A systematic evidence map of the association between cannabis use and psychosis-related outcomes across the psychosis continuum: An umbrella review of systematic reviews and meta-analyses

While the legal status and public perception of cannabis are currently changing in many countries, one of the important considerations from a public health viewpoint is its potential association with adverse health outcomes such as the development of psychosis. We conducted an umbrella review of systematic reviews and meta-analyses using the AMSTAR-2 to assess the quality of included reviews. We further created an evidence map to visualize and facilitate the overview of the published evidence synthesis on the association between cannabis use and all psychosis-related outcomes and risk moderators in healthy, high-risk, and clinical populations. Overall, we found 32 systematic reviews and meta-analyses. Based on a synthesis of current evidence, cannabis use is associated with subclinical psychosis states (psychotic-like experiences) and traits (schizotypal personality) in the healthy population, as well as earlier onset and development of psychosis. An association with the clinical-high-risk state for psychosis, attenuated psychosis symptoms and transition to psychosis in this population could not be confirmed. An association between cannabis use and psychosis outcomes in patients with psychotic disorder could solely be confirmed regarding relapse. Whether causal effects underlie those associations has not sufficiently been addressed in the evidence synthesis to date.


Introduction
Psychotic experiences such as delusions, hallucinations, and thought disorganization, appear to be continuously distributed and varying in their degree of symptom persistence and severity in the general population (van Os and Reininghaus, 2016).The psychosis continuum may be considered to range from subclinical manifestations in the form of schizotypal personality traits and psychotic-like experiences in the general population through attenuated psychotic symptoms in the prodromal phase of psychosis (also referred to as clinical-high-risk state/-CHR-state), to full-blown psychotic disorders at the other end.The latter rank among the most disabling disorders worldwide and impose a high economic burden on society linked to costs of caring and productivity losses (Vos et al., 2017).This burden further increases in individuals that relapse after their first episode of psychosis (Hong et al., 2009;Fasseeh et al., 2018) It is therefore not surprising that the search for potential predictors and moderators of risk for the development of psychosis and its relapse has been a major focus of research over the last decade.This is particularly important to support the development of individualized risk prediction models which may help personalize interventions based on individual risk (Sanfelici et al., 2020;Koutsouleris et al., 2021) An environmental risk factor, that has often been associated with psychotic disorders, is cannabis use.Cannabis is, after alcohol and tobacco, currently the most popular psychoactive substance worldwide (United Nations Office on Drugs and Crime, 2022).Particularly, with the growing trend of legalization of cannabis as well as increasing interest in its medicinal use worldwide, one of the pressing concerns is its association with psychotic symptoms and disorders.
Studies using randomized controlled designs have shown, that delta-9-tetrahydrocannabinol (THC), the primary psychoactive component of cannabis can induce transient psychotic-like experiences in healthy individuals and worsen existing symptoms in the clinical population (D 'Souza et al., 2005).Cross-sectional studies have shown that cannabis use is more frequent in individuals at clinical high-risk of psychosis (CHR) and those with established psychosis than in healthy controls (Addington et al., 2014;Pardo et al., 2021).Furthermore, several case-control and longitudinal cohort studies have reported that cannabis use is associated with increased odds and risk of onset of psychotic disorder (Andréasson et al., 1987;van Os et al., 2002;Zammit et al., 2002).However, the results of these studies are not always consistent and may even vary across different subpopulations, vulnerability levels, and outcomes of the psychosis continuum.Further factors that might moderate the effects are genetic variables and cannabis use-related variables such as frequency of use (occasional vs. chronic use), age of onset of cannabis use and potency/type of cannabis used (van der Steur et al., 2020;Kiburi et al., 2021).
In parallel with the growing evidence base of primary studies, there is also an increasing number of systematic reviews and meta-analyses that have attempted to summarize the evidence regarding the association of cannabis use with particular psychosis-related outcome of interest such as subclinical expressions of psychosis (psychotic-likeexperiences and schizotypal personality traits) as well clinically relevant psychotic symptoms and disorders in the general population, attenuated psychotic symptoms in CHR individuals and psychotic symptoms and relapse in individuals with established psychosis.Summary reports focusing on the association between cannabis use and those outcomes, either qualitative, in the form of systematic reviews or quantitative syntheses in the form of meta-analyses, have adopted slightly differing methodological approaches and decisions that have led to differing conclusions on the same topic.Existing evidence syntheses may also convey either more or less information on a potential causal relationship between cannabis use and psychosis.
Under such circumstances, umbrella reviews and evidence maps provide a useful bird's-eye view and appraisal of the extant evidence.Assessing existing systematic reviews and meta-analyses against a quality metric allows a more nuanced and robust understanding of the current evidence, and also helps highlight any gaps that remain.Here, we employ this approach to bring together all existing evidence from systematic reviews and meta-analyses to synthesize and present current best evidence on the effects of cannabis use on psychosis-related outcomes and moderating factors across the psychosis continuum.
We focused on systematic reviews and meta-analyses of observational studies, as our aim was to investigate the association between recreational cannabis use and persistent rather than transient psychosisrelated outcomes.Often it is challenging to disentangle transient psychotomimetic effects of cannabis from acute intoxication under the influence of cannabis.To the best of our knowledge, no studies to date have randomized participants to sustained cannabis exposure to investigate non-transient psychosis-related outcomes.Therefore, despite their well-known limitations in terms of causal inferences, we only included systematic reviews and meta-analyses summarizing data from observational studies, which represent the current best available evidence.

Method
We created a systematic evidence map using the umbrella review approach of existing systematic reviews and meta-analyses.The protocol for this study, including the methodology for the review process was preregistered on the open-access database PROSPERO before starting the review process (registration number: CRD42021257756).Not enough statistical information on the data that authors of existing systematic reviews had extracted from the primary studies was forthcoming.Hence, we did not re-analyse the primary study data included in the reviews quantitatively as originally planned, but instead summarized the statistical data narratively and presented the results in the form of forest plots and evidence maps.

Search method
We systematically searched the Web of Science (including the databases science citation index expanded, emerging sources citation index, social sciences citation index, Arts & humanities citation index, conference proceedings citation index, and Medline) to identify systematic reviews or meta-analyses published from the inception of these databases to the present time.The first systematic search was performed on the 1st of July 2021, and this was updated on the 27th of January 2023, searching article titles and abstracts using the following search terms: (AB= (schizo* OR psychotic* OR psychosis OR psychoses )) OR (TI= (schizo* OR psychotic* OR psychosis OR psychoses )) AND (AB= (((systematic) AND (review) ) OR meta-analysis) ) OR (TI= (((systematic) AND (review) ) OR meta-analysis)) AND (AB= (marij* OR cannab* OR marih* OR Hashish* OR Hemp*) ) OR (TI= (marij* OR cannab* OR marih* OR Hashish* OR Hemp*) ).
Further, we manually searched reference lists to identify any additional relevant articles.

