A, Type 1 cannabinoid (CB1) receptors are mainly localized on the neuronal presynaptic terminal, possibly mediating retrograde inhibitory feedback and neurotransmitter release. B, CB1 receptors are also localized on the soma of cholecystokinin (CCK)-positive interneurons, contributing to the synchronization of pyramidal cells cortical firing, which is involved in the genesis of γ oscillations. C, Given their high expression on immune cells, type 2 cannabinoid (CB2) receptors are mainly associated with the peripheral functions of the endocannabinoid system, particularly the regulation of immune processes and vascular permeability. However, CB2 receptors are expressed in the human brain, not only on immunomodulatory components (microglia and blood vessels) but also on neural tissues in which they may regulate dopamine-enriched areas. AEA indicates anandamide; 2-AG, 2-arachidonoylglicerol; DGL, diacylglycerol lipase; FAAH, fatty acid amide hydrolase; MGL, monoacylglycerol lipase; NAPE-PLD, N-acyl-phosphatidylethanolamine phospholipase D; PV, parvalbumin.
Weights obtained from random-effects analysis. Horizontal lines represent 95% CIs. The arrow indicates that the upper limit of the CI for that study is equal or superior to the upper limit of the SMD indicated at the bottom of the graph. The diamond shows the overall pooled SMD. The vertical line is the line of no effect, representing no difference between healthy controls and schizophrenia.
eMethods 1. MOOSE Checklist for Meta-analyses of Observational Studies
eMethods 2. Research Protocol
eMethods 3. AMSTAR-2 Checklist
eFigure. PRISMA Flow Diagram
eResults 1. AEA Levels in the CSF: Stage, Antipsychotic Medications, Cannabis Use
eResults 2. AEA Levels in the Blood: Stage, Phase, Antipsychotic Medications, Cannabis Use
eTable 1. Excluded Studies and Reasons for Exclusion
eTable 2. Risk of Bias
Customize your JAMA Network experience by selecting one or more topics from the list below.
Minichino A, Senior M, Brondino N, et al. Measuring Disturbance of the Endocannabinoid System in Psychosis: A Systematic Review and Meta-analysis. JAMA Psychiatry. 2019;76(9):914–923. doi:10.1001/jamapsychiatry.2019.0970
Is the endocannabinoid system abnormal in people with psychosis?
In this systematic review and meta-analysis of 18 studies, a higher tone of the endocannabinoid system was observed in people with psychosis, a finding that was consistent across all stages of illness and independent of antipsychotic treatment and current cannabis use. This increased tone was inversely associated with the severity of psychosis symptoms and was normalized after remission of psychosis.
The endocannabinoid system appears abnormal in people with psychotic symptoms and provides a useful target for illness measurement and treatment.
The endocannabinoid system (ECS) is a lipid-based endogenous signaling system. Its relevance to psychosis is through the association between cannabis use and the onset and course of illness and through the antipsychotic properties of cannabidiol, a potential ECS enhancer.
To conduct a systematic review and meta-analysis of the blood and cerebrospinal fluid (CSF) measures of the ECS in psychotic disorders.
Web of Science and PubMed were searched from inception through June 13, 2018. The articles identified were reviewed, as were citations to previous publications and the reference lists of retrieved articles.
Original articles were included that reported blood or CSF measures of ECS activity in patients with psychotic illnesses and in healthy controls.
Data Extraction and Synthesis
PRISMA guidelines, independent extraction by multiple observers, and random-effects meta-analysis were used. Heterogeneity was assessed with the I2 index. Sensitivity analyses tested the robustness of the results.
Main Outcomes and Measures
The clinical relevance of ECS modifications in psychotic disorders was investigated by (1) a quantitative synthesis of the differences in blood and CSF markers of the ECS between patients and healthy controls, and (2) a qualitative synthesis of the association of these markers with symptom severity, stage of illness, and response to treatment.
