and non-clinical expression of psychosis. For example, as abovementioned, Kuepper  and colleagues failed to replicate the interaction shown between childhood trauma and cannabis use [12-15]. This discrepancy could be owing to sampling variation or different time of follow-up  but it might be also possible that COMT genotypes play a role in this association considering our results. With regard to the interaction between cannabis and COMT-Val158Met polymorphism, several studies examining different aspects of the psychosis phenotype (psychotic symptoms, psychotic disorders, age of onset or duration of untreated psychosis) have yielded inconsistent results [11, 25-27]. Also, a failure to find such interaction effect has been reported . As our findings suggest that psychosis-inducing effects of cannabis have opposite effects depending on the COMT genotypes but only among those exposed to childhood abuse; future studies testing this gene–environment interaction effect may consider including childhood trauma in this association if the measure is available.
The fact that exposure to both childhood abuse and cannabis was associated with higher scores of positive PEs in Val carriers may be explained by sensitization involving dopaminergic signalling. Evidence from animal studies suggests a possible interaction (exposure to one factor increases sensibility to the effects of the other factor) between stress and THC. Rats living under normal conditions (i.e. access to water and food), that were exposed to THC, showed only minor behavioural changes and no change in dopaminergic transmission . In contrast, under stressful conditions (i.e. isolation and food deprivation), THC administration had marked behavioural consequences and was associated with a significant increase in dopamine uptake . Similarly, it has been shown in humans that the psychosis-inducing effect of cannabis may be stronger in subjects exposed to early stress . Our results indicate that variability in the COMT gene confers different neurobiological vulnerability to cannabis use in the risk of developing PEs. In accordance with previous studies, Val carriers are more vulnerable to the psychosis-inducing effects of cannabis than Met/Met individuals [25-27], but only when exposed to childhood abuse. Consistent with previous studies indicating that Met carriers were more vulnerable to stress than carriers of the Val/Val genotype , Met carriers were vulnerable to the psychosis-inducing effects of childhood abuse, but only when they did not use cannabis. Previous evidence indicates that the risk of psychosis did not increase in Met carriers of the COMT gene who used cannabis . However, in the current study, individuals exposed to childhood abuse who are homozygous for the Met allele appeared to be able to use cannabis without any increase in risk of developing PEs. It might be possible that cannabis may exert some benefit effect in certain individuals. Indeed, it has been suggested that cannabis use alleviate the stress associated with childhood traumatic events and the experience of PEs . However, such conclusions cannot be drawn from our results, thus this result needs replication and must be interpreted with caution.
In this regard, although the findings of gene–environment interaction studies have been exciting, there is increasing concern about the reliability and contribution of such results to the understanding of complex traits such as PEs . Dismissal of gene–environment interaction studies arises mainly as a result of the failure to replicate [52, 53]. As there are powerful reasons to expect that gene–environment interaction effects are involved in the aetiology of complex traits and psychiatric disorders , the debate is more focused on the reliability and clinical relevance of such findings . To prevent false positive results or statistically significant results that may not represent true insights, the current study was developed with an a priori hypothesis that guided the choice of the gene, the polymorphism and the environmental risk factors that were explored. Moreover, as abovementioned, power analyses are specified and correction for multiple testing was applied. Furthermore, the use of cannabis after exposure to childhood abuse had opposite effects on positive PEs depending on the COMT genotypes. This pattern of results coincides with the epidemiological definition of qualitative interaction. A qualitative interaction refers to an inverse or crossover effect from a given variable (e.g. cannabis exposure) according to differences in another variable (e.g. COMT genotypes) [51, 55]. Although these type of interactions have only rarely been observed in medicine, the implications of qualitative or crossover interactions are believed to have a clear biological meaning and be more helpful than the ones derived from quantitative or non-crossover interactions .
The results of this study should be interpreted in the context of its limitations. First, we used a relatively small sample size to detect a three-way interaction, replication in larger samples with higher statistical power are needed to confirm these findings. Second, the characteristics of the sample – young age, educational level, no history of psychiatric treatment – need to be considered when generalizing the present findings. Also, as substance abuse constituted an exclusion criterion, heavy cannabis users, who experience problems in their daily life because of their cannabis consumption, were not included in the study. Therefore, although the sample is drawn from the general population, the representativeness of the sample is limited by these characteristics. Third, no main genetic effects for COMT-Val156Met polymorphism on PEs were found in the current study. As the power to detect interactions is typically lower than the power to detect main effects , well-powered studies should be able to detect statistically significant main genetic effects unless a qualitative interaction effect is detected as is the case for this study. In qualitative interactions, main effects are cancelled out; therefore, the lack of significant main genetic effects in this study should not compromise the reliability of the reported results. Fourth, the cross-sectional nature of the design does not allow causal inference. Fifth, childhood abuse was measured retrospectively, which may constitute an inherent source of bias. Furthermore, this instrument has not been yet validated in Spanish population. That said, the Childhood Trauma Questionnaire has been validated in several European countries including Dutch and Swedish populations [57, 58] and is considered a reliable measure of childhood adversity . Finally, frequency of cannabis use was dichotomously defined in this study, and other parameters that have been related to the expression of psychotic symptoms such as onset, duration or potency of cannabis consumed [7, 34, 59, 60], were not specified. Furthermore, biological samples for confirming drug use by means of laboratory techniques were not available in this study.
Of note, we would like to stress the fact that consistent evidence indicates that cannabis may induce psychosis and/or worse psychotic symptoms [5-7]. Therefore, public health message about the potential risk of cannabis use should not be modified by results indicating that its use may not be harmful for a subgroup of the population.
To conclude, our findings suggest that the psychosis-inducing effects of childhood abuse and cannabis use are moderated by the Val158Met polymorphism of the COMT gene, which supports a gene–environment–environment interaction effect. Cannabis use after exposure to childhood abuse may have opposite effects on the risk of PEs development, depending on the COMT genotypes.