Selection criteria
Systematic reviews or meta-analyses examining cannabis use as a risk factor for any psychosis-related outcome (e.g.psychotic-like experiences, schizotypy, attenuated psychosis symptoms, transition to psychosis, psychotic symptoms, development of psychosis, psychosis relapse) in any population (e.g.healthy or general population, clinical high-risk population, patients with established psychotic disorder) were eligible for inclusion.
We assessed articles written in English, German, French, Spanish or Portuguese for inclusion as those were the languages accessible to the authors.
Conference abstracts, umbrella reviews, and systematic reviews and meta-analyses of non-observational studies (e.g.those that investigated the effects of THC or Cannabidiol randomized-controlled trials or neuroimaging studies) and those written in another language than those stated above were excluded from consideration.

Data extraction and quality assessment
Two authors (ED and RP) independently screened titles and abstracts of the retrieved articles to exclude irrelevant studies.The full texts of the remaining articles were then checked against the selection criteria.After both screening phases, any disagreements were discussed and resolved with the first author JMG and the senior author SB.
The data extraction was completed in duplicate by pairs of researchers working independently (ED, RP, AV, YS and JMG).Afterwards, the data was cross-checked, and any discrepancies were resolved in discussion with one of the researchers (JMG).From the selected systematic reviews and meta-analyses, we extracted title, authors, country of the primary author, year of publication, number of primary studies included in the review, study design of included primary studies, any statistical data estimated as part of the meta-analyses as well as data from and the risk of bias assessments from primary studies reported in the systematic reviews/ meta-analyses.All data was extracted and stored in excel spreadsheets that were pre-formatted and customized for this study.
To assess the quality of included systematic reviews and metaanalyses we applied the 16-item AMSTAR-2 tool for systematic reviews of healthcare interventions (Shea et al., 2017).This was also done in duplicate with pairs of researchers (ED, RP, JMG, DB and YS) working independently for every included systematic review with discrepancies resolved through discussion or by consulting the senior researcher (SB).We determined the overall rating of the AMSTAR following the suggestions given in the source publication of AMSTAR 2 (Shea et al., 2017) by categorizing the reviews as high, moderate, low or critically low quality based on the extent of critical flaws.The following items were considered as critical flaws when the answer to the Item was "no": Item 4 ("Did the review authors use a comprehensive literature search strategy"); Item 9 ("Did the review authors use a satisfactory technique for assessing the risk of bias (RoB) in individual studies that were included in the review; Item 11 ("If meta-analysis was justified did the review authors use appropriate methods for statistical combination of results?";Item 13 ("Did the review authors account for RoB in individual studies when interpreting/discussing the results of the review?"; and Item 15 ("If they performed quantitative synthesis did the review authors carry out an adequate investigation of publication bias and discuss its likely impact on the results of the review?).Reviews were classified as "high" quality when they had 0 critical flaws (or with up to 1 non-critical flaw); "moderate" when they had no critical flaw but more than one non-critical flaw; or "low" when they had one critical flaw and "critically low" with more than one critical flaw.

Evidence map and summary
Using the R-package "DiagrammeR" by Iaonne (Iaonne, 2018) we created a visualized overview based on the extracted data from the systematic reviews in the form of a network (JMG).We created an overall evidence map of cannabis use and its association with psychosis related outcomes in different populations (healthy population, CHR population and psychosis population) which summarized the authors of the systematic reviews the outcomes investigated (schizotypy, psychotic-like-experiences, attenuated psychosis symptoms/CHR symptoms, CHR state, psychotic symptoms, clinical diagnosis of psychosis, psychotic symptoms in psychosis, psychosis relapse) the primary studies that were included in these reviews, as well as study design, statistical significance and the risk of bias assessments of the latter.We only used the data that was provided in the published papers and supplementary materials of included systematic reviews and did not additionally extract data from the primary studies to summarize data as reported in the reviews and outline any gaps in reporting.Data was either extracted from tables, manuscript text, forest plots or figures in the main paper of included systematic reviews or their supplementary material.To categorize the data in the different population outcome groups, we oriented based on the categorization and study aims set out by the authors of respective systematic reviews.For example, if the study aim was to summarize the association between cannabis use and subclinical psychotic experiences we had included this review in the respective section and evidence map for "psychotic-like experiences", even if the same primary study papers or same assessment tools had been used to summarize the association between cannabis use and the development of psychotic symptoms or full-blown psychosis in the healthy population in other reviews.For the purposes of the present manuscript, we have used the term "psychotic-like-experiences" only when referring to subclinical psychotic experiences measured in the healthy population but not in the psychosis population.If symptoms had been measured in individuals with established psychosis we have consistently referred to them as "psychotic symptoms".If the review included studies that investigated a range of outcomes and did not summarize the different outcomes in separate sections or analyses on those outcomes, we have summarized those with the term "any psychosis outcome".Similarly, if the aim of the review authors was to investigate whether cannabis use was associated with earlier age of onset of psychosis, we have reported the review under "healthy population", although the studies summarized had included studies that assessed the age of onset of psychosis retrospectively in patients with psychosis.
For the risk of bias/quality assessment, different tools have been used by the authors of published reviews.We re-coded the ratings done by the systematic reviews to facilitate their inclusion and comparability in our visualized overview using the quality categories of "high" (low risk of bias/high quality), "medium" (medium risk of bias/medium quality) and "low" (high risk of bias/low quality).We recoded the ratings that used the Newcastle Ottawa Scale (NOS (Wells et al., 2000)) as "high" with a NOS rating of 7-9, "medium" with a NOS rating of 4-6 and "low" with a NOS rating of 0-3.We adapted the categories from the ratings carried out using the " Risk Of Bias In Non-randomized Studies of Interventions" (Robins-I (Sterne et al., 2016)) tool or those done using the "A Cochrane Risk Of Bias Assessment Tool: for Non-Randomized Studies of Interventions" (ACROBAT-NRSI (Sterne et al., 2014)) tool such that we re-coded the ratings as "high", "medium" or "low" in quality to correspond to the ratings of "low", "moderate" or "serious" risk of bias based on ROBINS-I or ACROBAT-NRSI as reported in the systematic reviews or meta-analyses.We also re-coded the ratings from reviews that used the "strengthening the reporting of observational studies in epidemiology" (STROBE) statement checklist (von Elm et al., 2007) such that ratings of as 0-3 were coded as "low", 4-5 as "medium" and 6-7 as "high" quality.
To visualize the information given in the systematic reviews in the evidence map we used different forms and colours for the different study types and outcomes of the evidence synthesis and primary evidence (see Fig. 1 and 1-extended in supplementary material).To facilitate its display, we further divided the overall network map into subnetworks based on the population and outcome that has been investigated.
Here, for ease of visualization, we report the simpler versions of the evidence maps that depict solely the details of the systematic reviews and meta-analyses rather than the primary studies.The extended maps that include all the details stated above are reported in the supplementary material.