A total of 18 studies were included. Three individual meta-analyses were performed to identify the differences in ECS markers between people with schizophrenia and healthy controls. Five studies, including 226 patients and 385 controls, reported significantly higher concentrations of anandamide in the CSF of patients than controls (standardized mean difference [SMD], 0.97; 95% CI, 0.67-1.26; P < .001; I2 = 54.8%). In 9 studies, with 344 patients and 411 controls, significantly higher anandamide levels in blood were found in patients, compared with controls (SMD, 0.55; 95% CI, 0.05-1.04; P = .03; I2 = 89.6%). In 3 studies, involving 88 patients and 179 controls, a significantly higher expression of type 1 cannabinoid receptors on peripheral immune cells was reported in patients compared with controls (SMD, 0.57; 95% CI, 0.31-0.84; P < .001; I2 = 0%). Higher ECS tone was found at an early stage of illness in individuals who were antipsychotic naïve or free, and it had an inverse association with symptom severity and was normalized after successful treatment. Moderate to high level of heterogeneity in methods was found between studies.
Conclusions and Relevance
Testing clinically relevant markers of the ECS in the blood and CSF of people with psychotic illness appears possible, and these markers provide useful biomarkers for the psychotic disorder; however, not all studies accounted for important variables, such as cannabis use.
PROSPERO identifier: CRD42018099863
The endocannabinoid system (ECS) is an endogenous biological signaling system that encompasses lipid-based mediators, their synthesizing and degrading enzymes, and 2 main receptors (Figure 1).1,2 The ECS regulates a series of physiological functions throughout the body, including cognition,3 sleep,4 energy metabolism,5 and inflammation.6 In the brain, the ECS modulates different neurotransmitter systems, including dopamine,7 glutamate,8 and γ-aminobutyric acid.1 The ECS mainly uses 2 major lipid-based endogenous mediators, anandamide and arachidonoyl-sn-glycerol, that act through type 1 and type 2 cannabinoid (CB1 and CB2)2 receptors (Figure 1). The main endocannabinoid-synthesizing enzymes are the N-acyl phosphatidylethanolamine phospholipase D and the diacylglycerol lipase.2 Catabolism is primarily regulated by the fatty acid amide hydrolase and the monoacylglycerol lipase.2
The endogenous activity of the ECS is affected by exogenous cannabinoids, such as THC (delta-9-tetrahydrocannabinol), the main psychoactive components of cannabis, and cannabidiol (CBD).3,9-11 Impairment of the ECS after cannabis consumption has been associated with increased risk of psychotic illness.12-15 By contrast, enhancement of the ECS with CBD has shown anti-inflammatory and antipsychotic outcomes in both healthy study participants and in preliminary clinical trials on people with psychotic illness or at high risk of developing psychosis.11,16-18 Furthermore, the ECS modulates the expression of N-methyl-D-aspartate receptors, which play a key role in the onset of psychotic illnesses.19-22 For these reasons, the ECS may be an important system for understanding the factors associated with and effective treatments for psychosis. However, to date it has not been established whether measuring clinically meaningful ECS activity in the blood or cerebrospinal fluid (CSF) is possible.23-25 We explored this question with a systematic review and meta-analysis of the available evidence on blood and CSF markers of the ECS in psychotic illnesses.
This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline26 (eFigure in the Supplement). The protocol was registered in PROSPERO (eMethods 1-3 in the Supplement).
Search for published studies was conducted from inception to June 13, 2018, in Web of Science and PubMed. Titles and abstracts of retrieved publications were imported into Mendeley (https://www.mendeley.com). All studies that did not meet the inclusion criteria were excluded, with the reasons for exclusion documented in eTable 1 in the Supplement. The entire search process was conducted independently by 2 of us (M.S. and S.Z.), and disagreements at the final stage were resolved by consensus. One of us (A.M.) extracted data from all included studies into an electronic summary table, which was checked by another one of us (N.B.).
We used the Newcastle-Ottawa Scale that has been adapted for cross-sectional studies (eTable 2 in the Supplement) to assess the quality of included studies. Psychotic illnesses included schizophrenia, schizoaffective disorders, first-episode psychosis, and prodromal stage.
We investigated the clinical relevance of ECS modifications in psychosis by (1) comparing the differences in blood and CSF markers of the ECS between patients and healthy controls (primary outcome), (2) reporting data on the correlation analysis between severity of positive and negative symptoms and blood and CSF markers in patients, (3) reporting data on the stage-related modifications of blood and CSF markers in patients and healthy controls, and (4) reporting data on the association between the modifications of blood and CSF markers and the response to treatment (secondary outcomes).
A quantitative synthesis of the differences in ECS markers between healthy controls and patients was provided when 2 or more studies were available. Medians, SEs, and interquartile ranges were transformed to means and SDs after a validated procedure.27 When needed, data were obtained from graphs using a web-based validated tool (WebPlotDigitizer; Automeris).28 All meta-analyses were conducted in Stata, version 13.0 (StataCorp LLC). The standardized mean difference (SMD) was used as summary statistics. Between-study heterogeneity was assessed by calculating Higgins I2 on the basis of Cochrane Q indexes. Because meta-analyses of observational studies are supposed to be characterized by substantial heterogeneity, we used random-effects models.29 To assess the robustness of the results, we performed sensitivity analyses by sequentially removing single studies and rerunning the analysis.30 When feasible, we performed additional descriptive analyses to explore the putative association of a priori–identified clinical subgroups with findings.
Clinical subgroups were identified on the following basis: (1) illness stage, with 3 subgroups: prodromal, first-episode, and multiepisode; (2) illness phase, with 2 subgroups: acutely ill (ie, inpatients or outpatients clearly defined as acutely ill) and clinically stable (ie, outpatients not defined as acutely ill); (3) antipsychotic medication use, with 2 subgroups: antipsychotic-free or antipsychotic-naïve and antipsychotic-treated; and (4) cannabis use, with 2 subgroups: current user (ie, patients included in studies in which current cannabis use was not an exclusion criterion) and not current user (ie, patients with past use or no previous use).
Eighteen studies11,24,31-46 met the inclusion criteria (Table and eFigure in the Supplement). These articles used different methodologies to investigate the ECS. Studies on CSF markers consistently reported data on anandamide levels only. Studies on blood markers reported information on either blood levels of anandamide or expression (messenger RNA/protein) of ECS receptors and enzymes on peripheral immune cells.
Five studies24,32-34,47 reported data on levels of anandamide in the CSF, including 226 patients and 385 controls. Across these 5 studies,24,32-34,47 significantly higher concentrations of anandamide were found in the CSF of patients compared with healthy controls (SMD, 0.97; 95% CI, 0.67-1.26; P < .001), with evidence of moderate heterogeneity (I2 = 54.8%) (Figure 2). This heterogeneity was explained by excluding 1 outlying study,24 with results remaining statistically significant (SMD, 0.81; 95% CI, 0.60-1.02; P < .001; I2 = 0%). Increased CSF anandamide levels were found at any stage of psychosis (prodromal, first-episode, and multiepisode) (eResults 1 in the Supplement) in individuals who were antipsychotic-free and antipsychotic-naive and current users or nonusers of cannabis.
None of the studies into CSF anandamide levels in psychosis reported data on clinically stable patients. Thus, describing data on the basis of illness phase was not possible.