Results
For an overview of the systematic review process see PRISMA flow diagram in Fig. 2. Overall, we identified 32 reviews on cannabis use and its association with psychosis-related outcomes (summarizing data from healthy individuals in 12 systematic reviews; from CHR individuals in seven, from psychosis patients in six systematic reviews and from moderators of risk in nine).For an overview of all the populations and outcomes that have been investigated please see Fig. 3.
The key evidence from the included systematic reviews and metaanalyses is summarized below using a two-tiered approach, which is further visualized in the network maps and forest plots.At the first level, they are organized in terms of the population investigated in the systematic reviews and then further grouped under outcomes of interest considered.
Last, information on risk moderators were grouped in terms of the variable investigated.The results of our AMSTAR-2 evaluation of the included systematic reviews and meta-analyses are reported in Table 1.A final overview of the key questions investigated in the included systematic reviews and meta-analyses as well as the summary conclusions based on the strength (as assessed based on consistency of conclusions from multiple independent reviews) and quality (as assessed by their AMSTAR-2 rating) of evidence are summarized in Table 2.

Cannabis use and its association with psychotic-like-experiences
As visualized in Fig. 4, three systematic reviews (including two metaanalyses) (van Os et al., 2009;Linscott and van Os, 2013;Ragazzi et al., 2018) investigated the association between cannabis use and the risk of psychotic-like-experiences in healthy individuals with effect sizes ranging from OR 1.75 to OR 1.77 for incidence to OR 2.51 to 2.59 for prevalence (see Fig. 5).No pooled effect-size estimate has been computed for the frequency of subclinical psychotic-like experiences in the reviews published to date.
Overall, the current evidence seems to agree that cannabis use is associated with increased prevalence (two meta-analyses rated as critically low quality showing evidence in support), incidence (two metaanalyses rated as critically low quality showing evidence in support) and severity of psychotic-like experiences (one narrative systematic review rated as critically low quality (see Table 2).However, as all three systematic reviews received a "critically low" AMSTAR-2 rating, the evidence synthesis may be biased (see Table 1).For example, none of them included a proper risk of bias/study quality assessment of included studies.Ragazzi et al. (2018) attempted to do so using the STROBE guidelines, however, did not report their results (as shown in Fig. 4).
None of the reviews investigated the potential causal relationship between cannabis use and PLEs.Ragazzi et al. (2018) who mostly summarized the evidence of cross-sectional studies, pointed out that reverse causality may have led to the association as the included studies did not rule out the possibility that people with PLEs may have used cannabis as a form of self-medication for their symptoms.

Cannabis use and its association with psychometric schizotypy
We found only one systematic review and meta-analysis that investigated the link between cannabis use and psychometric schizotypy and this was carried out by Szoke et al. (2014) Szoke et al. (2014) (AMSTAR-2 rating: critically low), see Fig. 6.The authors reported a significant association between lifetime cannabis use and higher schizotypy score across all schizotypy dimensions (total score, positive, negative and disorganization dimensions), and with current cannabis use in all dimensions aside from the negative dimension (see Fig. 6) with the effect sizes ranging from standardized mean difference (SMD) of 0.10 for the negative dimension up to SMD 0.44 for the positive dimension (see Fig. 7).Due to the sparsity of longitudinal studies the authors restricted their inclusion criteria to cross-sectional studies.They discussed that lack of longitudinal studies limited the ability to disentangle whether the association between cannabis use and schizotypy was causal in nature.Interpreting the current available evidence from longitudinal studies that suggest a bidirectional association between cannabis use and schizotypy, the authors further speculated that cannabis use may be one of many potential triggers that gradually increase schizotypal symptoms in vulnerable individuals.

Cannabis use and its association with the development of psychosis
As depicted in Fig. 8, we found two meta-analyses and two narrative systematic reviews reporting on the association between cannabis use and psychotic symptoms (Semple et al., 2005;Henquet et al., 2008;Marconi et al., 2016) with effect sizes ranging from odds ratio (OR) of 1.90 for lifetime cannabis use to OR 3.59 for most frequent cannabis use (see Fig. 5).
As depicted in Fig. 9, the same systematic reviews, as well another meta-analysis by Moore and Colleagues (Moore et al., 2007) further reported data on the association between cannabis use and the development of clinically relevant psychotic disorder with effect sizes ranging from OR 2.37 for lifetime cannabis use up to OR 5.07 for most frequent cannabis use (see Fig. 5).
As shown in Fig. 10, three meta-analyses and two narrative systematic reviews reported a significant positive association between cannabis use (lifetime, frequent and any cannabis use) and any psychotic outcome summarizing data from cross-sectional, longitudinal, and (prospective) cohort studies (Moore et al., 2007;Marconi et al., 2016;Polkosnik et al., 2021;Sorkhou et al., 2021) with effect-sizes ranging from OR 1.41 for lifetime cannabis use up to OR 3.90 for most frequent cannabis use (see Fig. 5).Overall, the evidence seems to agree that cannabis use is associated with increased psychotic symptoms (four critically low-quality systematic reviews showing evidence in support), and the onset of a psychotic disorder (four critically low-quality and one medium-quality systematic review showing evidence in support) and also shows an association with the broader category of any psychosis outcome (two medium-quality, one low quality and two critically low-quality systematic reviews in support).Henquet et al. (2008) and Le Bec et al. (2009) tried to address the question of causality through narratively comparing the study data with established criteria for causality such as temporality (exposure precedes the outcome), reproducibility (studies consistently report an association between exposure and outcome), a dose-response relationship and the existence of a plausible biological mechanism that links exposure and outcome.Both reviews came to the conclusion that those criteria seem to be fulfilled and that cannabis use may be a component cause for psychosis, but that more research was still needed to confirm this assumption.Henquet et al. (2008) also noted that given the current data a bidirectional relationship between cannabis use and psychosis was more likely than reverse causation.Furthermore, the authors noticed that the association does not seem to be due to confounding only, as although effect sizes in the primary studies decreased after controlling for confounders the statistical associations did not go away.Moore et al. (2007) assessed the strength of evidence of a causal relationship by quality-rating longitudinal primary studies based on how potential non-causal explanations (bias and confounding) were accounted for in each study.They could not find any evidence for a non-causal relationship reviewing the current data but outlined the limitations of drawing causal inference from observational studies such as residual confounding due to unobserved variables.Marconi et al. (2016) included only cohort studies that investigated the development of Table 1 AMSTAR-2 ratings of the quality of systematic reviews and meta-analyses.
The AMSTAR items D1-D16 are listed here.D1 -Did the research questions and inclusion criteria for the review include the components of PICO?, D2 -Did the report of the review contain an explicit statement that the review methods were established prior to conduct of the review and did the report justify any significant deviations from the protocol?, D3 -Did the review authors explain their selection of the study designs for inclusion in the review?, D4 -Did the review authors use a psychotic symptoms in individuals who did not show any psychotic symptoms at baseline to avoid reverse causation, however admitted that the results of their meta-analysis would still not unequivocally establish a causal link between cannabis and psychosis.Sorkhou et al. (2021) did not make any inferences about causality.Based on their narrative summary of the current evidence, Polkosnik et al. (2021) concluded that it provided "compelling evidence that cannabis may be a causative agent for symptoms of mental illness in otherwise healthy populations" and "possibly a causative factor for promoting psychosis illness progression to schizophrenia in high-risk individuals."

Cannabis use and its association with earlier age of onset of psychosis
As shown in Fig. 11, we found two meta-analyses and one systematic review that investigated whether cannabis use is associated with earlier age of onset of psychotic disorders (Large et al., 2011;Myles et al., 2016;van der Steur et al., 2020) and reported effect-sizes ranging from SMD 0.40-0.41(see Fig. 7).
Overall current evidence synthesis seems to agree that cannabis use is associated with an earlier onset of psychosis (two low-quality and one critically low-quality systematic reviews in support, see Table 2).
Based on their results, Large et al. ( 2011) concluded that their findings would further support the hypothesis of a causal relationship between cannabis use and psychosis.The authors speculated whether cannabis use would precipitate psychosis in vulnerable individuals by means of gene-environment interactions.They also stated that their findings did not support the assumptions that use of other substances such as alcohol confounded the association.Myles et al. (2016) did not find evidence that tobacco use confounded the association between cannabis use and earlier age of psychosis.Hence, they concluded that cannabis was likely an independent cause of earlier onset of psychosis.

Cannabis use and its association with the CHR state
As depicted in Fig. 12, we found two systematic reviews with metaanalyses reporting on the association of cannabis use with the CHR state (Carney et al., 2017;Fusar-Poli et al., 2017) with effect-sizes ranging from OR 1.56 for current cannabis use up to OR 5.49 for current cannabis use disorder (see Fig. 5).The systematic review by Fusar-Poli et al. (2017) found no evidence that CHR individuals used cannabis more often than healthy controls (see Figs. 5 and 12).However, this effect-size was characterized by a significantly high proportion of total variability due to between-study heterogeneity (see Fig. 5) questioning its precision.Furthermore, a meta-analysis in the same year by Carney et al. (2017) found that current cannabis use was not significantly associated with the CHR state but that lifetime cannabis use and current cannabis use disorder were (see Figs. 12 and 5).
Overall, the current evidence synthesis is conflicting and whether cannabis use may be associated with the CHR state remains unclear (one critically low-quality systematic review showing evidence against, one low quality systematic review in support, see Table 2).
Fusar-Poli et al. ( 2017) stated that due to the unknown pathophysiology of psychosis and therefore non-existence of causal risk factors they only investigated correlational risk factors for being in a CHR state in their review.In contrast, Carney et al. (2017) did not discuss the issue of causal inferences.

Cannabis use and its association with psychotic symptom severity in CHR individuals
Fig. 13 shows that we further found two narrative systematic reviews and one meta-analysis investigating the association between cannabis use and attenuated psychotic symptom severity in individuals at CHR. (van de Meer et al., 2012;Carney et al., 2017;Farris et al., 2020) with effect-sizes ranging from SMD − 0.03 for negative CHR symptoms to SMD 0.16 for positive CHR symptoms (see Fig. 7).The systematic reviews by Farris and colleagues (AMSTAR: "critically low") and Carney et al. (van de Meer et al., 2012) (AMSTAR: "critically low") included one paragraph each synthesizing the study results on the association between cannabis use and CHR symptom severity qualitatively.Although van de Meer et al. (2012) summarized studies that reported inconsistent pattern of results ranging from no association to increased symptoms or decreased symptoms in cannabis users, the authors concluded that cannabis use was "most often associated with more severe high-risk symptomatology within the CHR population".On the other hand, Farris et al. ( 2020) reported that only one of the 5 studies showed a significant association.Farris et al. (2020) stated the results of their study should be interpreted with caution, as the primary data used in the study was not adjusted for confounders.
The meta-analysis from Carney et al. (2017) (AMSTAR: "low") summarized the evidence quantitatively and neither found a significant pooled effect-size for total positive nor for total negative symptoms (see Fig. 7).However, the researchers found a significant association of cannabis use with the subscales of suspiciousness (k (number of primary studies) = 3, hedge's g(bias corrected SMD)= 0.21 CI 0.02-0.39,I 2 (percentage of total variability due to between study heterogeneity)= 0.00) and unusual thought content (k = 4, hedge's g = 0.270.07-0.48I 2 = 30.29).However, the latter results did not take any confounders into account.
Bringing these heterogeneous findings together, it seems that, to date, there is not enough evidence to conclude that cannabis use is associated with increased attenuated psychotic symptom severity in CHR-individuals (one critically low-quality review showing evidence in support while one low and one critically low quality review showing evidence against, see Table 2).There might be an effect for different subtypes of CHR symptoms but this needs further investigation.

Cannabis use and its association with transition to psychosis in CHR individuals
As depicted in Fig. 14, our search revealed six systematic reviews including three meta-analyses and two narrative systematic reviews on the association of cannabis use with the risk of transition from the CHR state to psychosis (van de Meer et al., 2012;Kraan, 2017;Farris et al., 2020;Oliver et al., 2020;van der Steur et al., 2020) with effect-sizes ranging from OR 1.11 for any cannabis use to OR 1.75 for current cannabis abuse (see Fig. 5).Whereas three of the reviews did not confirm a positive association, a meta-analysis by Kraan (2017) reported a significant effect of current cannabis abuse or dependence but not lifetime cannabis use.However, closer inspection of the evidence map (Fig. 14), comprehensive literature search strategy?, D5 -Did the review authors perform study selection in duplicate?D6 -Did the review authors perform data extraction in duplicate?, D7 -Did the review authors provide a list of excluded studies and justify the exclusions?, D8 -Did the review authors describe the included studies in adequate detail?D9 -Did the review authors use a satisfactory technique for assessing the risk of bias (RoB) in individual studies that were included in the review?, D10 -Did the review authors report on the sources of funding for the studies included in the review?, D11 -If meta-analysis was justified did the review authors use appropriate methods for statistical combination of results?, D12 -If meta-analysis was performed did the review authors assess the potential impact of RoB in individual studies on the results of the meta-analysis or other evidence synthesis?, D13 -Did the review authors account for RoB in individual studies when interpreting/ discussing the results of the review?, D14 -Did the review authors provide a satisfactory explanation for, and discussion of, any heterogeneity observed in the results of the review?D15-If they performed quantitative synthesis did the review authors carry out an adequate investigation of publication bias (small study bias) and discuss its likely impact on the results of the review?D16-Did the review authors report any potential sources of conflict of interest, including any funding they received for conducting the review?.Critical Items are depicted in green.Y -Yes, PY -Partial Yes, N -No, NMC-no meta-analysis conducted.indicates that the latter significant overall effect-size (see Fig. 5) was only based on one significant study.This study also suggested that the association was confounded by alcohol use, as outlined in the manuscript of Kraan and colleagues.On the basis of this meta-analysis and other studies, the recent narrative review by van der Steur et al. ( 2020) (AMSTAR-2: "low") concluded that not only frequent cannabis use and dependence, but also lifetime cannabis use increased the risk for transition to psychosis.The authors also suggested that the association was more pronounced in heavy cannabis users and that it might be confounded by alcohol use.Interestingly, the latter review included significant results of studies that have not been included in the other reviews that cited the same studies.The latest meta-analysis by Oliver et al. ( 2020) which included more studies (data from clinical cohorts as well as those nested within randomized controlled trials), did not report a significant pooled effect-size.However, in this synthesis the authors did not investigate the effect of frequency of cannabis use.A later systematic review by Polkosnik et al. (2021) included only a few studies with mixed results that were already included in the other older reviews and came to the opposing conclusion, that cannabis use may increase the risk of transition to psychosis in CHR.
Taking those conflicting findings together, it still remains unclear whether there exists an association between cannabis use and transition to psychosis in CHR (one critically low-quality review and two lowquality review in support, three critically low systematic reviews showing evidence against, see Table 2).While it has been suggested that the effect varies with the extent of use (in terms of frequency and quantity of use), this warrants further investigation.

Cannabis use and its association with psychotic symptom severity in the psychosis population
We found three reviews that investigated whether cannabis use increased psychotic symptoms in patients with psychotic disorders with mixed conclusions (see Fig. 15) (Zammit et al., 2002;Schoeler et al., 2016;Sabe et al., 2020) and effect-sizes ranging from SMD − 0.35 for negative symptoms and SMD − 0.30 for positive symptoms in the comparison of patients who discontinued cannabis use compared to non-users (indicating less frequent symptoms in those patients discontinuing cannabis use compared to those who never used cannabis) up to SMD 0.41 for negative symptoms and SMD 0.26 for positive symptoms in those who continued use compared to those who discontinued use (indicating more psychotic symptoms in those who continued using cannabis during their illness compared to those who discontinued cannabis use).Zammit et al. (2008) (AMSTAR-2:"Critically low") found inconsistent results for the association with positive symptoms and no association with negative symptoms.The meta-analysis by Schoeler et al. (2016) suggested that continued cannabis users experienced higher positive symptoms compared to non-users (see Fig. 7), however, this result seems to be driven by only two low quality significant studies as can be drawn from the evidence map (see Fig. 15).No significant group differences in positive psychotic symptoms were found in the comparison of continued users with discontinued users or non-users with discontinued users (see Fig. 7).Furthermore, no significant group differences were found in the negative symptom dimension (see Fig. 7).
Similarly, Sabe et al. (2020) (AMSTAR: "low") found no significant This table summarizes the main research questions answered by our review using the outcomes yes vs no vs unclear based on the opinion that we gained while evaluating the whole systematic review evidence.PLE-psychotic-likeexperiences, cCUcontinued cannabis use, dCUdiscontinued cannabis use.
* gives the number of critically low(A), low (B), moderate (C) and high (D) quality reviews to a specific question using the format (A|B|C|D).

Fig. 4.
Evidence map of cannabis use and its association with psychotic-like experiences.This figure depicts all the existing evidence syntheses published about cannabis use and its association with psychotic-like experiences, which are two systematic reviews with meta-analyses (diamond shape) and one narrative systematic review (rectangle).Green colour indicates that a (significant) positive association between cannabis use and the outcome has been reported.Grey colour indicates a non-significant association.White fill indicates that no data from the primary studies has been reported in the systematic review.The dark red colour of the arrows indicates the result of our Overall-AMSTAR-rating (Critically low).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) differences between negative symptoms in cannabis-using psychosis patients and no drug users (see Fig. 7).However, recent cannabis abstainers seemed to have less severe negative symptoms compared to no drug users (Figs. 7 and 15).They found a significant difference between current cannabis users and non-drug-users in terms of positive symptoms (Fig. 7).However, this association was no longer significant when only high-quality studies were included (SMD = 0.07; 95 % CI − 0.03 -0.17; p = 0.18, I 2 = 5.4 %).No differences were found comparing recent cannabis abstainers with no drug users in their positive symptom severity (see Fig. 7).
To sum up, the results of the current research evidence seems to be conflicting.It remains unclear whether cannabis use is associated with higher positive symptoms (one critically low quality review showing evidence in support, another review with critically low quality suggesting equivocal evidence and a third review with medium quality showing evidence against) and whether discontinued cannabis use may be associated with less severe negative symptoms compared to non-use (one critically low quality review showing evidence against, one critically low review suggesting equivocal evidence and one medium quality review showing evidence in support, see Table 2).

Transition from cannabis-induced psychotic disorders to schizophrenia
The systematic review by Murrie et al. (2020) meta-analytically synthesized the transition rates from substance-induced psychotic disorders to schizophrenia and found a statistically significant pooled transition rate from cannabis-induced psychotic disorders to schizophrenia (k = 6, 34 %, N = 3040, Q = 137, p < 0.0001, I 2 96 %) as  Fig. 8. Evidence map of cannabis use and its association with psychotic symptoms in the healthy population this figure depicts the association between cannabis use and psychotic symptoms, which were investigated in three meta-analyses (diamond shapes) and one narrative systematic review (square).All reviews were rated as critically low based on AMSTAR-2 (dark red arrows pointing from the outcomes to the systematic reviews).Green colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)visualized in Fig. 16.

Cannabis use and its association with relapse in the psychosis population
As shown in Fig. 17, we found one narrative systematic review and one meta-analysis that reported on the effect of cannabis use on the risk of relapse in psychosis (Zammit et al., 2002;Schoeler et al., 2016) with effect sizes ranging from SMD 0.02 for discontinued use vs. non-use to SMD 0.36 for continued use vs. non-use (see Fig. 7).The first systematic review that showed evidence for the association of cannabis with increased relapse reported on evidence for a dose-response effect and increased effect of continued cannabis use.The authors however also noted that the effects decreased when confounders were controlled (Zammit et al., 2008) .Schoeler et al. (2016) (AMSTAR: "critically low") investigated the effects of continued cannabis use more closely.Those authors meta-analytically compared continued, discontinued and non-cannabis-using patients in their relapse outcome.Their results suggested that patients who continued using cannabis were more likely to suffer from relapse than non-using patients and discontinuing users (see Fig. 7).Discontinuation compared to non-use did not appear to make a statistically significant difference.
Overall, the evidence synthesis regarding the association between cannabis use and relapse appears to suggest an association (two critically low-quality reviews showing evidence in support, see Table 2).
Whether this association reflects a causal relationship was not specifically investigated in either of the reviews.Zammit et al. (2008) however stated that due to the limited adjustment of confounders in primary studies, confidence about a causal relationship must be low.Still, they did not suspect reverse causation to have produced the effects, as only longitudinal studies were included.Similarly, Schoeler and colleagues admitted that they were not able to control for important confounders such as medication adherence, engagement with services or other drug use.

Cannabis use and its association with any psychotic outcome in the psychosis population
The medium-quality narrative review by Athanassiou et al. ( 2021) tried to enhance inferences drawn from their review by only including prospective longitudinal studies to investigate the association between cannabis use and any psychotic outcome.They further looked at the confounders that have been adjusted for in the primary studies included.Although the authors interpreted their results as indicative of a positive association overall (see Fig. 18), they also mentioned that those studies that controlled for important confounders such as other substance use, mostly did not show any significant association between cannabis use and the psychosis outcome that was investigated in the psychosis population.In addition, further important confounders such as medication adherence and other clinical and social intervention services were not accounted for in the included studies.A pooled effect size estimate has not been computed in the reviews published to date.(Fig. 19)

Environmental moderators
In the included systematic reviews and meta-analyses (Hosseini and  , 2020;Kiburi et al., 2021;Polkosnik et al., 2021;Petrilli et al., 2022;Robinson et al., 2022), we found seven cannabis use-related and other environmental variables that have been investigated as moderators of the relationship between cannabis use and psychosis related outcomes (see Fig. 19).The following moderators have been identified as meaningful or significant: age of first cannabis use, adolescent cannabis use, pattern or frequency of cannabis use, type or potency of cannabis and childhood trauma (see Fig. 20).Other substance use has been shown to either increase or confound the association between cannabis use and psychosis related outcomes.
Earlier age of onset of cannabis use and pattern/frequency of cannabis use were the factors that were investigated most frequenctly and found to be associated with increased risk of subclinical, attenuated and clinical relevant psychotic outcomes measured in the healthy, CHR and psychosis populations respectively.However the review authors could not fully rule out confounding through other environmental and genetic risk moderating factors or disentangle individual contributions of the cannabis use related factors such as age of onset of cannabis use, frequency of use and potency of use.The authors expected a similar complicated interplay with childhood-trauma that may additionally Fig. 12.Evidence map of the association between cannabis use and the CHR state.This figure depicts the association between cannabis use and the CHR state (Clinical-high-risk state for psychosis).We found 2 meta-analyses (diamonds) on this topic.One review was rated as critically low based on AMSTAR-2 (dark red arrow), the other as low quality (red arrow).The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The grey fill colour indicates a non-significant association.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)Fig. 13.Evidence map of cannabis use and its association with CHR symptom severity in the CHR population.This figure depicts the association between cannabis use and CHR (Clinical-high-risk state for psychosis) symptom severity.We found one meta-analysis (diamond) on this topic and two narrative systematic reviews (squares).Two reviews were rated as critically low based on AMSTAR-2 as indicated by a dark red arrow pointing from the outcome to the review) and one as low quality as indicated by a red arrow.Green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.Grey fill colour indicates a non-significant association.No fill-colour indicates that no information of the statistical significance of primary studies was given.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)increase the risk of mentioned factors.

Genetic moderators
As visualized in Fig. 21, included systematic reviews and metaanalyses (Vaessen et al., 2018;van der Steur et al., 2020;Kiburi et al., 2021;Carvalho and Vieira-Coelho, 2022) investigated single nucleotide polymorphisms (SNPs) in the following genes as potential moderators of the association between cannabis use and psychosis-related outcomes: Nitric oxide synthase 1 adaptor protein (NOS1AP) gene (no association); Dopamine receptor D2 (DRD2) gene (potential moderator); Brain-derived neurotrophic factor (BDNF) gene (association unclear); Fatty acid amide hydrolase (FAAH) gene (association unclear); AKT1 (serine/threonine kinase 1) gene (potential moderator) and Catechol-O-methyl transferase (COMT) gene (association unclear).Figs, 20 and 22 show the pooled effect sizes that were extracted from the meta-analyses.In addition, genome-wide association studies included in the latest review by Carvalho and Vieira-Coelho (2022) suggested that SNPs on the CHRM3 (cholinergic receptor muscarinic 3) gene and on the P2 × 7 (P2X purinoceptor 7) gene may play a role in the complex interaction between cannabis use and psychosis.Furthermore, qualitative evidence synthesis investigating potential moderating effects of familial predisposition and prodromal symptoms at baseline provided unclear results.Finally, mendelian randomization studies suggested a bidirectional association between cannabis use and psychosis (no pooled effect-sizes for the moderating effect of familial predisposition and mendelian randomization studies have been published in the systematic reviews to date (Fig. 22).