Nine studies24,33-39,47 reported data on levels of anandamide in the blood, including 344 patients and 411 healthy controls. One study35 was excluded from the quantitative synthesis because it analyzed anandamide levels in whole blood, a method believed to produce heterogeneous results, as compared with plasma and serum extraction.24,33 Across the remaining 8 studies, significantly higher concentrations of anandamide were found in the blood of patients compared with controls (SMD, 0.55; 95% CI, 0.05-1.04; P = .03), with evidence of substantial heterogeneity (I2 = 89.6%) (Figure 3). Sequentially removing single studies consistently reduced heterogeneity, with the results remaining statistically significant (SMD, 0.33; 95% CI, 0.11-0.55; P = .003; I2 = 31.1%). This result is in line with findings in the only study excluded from the quantitative synthesis, which showed a substantial increase in whole-blood anandamide levels of patients with acute psychosis compared with controls.35 The overall pooled estimate noted the increased anandamide levels in blood of patients compared with controls, but when the studies were divided according to the a priori–identified clinical subgroups, none of the subgroups substantially differed from the controls. However, larger SMDs were found when individuals with multiepisode and acute illness were compared with controls (eResults 2 in the Supplement), suggesting that the overall findings might have been driven by these 2 clinical subgroups. Anandamide levels in the CSF and blood did not correlate in any of the studies that reported data on both body compartments.24,33,34,47
Five studies reported data on CB1 receptor expression on peripheral immune cells and whole blood of patients compared with healthy controls.40-44 Owing to the lack of extractable data, a meta-analytic synthesis was performed by including only 3 of these 5 studies.41,43,44 Across the 3 studies, involving 88 patients and 179 controls, a significantly higher expression of CB1 receptors was observed in patients compared with healthy controls (SMD, 0.57; 95% CI, 0.31-0.84; P < .001; I2 = 0%) (Figure 4). These 3 studies were all conducted on patients with multiepisode psychosis and included mixed samples of patients who were medicated or unmedicated, clinically stable or acutely ill, and current or not-current cannabis users. The 2 studies not included in the quantitative synthesis reported mixed findings: one showed increased CB142 expression in patients compared with controls, and the other reported no differences.40
Owing to the lack of extractable data, performing a quantitative synthesis on CB2 receptor and ECS expression levels in patients compared with controls was not possible. Five studies reported data on CB2 receptor expression on peripheral immune cells of patients compared with healthy controls,31,40-43 with mixed findings: 2 studies indicated increased expression in patients,41,42 and 3 described no differences31,42,43 (Table). Three studies reported data on ECS-degrading enzymes on peripheral immune cells of patients compared with controls: 2 studies31,35 reported increased expression of degrading enzymes, whereas 1 study indicated no differences.43 Two studies31,43 reported data on ECS-synthesizing enzymes on peripheral immune cells of patients compared with controls. One described reduced expression in patients,31 and one described no differences.43
A downregulating association of cannabis use with CB2 expression in peripheral immune cells was suggested by 1 study.31 However, most studies on peripheral ECS receptors and enzymes reported no association between cannabis use and their findings.
In people with psychosis, the association between severity of positive and negative symptoms, cognitive performances, and ECS tone in the blood and CSF was investigated by 7 studies.24,31,33,34,40,41,46,47 Severity of positive and negative symptoms was associated with (1) decreased anandamide levels in CSF24,33 and (2) increased expression of CB1 and CB2 receptors in peripheral blood mononuclear cells (PBMC).41 Poor cognitive performances were associated with (1) lower anandamide levels in CSF,47 (2) lower anandamide levels in serum,34 (3) higher expression of CB1 and CB2 receptors in PBMC,40,41 and (4) reduced expression of ECS-synthesizing enzymes and increased expression of ECS-degrading enzymes in PBMC.46
Only 2 studies focused on the modifications of the ECS in the context of response to treatment as defined by standardized criteria.11,35 De Marchi et al35 showed that remission of symptoms after 3 months of treatment as usual was associated with a substantial reduction in whole-blood anandamide levels and in messenger RNA levels of fatty acid amide hydrolase and CB2 in PBMC of patients with schizophrenia. Leweke et al11 conducted a 1-month randomized clinical trial in 2012 to compare the efficacy of amisulpride compared with CBD in inpatients with schizophrenia. Cannabidiol induced a pretreatment to posttreatment increase in serum anandamide levels, which was associated with amelioration in positive symptoms.
To our knowledge, this is the first systematic review and meta-analysis of the clinical relevance of blood and CSF markers of the ECS in psychotic illness. The main findings were of increased anandamide levels in the CSF and blood and increased CB1 expression in peripheral immune cells of people with psychotic illness, compared with healthy controls. Higher ECS tone was found at an early stage of illness and in antipsychotic free/naive individuals, had an inverse association with symptoms severity and normalized following successful treatment.
In the whole-group analysis, anandamide in the CSF of patients with schizophrenia was elevated, in comparison to controls (Figure 2). This elevation was found in every stage of psychosis, from first episode to longstanding illness, and was independent of antipsychotic treatment and current cannabis use. Taken together, these findings suggest that anandamide in the CSF might represent a diagnostic marker of schizophrenia and that engagement of the ECS in the brain is protective in schizophrenia.48 Increased anandamide levels in the CSF were associated with reduced severity of both positive and negative symptoms and better cognitive performances in patients. Mechanistically, anandamide release in the brain may provide retrograde inhibition of the mesolimbic hyperdopaminergic state, resulting in reduced positive symptoms.2 The advantage for negative symptoms and cognition may be associated with the regulation of ECS mediators on the excitatory-inhibitory balance in the central nervous system, mediated by N-methyl-D-aspartate49 (Figure 1).
In the whole-group analysis, anandamide in the blood of patients with schizophrenia was elevated, in comparison to controls. None of the a priori–identified clinical subgroups showed significant differences in blood anandamide levels when compared with controls. However, the largest differences were found between patients with multiepisode and acute illness and healthy controls, suggesting that the findings in the overall group may have been driven by the elevation of anandamide concentration in these 2 clinical categories.
Blood levels of anandamide are associated with a number of peripheral sources, including inflammation, activation of the hypothalamic pituitary adrenal axis,50 modifications of the gut microbiome,51 alterations of metabolic functions,52 and medications.33 These peripheral sources may be associated with the increased concentration of anandamide seen in those with more chronic illness and in those with acute psychotic episodes.
The stage- and phase-related increase in serum anandamide does not seem to be associated with antipsychotic medications. Antipsychotic medications seem to downregulate anandamide blood levels (eResults 2 in the Supplement).
These peripheral variables in blood anandamide levels are likely associated with a lack of direct correlation between CSF and blood anandamide levels, as reported in 4 studies. Furthermore, the anandamide-hydrolyzing enzyme fatty acid amide hydrolase is highly expressed in cellular components of the blood-brain barrier,1 providing another possible factor in the discrepancies between peripheral and central anandamide findings.
However, as suggested by correlational analyses, peripheral anandamide levels might still have relevance for central brain processes. Serum anandamide levels and other blood markers of the ECS, such as the expression of cannabinoid receptors, have been associated with the severity of cognitive deficits in patients with schizophrenia. As suggested by Reuter and colleagues,34 the association between peripheral ECS markers and cognition might be the direct frontal regions draining directly into the bloodstream through the dural venous sinuses.34 Furthermore, as shown by 2 long-term studies by Koethe and colleagues,39,47 the engagement of ECS in the periphery is associated with reduced risk of developing schizophrenia in high-risk individuals. These findings further support the hypothesis32 that the activation of the ECS in the periphery is clinically relevant for psychotic illnesses.
Two studies of treatment response have shown that both olanzapine and CBD (a potential ECS enhancer) are associated with reduced anandamide levels in people with schizophrenia.11,35 This finding suggests that both drugs have a comparable action on the ECS, perhaps by reducing the need for the protective feedback operated by the ECS (olanzapine) or by potentiating its outcomes (CBD).
This meta-analytic synthesis showed an increase in CB1 receptors expression in the blood of patients compared with controls. Activation of CB1 has both anti-inflammatory and neuromodulatory outcomes.51 As for anandamide, different research groups have hypothesized that the increase in cannabinoid receptors in the periphery of patients with schizophrenia might represent a compensatory mechanism toward increased inflammation.41,43 A chronic proinflammatory status has been shown to characterize psychotic illnesses, independent of the phase or stage of the disorder,53-55 and is believed to be associated with long-term brain structure and functionality as well as symptom severity and cognitive performance.56,57 Exploring the association of stage, phase, current cannabis use, or antipsychotic medications with CB1 receptor expression was not possible because of the limited number of studies available.
Different methodologies and clinical confounders may be factors in the moderate or high heterogeneity observed. Some studies did not control for relevant variables, such as cannabis use38,44 (Table). Studies into blood levels of anandamide used different biological samples (whole blood,35 plasma,36,37 and serum24,33,34,39,47). In addition, the specific methods in each of the laboratories used to measure anandamide levels or expression of ECS are likely to vary considerably between studies, including the equipment used (eg, immunoassays, bioassays). Different assay procedures may yield different results, and certain platforms for assessment might be more sensitive than others. However, the use of random-effects models in the current study accounted for such between-study heterogeneity.