Discussion
To our knowledge, this is the first systematic evidence map that reports on the association between cannabis use and psychosis-related outcomes across the whole psychosis continuum and moderators of that risk.The main research questions that were investigated in the reviews and meta-analyses included in our umbrella review and our conclusions about the current state of evidence are summarized in Table 2. Based on this, it seems that research summarized to date has provided a clear answer to most of the questions listed, though some of the problems remain unresolved and need further investigation.Although there are issues around the methodological quality of systematic reviews and meta-analyses included here with several reviews ranging from critically low to moderate quality, they suggest a consistent association between cannabis use and most of the investigated psychosis-related outcomes across levels of the psychosis continuum from healthy individuals to patients with psychotic disorders.
Taking into account the amount and the consistency of available evidence from systematic reviews and meta-analyses as well as their quality, it seems on balance that there is clear evidence in support of an association between cannabis use and psychotic-like experiences and schizotypy.However, existing systematic reviews mostly focused on cross-sectional studies so that only an association rather than a probable causal relationship may be inferred.Future systematic reviews should therefore focus on longitudinal studies following up cohorts from before the onset of subclinical psychosis symptoms.Similarly, mostly crosssectional studies have examined the relationship between cannabis use and the clinical-high-risk state for psychosis.Therefore, one may only Another interesting outcome of our umbrella review is that the association between cannabis use and psychosis outcomes appear to be less pronounced in the CHR population.Considering all available evidence from existing systematic reviews and meta-analyses as well as their quality, whether cannabis use is associated with increased severity of CHR symptoms or with increased transition to a psychotic disorder, remains unclear.This is intriguing when considered in light of more consistent evidence of association between cannabis use and psychosisrelated outcomes in healthy individuals and those with established psychosis.Whether this reflects the possibility that the effects of cannabis use on transition to psychosis in CHR individuals is only evident in heavy cannabis users, as has been suggested (Kraan, 2017;van der Steur et al., 2020) or whether CHR cohorts in these studies, which typically recruit help-seeking individuals, are not fully representative of those at risk of transition (van Os and Guloksuz, 2017) remains to be systematically investigated in future studies.
Furthermore, the systematic reviews and meta-analytical evidence seems to agree that cannabis use in the general population is associated with the development of clinical psychosis.The effect sizes obtained in the meta-analyses that have been published around the same time differed from each other ranging from 1.5 to 3-fold increase in odds for the cannabis user group.This is most likely due to the fact, that the authors of different reviews sometimes extracted different effect-sizes (either unadjusted or adjusted for covariates or from different subgroups based on factors that they considered appropriate such as cannabis use level) from the same cohort or publication.Reviews also differed in terms their quantitative synthesis approaches.Some combined data from cross-sectional and longitudinal cohorts or observational and experimental studies or combined both the "softer" outcome of psychotic symptoms and the "harder" outcome of clinical psychosis diagnosis in the same analysis.However, when they separated those in subgroup or sensitivity analyses, it has been consistently shown that the effect-sizes were larger for association with the narrower clinical outcomes.
A major limitation of the present endeavour relates to our focus on only observational studies.As mentioned at the outset, observational study designs do not directly address the question whether any observed association between cannabis use and psychosis-related outcomes is causal in nature.
Although the language used in some of the systematic reviews implied causal effects (such as "effect of cannabis on"), only a few of the systematic reviews and meta-analyses intended to systematically address the question of whether cannabis actually causes those psychosis related outcomes by means of their evidence synthesis.Of those authors that have explored the causal question narratively by using Bradford Hill's Criteria or through discussing the plausibility of alternative non-causal theories such as the self-medication hypothesis or reverse causality, most considered the theory of cannabis as a component cause for psychosis and the existence of a bidirectional association between cannabis use and psychosis as most plausible.
From the review authors that tried to address the causality question by only including longitudinal studies in their review, none of them distinguished between primary studies using conventional analyses (such as (repeated) cross-sectional and longitudinal designs with single equation approaches) and more sophisticated advanced analytical Fig. 15.Evidence map of cannabis use and its association with psychotic symptom severity in the psychosis population.This figure depicts the association between cannabis use and psychotic symptom severity in individuals with established psychotic disorders.We found two meta-analyses (diamonds) on this topic and one narrative systematic review (square).Two reviews were rated as critically low based on AMSTAR-2 (dark red arrows), one as medium quality (orange arrow).The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The grey fill colour indicates a non-significant association.Blue colour indicates that the paper reported mixed results so that the association remains unclear and purple fill colour indicates an inverted effect (recent cannabis abstainers showed less symptoms than non-users).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)techniques that support more robust causal interpretations such as longitudinal studies that allowed for time-varying exposure and confounder-control.Future systematic reviews and meta-analyses should therefore synthesize information from those observational studies that target key issues for causal inference such as confounding and reverse causality using appropriate study designs and more advanced statistical analytical methods (e.g.target emulation trials, (genetic) instrumental variable methods, propensity scores, fixed-effects regression with time-varying covariates and G-methods such as G-formula, marginal structural models and G-estimation, etc.).
Furthermore, meta-analyses based on individual participant data raise the potential for new opportunities of using more advanced quantitative synthesis approaches and further facilitate investigation of environmental and genetic moderators of the relationship between cannabis use and psychosis related outcomes.While experimental cannabis/ THC administration studies offer the possibility to draw causal inferences by means of the study design, those carried out to date have generally examined transient psychosis-related outcomes.Arguably, they offer only limited evidence with regard to questions regarding the association between cannabis use and nontransient psychosis-related outcomes such as onset of a psychotic episode or its relapse.Although, we did not synthesize available evidence regarding the acute and transient effects of cannabis as examined under experimental conditions, it is worth noting that the results summarized herein broadly align with a recent systematic review of experimental studies investigating the acute and transient psychotomimetic effects of THC.For example, the most recent review by Hindley et al. (2020) synthesized evidence from experimental studies involving oral or inhaled administration of delta-9-tetrahydrocannabinol (THC), the main psychoactive component of cannabis in healthy human participants.They reported that THC administration was associated with transient psychotic symptoms, with stronger effects for positive compared to negative psychotic symptoms." Considering the plurality of experimental evidence and observational evidence in healthy individuals, one may argue the case for shifting of focus from further investigations of cannabis -psychosis association to evaluation of appropriate intervention approaches to mitigate harm in those most vulnerable.This is particularly important in light of the evolving legislative landscape and societal attitudes about cannabis use across many geographical jurisdictions.
Furthermore, existing experimental data that has shown that THC further increases psychotic outcomes in clinical populations (D 'Souza et al., 2005) aligns with those results of our umbrella review that indicate that adverse effects of cannabis may not be limited to the general population but may further worsen psychosis-related outcomes in not reported, I 2 -percentage of total variability due to between study heterogeneity, ACU-adolescent cannabis use.clinical populations, especially in patients with established psychosis.This may have important implications for health care professionals working with people at various stages of the psychosis risk and disease continuum with history of co-occurring cannabis use disorder as well as in more general addiction or public health settings.There may be important policy implications as well of the evidence highlighted here that suggests a greater risk of poor outcome in those with heavier and persistent use such as in those having cannabis dependence.In this regard, some of the risk modifiers such as frequency and potentially potency of cannabis used as well as an earlier age of onset of cannabis use may serve as useful pointers in focusing intervention strategies on those that may be most affected.
Our umbrella review further suggests that genetic factors potentially may play role in increasing the risk of development of psychosis in cannabis users.However, the existing evidence synthesis, mostly included studies only investigating one single or maximal two nucleoid polymorphisms at a time.Future meta-analyses should therefore summarize the evidence from large-scale genomic association studies, those employing polygenic risk scores and mendelian randomization approaches.Moreover, with a hopefully increasing number of studies investigating the long-term epigenetic effects of cannabis use based on DNA-methylation and histone protein regulation data and its association with psychosis outcomes future systematic reviews and meta-analysis may aim at summarizing and quantifying this data in order to better understand the effect of interaction between genetic risk and cannabis use on psychosis-related outcomes.
Another aspect that the systematic review evidence to date fails to address, is the comorbidity of cannabis use with other psychiatric disorders in otherwise healthy individuals.Current evidence implies that cannabis use is associated with a range of other comorbid psychiatric symptoms such as anxiety and depression and that cannabis users tend to self-medicate those symptoms with their use (Gobbi et al., 2019;Hasin and Walsh, 2020).However none of the reviews to date have investigated whether the presence of other comorbid psychiatric disorders may increase the vulnerability of harm from cannabis use or even act as confounder when assessing a causal relationship between cannabis use and psychosis in the general population.
The strength of this umbrella review lies in that it combines, visualizes and evaluates all available evidence across all the outcomes and subpopulations of the psychosis continuum, that have been studied to date.However, as the investigation of the psychosis continuum is still an emerging field, the terminology to describe the different dimensions of psychosis related outcomes and their measurement approaches are not yet standardized.Across the included systematic reviews and their primary studies, different terms were used for the same/similar concepts.On the other hand some terms such as "psychotic-like-experiences, "psychotic symptoms" and "psychotic experiences" were used interchangeably.Definitions were most of the time not provided and if so, Fig. 21.Evidence map of genetic moderators the association between cannabis use and psychosis related outcomes.This figure depicts all genetic moderators that have been investigated to influence the association between cannabis use and psychosis related outcomes and other studies that have been reported on genetic factors.We found eight different SNPs (single-nucleotide-polymorphisms) that have been investigated.The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The grey fill colour indicates a non-significant association.No fill colour indicates that no information of the statistical significance of primary studies given.Blue colour indicates that the paper reported mixed results so that the association remains unclear.The size of the circles corresponds to the number of primary studies in the systematic review.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)measurement tools for the outcomes investigated still varied strongly across the systematic reviews and included primary studies.
In addition to the limitation with regard to the content of the existing evidence syntheses, we found that the vast majority of the currently existing systematic reviews on cannabis use and psychosis-related outcomes have a critically low methodological quality, limiting confidence in their conclusions.Only two out of 32 reviews were rated as high quality, six reviews were classified as moderate, six as low, and the remaining as critically low in quality.However, this is not as surprising, and is consistent with the quality of systematic reviews as assessed with the AMSTAR-2 in other areas of research (Leclercq et al., 2020).It has been argued that low quality ratings reflect the current low methodological quality and confidence in systematic reviews in general (Pieper et al., 2021).What stands out from the results of our AMSTAR-2 evaluation is that many of the included reviews neither assessed the risk of bias of included studies nor discussed the influence of study quality on the results.Others used reporting guidelines such as STROBE (Schoeler et al., 2016;Ragazzi et al., 2018) to determine the quality of included studies.Future studies may therefore use validated quality assessment or risk of bias assessment instruments and in addition follow the STROBE guidelines to evaluate the reporting quality of included reviews as it was originally intended to be used for (Da Costa et al., 2011).Another limitation of our approach relates to the fact that we did not ourselves estimate pooled effect-sizes using all available published data related to the questions investigated in the literature.As our overall objective was to provide a high-level summary of existing expert knowledge for the interested reader and thereby also point out the inconsistencies and gaps in evidence and their interpretation, we summarized instead the pooled effect-sizes as well as qualitative review conclusions that have been reported in published reviews and meta-analyses and provided an assessment of the methodological quality of these approaches.

Conclusion
In summary, our umbrella evidence map aimed to provide an overview of all the evidence that has been synthesized by different researchers to date focusing on the broad topic of cannabis use and any psychosis-related outcomes.We provide a visually enriched summary of evidence that outlines the key evidence and their methodological quality, discrepancies between existing evidence and their interpretation where they exist and summary conclusions as well as gaps in evidence that may be addressed in future research, thereby facilitating a nuanced understanding of the current state of evidence in the field.

Fig. 1 .
Fig. 1.Visualization scheme of the systematic evidence map.This figure visualizes the colours and shapes that have been used to visualize the information in the evidence maps.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Overview of the populations and outcomes investigated.This figure shows all the populations and outcomes that have been included in this systematic evidence network map ranging from the healthy population (yellow) over the CHR population (Clinical high risk for psychosis, orange) to individuals with full-blown psychotic disorder (red).The severity of the psychotic outcomes is coloured on a spectrum from yellow (low) to red (high) ranging from subclinical to clinical psychotic outcomes.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig. 5. Forest plot showing the effect of cannabis use on psychosis-related outcomes (as summarized by odds ratios) in the healthy, high-risk and clinical population.knumber of reviews, Ntotal number of participants, n.r.-not reported, I 2 -percentage of total variability due to between study heterogeneity, Estimate -Odds ratio, CHR-clinical high-risk.

Fig. 6 .
Fig.6.Evidence map of cannabis use and its association with psychometric schizotypy.This figure depicts the existing evidence syntheses published about cannabis use and its association with schizotypy, which is one systematic review with meta-analysis (diamond shapes) including its cross-sectional primary studies (circles).The green colour indicates that a (significant) positive association between cannabis use and the outcome has been reported.The grey colour indicates a nonsignificant association.The dark red colour of the arrows pointing from the outcome to the review indicates the result of our Overall-AMSTAR-rating (Critically low).(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 9 .Fig. 11 .
Fig.9.Evidence map of cannabis use and the risk of development of a psychotic disorder.This figure depicts the association between cannabis use and its association with the development of a psychotic disorder.We found 4 meta-analyses (diamonds) and one narrative systematic review (square) reporting on this topic.Four reviews were rated as critically low based on AMSTAR-2 (dark red arrows pointing to the corresponding systematic reviews) and one as moderate quality (orange arrow pointing to the review).The green colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 14 .
Fig. 14.Evidence map of cannabis use and the risk of transition to psychosis in the CHR population.This figure depicts the association between cannabis use and the risk of transition to psychosis in the CHR population.We found three meta-analyses (diamonds) on this topic and three narrative systematic reviews (squares).Three reviews were rated as critically low based on AMSTAR-2 (dark red arrows pointing from the outcome node to the systematic review node), one as low quality (red arrow).The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.Grey fill colour indicates a non-significant association.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 16 .Fig. 17 .Fig. 18 .
Fig.16.Evidence map of transition from cannabis use induced disorder to schizophrenia.This figure depicts transition rates from cannabis-induced psychotic disorders to schizophrenia.We found one meta-analysis (diamond) on this topic.The review was rated as medium quality (indicated by the orange arrow pointing from the outcome to the systematic review) based on AMSTAR-2.The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 19 .
Fig. 19.Evidence map of environmental moderators of the association between cannabis use and all psychosis related outcomes.This figure depicts all cannabis-userelated and other environmental moderators that have been investigated to influence the association between cannabis use and psychosis related outcomes.We found 6 different environmental moderators.The green fill colour indicates that a (statistically significant) positive association between cannabis use and the outcome has been reported.The grey fill colour indicates a non-significant association.The size of the circles corresponds to the number of primary studies in the systematic review.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 22 .
Fig. 22. Forest-plot showing moderators on the effect of cannabis use on psychosis-related outcomes (as summarized by beta-coefficients).knumber of reviews, N total number of participants, n.r.-not reported, I 2 -percentage of total variability due to between study heterogeneity, Estimate -Betacoefficient.(For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Table 2
Main research questions and answers.

Table 2
(continued ) (continued on next page) J.M.Groening et al.

Table 2
(continued ) (continued on next page) J.M.Groening et al.

Table 2
(continued ) (continued on next page) J.M.Groening et al.