Shedding light on ECS activity in these disorders is particularly relevant given that exogenous cannabinoids are currently being tested as a novel therapeutic approach in multicentered randomized clinical trials (OPTiMiSE study).58-60 Evidence of the modifications of ECS tone in prodromal states of psychosis is limited but encouraging. If results are confirmed, enhancing ECS tone (eg, using CBD) in individuals with prodromal psychosis could represent a new therapeutic strategy for psychosis prevention,59,60 as suggested by a recent study.60
This systematic review and meta-analysis has limitations. First, most of the clinical subgroup descriptive analyses of anandamide in the CSF and blood were based on a small number of studies. Furthermore, the definition of cannabis use was not always provided in studies, and the noncurrent cannabis user group included individuals who used cannabis in the past, a variable that is known to be associated with ECS functions. Evidence suggests a nonlinear relationship between cannabis use and psychotic illness,10,61 which is likely associated with the complex interaction between endogenous ECS mediators and THC, a partial agonist at the CB1 receptors. Future studies need to use standardized measures to assess cannabis use to clarify the association of cannabis use with blood and CSF markers of the ECS.
Second, all of the studies into anandamide levels in the CSF of patients with schizophrenia were from a single research group and included patients and healthy controls from the same geographic area, likely limiting the generalizability of these findings. Analyses targeting a broader population need to be performed.
Third, despite our efforts to contact all of the authors to retrieve data, this meta-analysis on CB1 receptors in schizophrenia is missing information from 2 studies, and providing a quantitative synthesis of the findings on CB2 receptors and ECS enzymes was not possible. For a similar reason, in the stage-related clinical subgroup descriptive analysis of blood anandamide findings, Wang and colleagues38 was categorized among the multiepisode studies, but it actually included a minority of first-episode psychosis (23.5%).
Fourth, data on ECS prognostic properties are limited and should be further explored by future studies. Future studies should also clarify the diagnostic specificity of these ECS abnormalities, both within psychotic diagnoses and across psychiatric disorders.
This study highlights the diagnostic and prognostic potential of measuring ECS tone in the CSF and blood of patients with psychotic illnesses. Understanding ECS activity in these disorders is relevant owing to the ongoing trials on exogenous cannabinoids as a therapeutic approach.
Accepted for Publication: March 9, 2019.
Published Online: June 5, 2019. doi:10.1001/jamapsychiatry.2019.0970
Correction: This article was corrected on November 25, 2020, to fix a misspelled name in the byline.
Corresponding Author: Amedeo Minichino, MD, Department of Psychiatry, University of Oxford, Oxford OX3 7JX, United Kingdom (firstname.lastname@example.org).
Author Contributions: Drs Cipriani and Lennox have joint senior authorship. Dr Minichino had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Minichino, Senior, Godlewska, Burnet, Lennox.
Acquisition, analysis, or interpretation of data: Minichino, Senior, Brondino, Zhang, Cipriani, Lennox.
Drafting of the manuscript: Minichino, Senior, Brondino, Zhang, Burnet, Lennox.
Critical revision of the manuscript for important intellectual content: Minichino, Godlewska, Burnet, Cipriani, Lennox.
Statistical analysis: Minichino, Brondino, Cipriani.
Administrative, technical, or material support: Burnet.
Supervision: Godlewska, Burnet, Cipriani, Lennox.
Conflict of Interest Disclosures: Dr Godlewska reported other support from Janssen UK outside of the submitted work. Dr Cipriani reported support from the National Institute for Health Research (NIHR) Oxford Cognitive Health Clinical Research Facility, grant RP-2017-08-ST2-006 from NIHR Research Professorship, and grant BRC-1215-20005 from the NIHR Oxford Health Biomedical Research Centre. Dr Minichino is supported by a Medical Research Council (MRC, UK) Doctoral Studentship. No other disclosures were reported.
Disclaimer: The views expressed herein are those of the authors and do not necessarily reflect those of the UK National Health Service, the NIHR, or the UK Department of Health.
Create a personal account or sign in to: