Introduction
Cannabis is one of the most used psychoactive drugs worldwide (Wang et al., Reference Wang, Qin, Xing, Zhu and Jia2024). Estimates suggest that each year approximately 219 million people use cannabis globally [United Nations Office on Drugs and Crime (UNODC), 2023], and this number is increasing (UNODC, 2013). The rise in cannabis use, coupled with its legalization and decriminalization in several high-income countries, has raised concerns about its potential impact upon mental health [World Health Organization (WHO), 2016]. Indeed, cannabis use has been associated with anxiety (Kedzior & Laeber, Reference Kedzior and Laeber2014), depression (Lev-Ran et al., Reference Lev-Ran, Roerecke, Le Foll, George, McKenzie and Rehm2014), and suicidality (Shamabadi et al., Reference Shamabadi, Ahmadzade, Pirahesh, Hasanzadeh and Asadigandomani2023) in clinical settings, and up to 22% of users meet criteria for cannabis use disorder (CUD) (Leung et al., Reference Leung, Chan, Hides and Hall2020). Cannabis use has also been associated with severity of psychotic symptoms (Hindley et al., Reference Hindley, Beck, Borgan, Ginestet, McCutcheon, Kleinloog and Howes2020) and a fourfold increased risk of psychosis (Marconi et al., Reference Marconi, Di Forti, Lewis, Murray and Vassos2016), a relationship that appears dose-dependent (Petrilli et al., Reference Petrilli, Ofori, Hines, Taylor, Adams and Freeman2022; Polkosnik et al., Reference Polkosnik, Sorkhou and George2021).
Cannabis use is highly prevalent in people with a diagnosis of schizophrenia. Meta-analyses suggest the prevalence of CUD among this group is around 25% (Koskinen et al., Reference Koskinen, Löhönen, Koponen, Isohanni and Miettunen2010), and it is even higher (36%) in those with first episode psychosis (Hunt et al., Reference Hunt, Large, Cleary, Lai and Saunders2018). Despite qualitative accounts of the beneficial effects of cannabis use on hopelessness (Asher & Gask, Reference Asher and Gask2010), boredom, depression, anxiety, and psychotic symptoms (Parshotam & Joubert, Reference Parshotam and Joubert2015), and its restorative effect on identity and social networks (Wagstaff et al., Reference Wagstaff, Graham, Farrell, Larkin and Tatham2018), cannabis use in schizophrenia has also been associated with deleterious clinical outcomes including relapse (Bioque et al., Reference Bioque, Mezquida, Amoretti, García-Rizo, López-Ilundain, Diaz-Caneja and Bernardo2022), medication treatment resistance (Reid & Bhattacharyya, Reference Reid and Bhattacharyya2019), hospitalization (Patel et al., Reference Patel, Wilson, Jackson, Ball, Shetty, Broadbent and Bhattacharyya2016), and suicide (Serafini et al., Reference Serafini, Pompili, Innamorati, Rihmer, Sher and Girardi2012).
Suicide is the leading cause of unnatural death among people with a diagnosis of schizophrenia (Moreno-Küstner et al., Reference Moreno-Küstner, Guzman-Parra, Pardo, Sanchidrián, Díaz-Ruiz and Mayoral-Cleries2021). The risk of suicide is 10 times greater in this group compared to the general population (Correll et al., Reference Correll, Solmi, Croatto, Schneider, Rohani-Montez, Fairley and Tiihonen2022), and the prevention of suicide is considered an international public health priority (Lu et al., Reference Lu, Dong, Zhang, Zhu, Ungvari, Ng and Xiang2020). Cannabis use is increasingly prevalent in those who die by suicide (Mulligan et al., Reference Mulligan, Bojanić, Hunt, Baird, Turnbull, Kapur and Shaw2024a) and has been associated with attempted suicide (Waterreus et al., Reference Waterreus, Di Prinzio, Badcock, Martin-Iverson, Jablensky and Morgan2018) and suicidal ideation (Ricci et al., Reference Ricci, Ceci, Di Carlo, Di Muzio, Ciavoni, Santangelo and Maina2023a) in people with a diagnosis of schizophrenia. This has led many to hypothesize a role of cannabinoid-1 receptor disruption (Ceccarini et al., Reference Ceccarini, Kuepper, Kemels, van Os, Henquet and Van Laere2015) and dysregulation of the endocannabinoid system more generally (Volkow, Hampson & Baler, Reference Volkow, Hampson and Baler2017) in the emergence of suicidal behavior in this group. However, not all studies lend support for a link between cannabis and suicide (Heuschen et al., Reference Heuschen, Bolhuis, Zantvoord, Bockting, Denys, Lok and Schirmbeck2024; Naji et al., Reference Naji, Rosic, Dennis, Bhatt, Sanger, Hudson and Samaan2018; Reutfors et al., Reference Reutfors, Brandt, Jönsson, Ekbom, Sparén and Ösby2009). A small narrative review also concluded that recent data on this relationship is contradictory and insufficient to draw firm conclusions (Ricci et al., Reference Ricci, Cristofori, Passarello, Paggi, Cavallo, Ceci and Maina2023b).
To the authors knowledge, no studies have attempted to meta-analyze all existing data on the relationship between cannabis use and suicide in people with a diagnosis of schizophrenia. This paper fills this important gap by synthesizing research on the relationship between cannabis use, measured categorically or continuously, of any frequency, degree, or impact, at any time point, and individual suicide-related outcomes (i.e., suicide death, attempted suicide, and suicidal ideation). By pooling and analyzing data from longitudinal, case control, and cross-sectional studies, we aimed to provide a quantitative assessment of the nature of these relationships while considering potential sources of heterogeneity and bias.
Method
Search strategy
A systematic review and meta-analysis, pre-registered with PROSPERO (registration number: CRD42024437152), were conducted following guidelines from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al., Reference Moher, Liberati, Tetzlaff and Altman2009). The process of identifying and retrieving eligible articles involved two strategies: (1) an electronic literature search across three major bibliographic databases; including Medline (spanning 1946 to 2024), PsycINFO (spanning 1806 to 2024), and Embase (spanning 1974 to 2024); and (2) examining the reference lists and citations of relevant papers to identify studies not found in the initial searches (i.e., forward and backward citation tracking).
The search incorporated terms related to substance use and suicide in its various manifestations, including (substance use* OR substance abuse* OR substance misuse* OR substance dependen* OR substanc* OR alcohol* OR alcohol use* OR alcohol abuse* OR alcohol misuse* OR alcohol dependen* OR addict* OR drug* OR drug use* OR drug abuse* OR drug misuse* OR drug dependence OR cannabis OR marijuana OR cocaine OR heroin OR amphetamine* OR methamphetamine* OR smoking OR tobacco OR nicotine).af AND (suicide OR suicid* OR suicidal behavio* OR selfharm).af AND (psychosis OR psychoti* OR schizo* OR hallucinat* OR delusion* OR paranoi*).af. To broaden the search results and ensure all pertinent studies were identified, Medical Subject Headings (MeSH) were also utilized. Duplicate entries were systematically eliminated both electronically and manually. Broad search terms were used as a preliminary scoping review suggested that cannabis-related effects might be documented in studies addressing substance abuse in general, rather than exclusively focusing on cannabis.
Inclusion and validity
The review included longitudinal, case-control, and cross-sectional studies that met the following criteria: (1) they involved adults diagnosed with a schizophrenia-spectrum disorder (such as psychotic disorder, schizophrenia, schizoaffective disorder, delusional disorder, or schizophreniform disorder) based on DSM-III, DSM-III-R, DSM-IV, DSM-IV-TR, DSM-V, or comparable research diagnostic criteria [such as the International Classification of Diseases (ICD)], or diagnoses confirmed through psychiatrist or psychologist case note reviews or clinical interviews; (2) they contained subjective or objective measures of cannabis use (including self-reported, informant-reported, or clinician-reported composite, dichotomous, or continuous measures of current or past cannabis use using questionnaires, checklists, or diagnoses from case note reviews, clinical interviews, or urine analyses); (3) they included a specific measure of suicidality (including self-reported, informant-reported, or clinician-reported dichotomous or continuous measures of suicide death, current or past suicide attempts or suicidal ideation); and (4) they provided summary estimates of cannabis use and suicide-related outcomes that were sufficient for meta-analysis.
We excluded studies that were not written in English or French. Additionally, we excluded (1) case reports, conference proceedings, and review articles; (2) studies focusing on non-suicidal self-injury (NSSI), given phenomenological differences between NSSI and suicidality in terms of intention, frequency, and lethality (Cipriano et al., Reference Cipriano, Cella and Cotrufo2017); (3) studies that used composite measures of suicidality or combined various suicide-related variables (e.g., grouping suicide with attempted suicide, or attempted suicide with suicidal ideation); and (4) studies examining adults with substance-induced psychosis, those ‘at risk’ of psychosis, focusing on retrospective relationships before the onset of psychosis, or those reporting on a mixed patient sample where less than 50% had a schizophrenia-spectrum diagnosis.
Two researchers independently assessed eligibility using a two-stage process: an initial screening of titles and abstracts, followed by a full-text review. During the initial phase, L.M. and K.H. independently screened all titles and abstracts. If either researcher found a title or abstract suitable for inclusion, it was progressed to full article screening (independent agreement: 92.5%). In the final phase, whole articles were reviewed to inform final decisions on inclusion (independent agreement: 98.5%). All discrepancies between the two researchers were resolved through consultation with a third reviewer (GH) to reach a consensus.
Risk of bias was evaluated using a modified version of the Effective Public Health Practice Project (EPHPP) tool (Thomas et al., Reference Thomas, Ciliska, Dobbins and Micucci2004). This tool assessed the methodological quality of the studies across several domains, including (1) selection bias; (2) confounders; (3) data collection methods; (4) withdrawals/dropouts; and (5) statistical analyses, to provide an overall quality score. L.M. and K.H. independently assessed study quality and discrepancies were discussed and resolved with input from a third researcher (G.H.). There was substantial agreement on the global scores assigned by the coders (92.5%).
Data extraction
All data were independently extracted by L.M. and K.H and were cross-checked to ensure accuracy. Authors were contacted if relevant data were missing from a study. Descriptive variables extracted included: year of study, country, study methodology, sample size, age of the sample, number and age of cases/controls with a suicide outcome (if applicable), gender, sample diagnoses, diagnosis measure, cannabis measure, cannabis use timeframe, suicidality measure, baseline total Positive and Negative Syndrome Scale (PANSS; Kay et al., Reference Kay, Fiszbein and Opler1987) score (if available), and statistical information needed to calculate effect sizes. The baseline total PANSS score was prioritized over other measures like the Brief Psychiatric Rating Scale (BPRS; Overall & Graham, Reference Overall and Gorham1988) as it includes both specific symptoms of schizophrenia and general psychopathology in one composite score.
To avoid data duplication, a hierarchy was constructed to guide decisions about extraction and improve the homogeneity of included studies. When studies reported multiple cannabis variables (e.g., binary and continuous), binary variables were preferentially chosen, as most studies measured cannabis by the presence or absence of use (or CUD). If cannabis use or specific suicide outcomes were recorded at various time points (e.g., recent and lifetime) or using self-reported and clinician reported measures, summary effects were pooled to create a single combined outcome metric. In cases where identical samples were used across different studies, eligible papers were selected sequentially based on pre-determined criteria: (1) the presence of binary cannabis variables, (2) the number of suicide outcome variables, or (3) sample size.
Effect size computation and statistical analysis
All analyses were conducted using Stata version 14.0 (StataCorp, Reference StataCorp2013). Summary effects were calculated using a random-effects model with DerSimonian-Laird estimation, anticipating variability in study methods, measures, and sample characteristics. Six models were developed to summarize effects across different suicide-related outcomes: (1) odds of suicide death, (2) risks of suicide death, (3) odds of attempted suicide, (4) risks of attempted suicide, (5) odds of suicidal ideation (including suicidal thoughts and plans), and (6) risks of suicidal ideation. The primary outcome metrics were Odds Ratios (ORs) and Hazard Ratios (HRs). When available, unadjusted ORs and HRs were extracted from all eligible studies. If these were not reported, adjusted HRs were used, or both ORs and their 95% confidence intervals were estimated from available descriptive statistics using 2 x 2 tables based on standard computational techniques for dichotomous data (Fleiss & Berlin, Reference Fleiss and Berlin2009). For studies presenting correlational or between-group analyses for continuous data, the correlation statistic, sample size, mean, standard deviation, and significance (p-value) were converted to ORs using Comprehensive Meta-Analysis (CMA) software (Borenstein, Reference Borenstein, Egger, Higgins and Smith2022).
Heterogeneity was assessed using Cochran’s Q and I 2 statistics to determine and quantify the variance due to true heterogeneity (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). An I 2 value below 25% was considered low inconsistency, 25%–75% medium inconsistency, and above 75% high inconsistency. Publication and selection bias were evaluated by visual inspection of funnel plots and supplemented by Egger’s test for funnel plot asymmetry (Egger et al., Reference Egger, Smith, Schneider and Minder1997). Due to the risk of inflated type-1 error with Egger’s test, particularly for dichotomous outcomes with small sample sizes, it was only applied to outcomes with ten or more data points (Deeks et al., Reference Deeks, Higgins, Altman, Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2019). The ‘Trim-and-Fill’ method was also used to adjust for publication or selection bias on outcomes with ten or more data points (Duval & Tweedie, Reference Duval and Tweedie2000).
Sensitivity analyses were performed to assess the impact of omitting specific studies on the overall summary estimates when studies held significant weighting or outliers were identified. For analyses with significant heterogeneity and including ten or more data points (Deeks et al., Reference Deeks, Higgins, Altman, Higgins, Thomas, Chandler, Cumpston, Li, Page and Welch2019), meta-regression analysis was used to explore whether factors such as publication year, age of sample, or baseline PANSS total score influenced the relationship between cannabis use and suicide-related outcomes. Subgroup analyses were also conducted to examine the effects of gender, study design, study quality, or illness course (first episode vs. chronic schizophrenia samples).
Results
60,522 studies were identified via database search. After title screening, 3,950 articles were screened by abstract, 159 full texts were scrutinized and 30 were included in the final coding phase. We excluded five studies at this stage as there was insufficient data to calculate summary effects. However, four of these studies were re-instated following contact with authors who shared previously unpublished data (Fekih-Romdhane et al., Reference Fekih-Romdhane, Abassi, Ghrissi, Loch, Cherif, Damak and Cheour2023; Fridman et al., Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023; Golay et al., Reference Golay, Reitzel and Conus2023; Lahteenvuo et al., Reference Lähteenvuo, Batalla, Luykx, Mittendorfer-Rutz, Tanskanen, Tiihonen and Taipale2021). This resulted in 29 studies comprising 36 separate samples (see Figure 1). Of these, eight samples concerned cannabis and suicide death, 22 concerned cannabis and attempted suicide, and six concerned cannabis and suicidal ideation (See Table 1 and Appendix B and C).
Figure 1. PRISMA Diagram.
Table 1:. Summary of eligible studies
Note: A, Abuse; D, Dependence; FEP, First Episode Psychosis; GAP, Genetics and Psychosis.
Study characteristics
Twenty-nine studies were appraised: eighteen employed a cross-sectional/case control design and eleven utilized a longitudinal design. Studies took place at various sites worldwide: eighteen were conducted across Europe (Sweden, United Kingdom, Denmark, Finland, France, Hungary, Italy, Spain, Switzerland, and the Netherlands); four were conducted across North America (United States, Canada); three were conducted in Australia; three were conducted across Africa (Morocco, Tunisia, South Africa); and one was conducted in Asia (Israel). Overall, sixteen studies included people with a chronic schizophrenia diagnosis and thirteen utilized a first episode psychosis sample. Cannabis use was assessed primarily by the presence of absence of CUD: fifteen studies using ICD, DSM, DIGS, CIDI, or MINI diagnostic interviews; six via scrutiny of medical records; and four via study questionnaire. The remaining four studies utilized either validated measures comprising the Drug Use Scale (DUS; Drake et al., Reference Drake, Mueser and McHugo1996), the Cannabis Use Disorder Identification Test – Revised (CUDIT-R; Adamson et al., Reference Adamson, Kay-Lambkin, Baker, Lewin, Thornton, Kelly and Sellman2010), and Cannabis Experience Questionnaire (CEQEU-GEI; Di Forti et al., Reference Di Forti, Quattrone, Freeman, Tripoli, Gayer-Anderson, Quigley and van der Ven2019) or clinician-rated items pertaining to the frequency of cannabis use over the last month.
Quality assessment
Of the 29 eligible studies, one was deemed to be of weak quality, seventeen were deemed to be of moderate quality, and eleven were deemed to be of strong quality. Across studies, most common weaknesses were due to selection bias (representativeness of sample to wider population), and presence of confounders (See Appendix D).
Association between cannabis use and suicide death
Four papers quantified the association between cannabis use and suicide death using ORs (Bornheimer et al., Reference Bornheimer, Bagge, Overholser, Brdar, Matta, Kitchen and Stockmeier2024; Dutta et al., Reference Dutta, Murray, Allardyce, Jones and Boydell2011; Koola et al., Reference Koola, McMahon, Wehring, Liu, Mackowick, Warren and Kelly2012; Reutfors et al., Reference Reutfors, Brandt, Jönsson, Ekbom, Sparén and Ösby2009). Two studies employed a cross-sectional/case control design (Bornheimer et al., Reference Bornheimer, Bagge, Overholser, Brdar, Matta, Kitchen and Stockmeier2024; Reutfors et al., Reference Reutfors, Brandt, Jönsson, Ekbom, Sparén and Ösby2009), and two utilized a longitudinal design (Dutta et al., Reference Dutta, Murray, Allardyce, Jones and Boydell2011; Koola et al., Reference Koola, McMahon, Wehring, Liu, Mackowick, Warren and Kelly2012). Utilizing a retrospective case control design, Reutfors et al. (Reference Reutfors, Brandt, Jönsson, Ekbom, Sparén and Ösby2009) reported a non-significant relationship between cannabis use and suicide death in 84 age-matched cases from Sweden with a schizophrenia diagnosis. Bornheimer et al. (Reference Bornheimer, Bagge, Overholser, Brdar, Matta, Kitchen and Stockmeier2024) reported a similar result utilizing a psychological autopsy method. They found that cannabis use was not significantly associated with suicide death in a mixed, heterogeneous, US sample of 57 patients. These studies were rated strong and moderate in quality, respectively, and suggest a non-significant relationship between cannabis use and suicide death in available cross-sectional studies. Comparably, a longitudinal, epidemiological study of several UK patient cohorts with FEP (n = 2,132) (i.e., London, Nottingham, Dumfries, and Galloway) found that cannabis use did not significantly predict suicide death over time (Dutta et al., Reference Dutta, Murray, Allardyce, Jones and Boydell2011) nor did a US study of people with chronic schizophrenia (n = 762) (Koola et al., Reference Koola, McMahon, Wehring, Liu, Mackowick, Warren and Kelly2012). While these studies were rated as strong and moderate in quality, respectively, owing to few suicide cases in the latter study, this finding should be interpreted with caution. Collectively, the meta-analysis showed that cannabis use did not significantly increase the odds of suicide in people with schizophrenia (k = 4, OR = 1.34, 95% CI = 0.81 – 2.21) (See Figure 2). Q and I 2 tests indicated that between-sample heterogeneity was low and non-significant (I 2 = 14.9%, p = 0.32).
Figure 2. Forest plot for the meta-analysis examining odds ratios between cannabis use and suicide.
Three papers comprising four samples reported HRs (Hjorthoj et al., Reference Hjorthøj, Østergaard, Benros, Toftdahl, Erlangsen, Andersen and Nordentoft2015; Lahteenvuo et al., Reference Lähteenvuo, Batalla, Luykx, Mittendorfer-Rutz, Tanskanen, Tiihonen and Taipale2021; Ostergaard et al., Reference Østergaard, Nordentoft and Hjorthøj2017). All studies employed a longitudinal design and analyzed nation-wide, register-based cohorts, prospectively. In a large scale, cohort, study of Danish people with schizophrenia (n = 41,470), Hjorthoj et al. (Reference Hjorthøj, Østergaard, Benros, Toftdahl, Erlangsen, Andersen and Nordentoft2015) found that cannabis use was not significantly associated with risks of suicide death over time. However, this result just failed to reach significance. Ostergaard et al. (Reference Østergaard, Nordentoft and Hjorthøj2017) reported similar findings in a separate cohort of Danish people with a schizophrenia-spectrum diagnosis (n = 35,625). Both current and past cannabis use were not significantly associated with risks of suicide, but only current use just failed to reach significance. Comparatively, whereas Lahteenvuo et al. (Reference Lähteenvuo, Batalla, Luykx, Mittendorfer-Rutz, Tanskanen, Tiihonen and Taipale2021) found a significant relationship between cannabis use and suicide in a cohort of Finnish people with schizophrenia and substance use, this finding was not replicated in a parallel Swedish cohort. However, differences in the length of follow up could partially explain these results (Finnish cohort – 22 years, Swedish cohort – 11 years). All available longitudinal studies were rated moderate in quality and provided a mixed picture of the relationship between cannabis use and risks of suicide. Yet, collectively, the meta-analysis showed that cannabis use significantly increased the risk of suicide death in people with schizophrenia (k = 4, HR = 1.21, 95% CI = 1.04 – 1.40) (See Figure 3). Between-sample heterogeneity was low and non-significant (I 2 = 8.5%, p = 0.35). This result persisted after the removal of studies holding significant weighting (Ostergaard et al., Reference Østergaard, Nordentoft and Hjorthøj2017) (k = 3, HR = 1.37, 95% CI = 1.07 – 1.77).
Figure 3. Forest plot for the meta-analysis examining hazard ratios between cannabis use and suicide.
Association between cannabis use and attempted suicide
Overall, cannabis use significantly increased the odds of attempted suicide in people with a diagnosis of schizophrenia (k = 21, OR = 1.40, 95% CI = 1.16 – 1.68) (See Figure 4). Furthermore, data from one large-scale, nation-wide, Danish prospective cohort study (n = 35,625) found that cannabis use significantly increased the risk of attempted suicide (HR = 1.16, 95% CI = 1.05 – 1.34) (Ostergaard et al., Reference Østergaard, Nordentoft and Hjorthøj2017). Q and I 2 tests indicated that between-sample heterogeneity was medium and significant in analyses of odds between cannabis use and attempted suicide (I 2 = 39.6%, p = 0.03). Publication or other selection bias was low and non-significant (β = .22, SE = .13, p = 0.55) (See Appendix E). Trim and Fill analysis also identified no missing studies. Across all analyses of attempted suicide, one study was an outlier. Sensitivity analyses revealed the relationship between cannabis use and attempted suicide decreased but remained significant following the removal of Dervaux et al. (Reference Dervaux, Laqueille, Bourdel, LeBorgne, Olié, Lóo and Krebs2003) whose outcome metrics lay outside of the 95% CIs of the aggregated summary effect (k = 20, OR = 1.34, 95% CI = 1.14 – 1.57) (See Appendix F).
Figure 4. Forest plot for the meta-analysis examining odds ratios between cannabis use and attempted suicide.
Subgroup analyses showed that cannabis use was significantly associated with attempted suicide in cross-sectional / case control studies (k = 17, OR = 1.44, 95% CI = 1.14 – 1.81, I 2 = 48.7%, Q = 31.18), but effects just failed to reach significance in longitudinal studies (k = 4, OR = 1.40, 95% CI = 0.97 – 1.68, I 2 = 0.0%, Q = 1.86) (See Appendix G & K). Furthermore, cannabis use was significantly associated with attempted suicide in studies using male only (k = 3, OR = 1.68, 95% CI = 1.16 – 2.44, I 2 = 0%, Q = 1.08), female only (k = 2, OR = 2.07, 95% CI = 1.27 – 3.39, I 2 = 0%, Q = 0.27), and gender combined samples (k = 14, OR = 1.33, 95% CI = 1.04 – 1.69, I 2 = 49.9%, Q = 25.94) (See Appendix H & K). In addition, whereas summary effects of studies investigating samples with First Episode Psychosis (FEP) just failed to reach significance (k = 9, OR = 1.24, 95% CI = 0.99 – 1.55, I 2 = 0.0%, Q = 5.70), those employing chronic presentations were significant (k = 12, OR = 1.53, 95% CI = 1.16 – 2.03, I 2 = 59.2%, Q = 26.97) (See Appendix I & K). Summary effects were also significant in studies rated as strong (k = 9, OR = 1.25, 95% CI = 1.06 – 1,47, I 2 = 0.0%, Q = 7.35) and moderate in methodological quality (k = 11, OR = 1.48, 95% CI = 1.01 – 2.17, I 2 = 46.3%, Q = 18.61). One study rated weak in quality was also significant (Koubaa et al., Reference Koubaa, Aden and Barrimi2023) (OR = 2.77, 95% CI = 1.58 – 4.85) (See Appendix J & K). Meta-regression showed that neither publication year (k = 21, β = 0.99, SE = 0.02, p = 0.79), mean age of cases (k = 9, β = 1.0, SE = 0.03, p = 0.86), nor baseline total PANSS score (k = 4, β = 0.99, SE = 0.04, p = 0.88) influenced the overall summary estimate.
Association between cannabis use and suicidal ideation
Six papers provided statistical information sufficient to calculate the association between cannabis use and suicidal ideation using ORs (Fridman et al., Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023; Heuschen et al., Reference Heuschen, Bolhuis, Zantvoord, Bockting, Denys, Lok and Schirmbeck2024; Phalen et al., Reference Phalen, Jones, Davis, Sarpal, Dickerson, Vatza and Bennett2024; Ricci et al., Reference Ricci, Di Muzio, Ceci, Di Carlo, Mancusi, Piro and Maina2024; Salagre et al., Reference Salagre, Grande, Jiménez, Mezquida, Cuesta and Llorente2021; Sicotte et al., Reference Sicotte, Iyer, Lacourse, Séguin and Abdel-Baki2023). Two studies utilized a cross-sectional/case control design (Fridman et al., Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023; Heuschen et al., Reference Heuschen, Bolhuis, Zantvoord, Bockting, Denys, Lok and Schirmbeck2024) and four employed a longitudinal research design (Phalen et al., Reference Phalen, Jones, Davis, Sarpal, Dickerson, Vatza and Bennett2024; Ricci et al., Reference Ricci, Di Muzio, Ceci, Di Carlo, Mancusi, Piro and Maina2024; Salagre et al., Reference Salagre, Grande, Jiménez, Mezquida, Cuesta and Llorente2021; Sicotte et al., Reference Sicotte, Iyer, Lacourse, Séguin and Abdel-Baki2023).
Utilizing a retrospective cross-sectional design, Fridman et al. (Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023) compared the frequency of suicidal ideation, extracted from medical records, in cannabis and non-cannabis using Israeli patients with a diagnosis of schizophrenia (n = 144). Reporting previously unpublished data, they found a non-significant relationship between cannabis use and suicidal ideation. In a larger study, Heuschen et al. (Reference Heuschen, Bolhuis, Zantvoord, Bockting, Denys, Lok and Schirmbeck2024) reported similar findings in a Dutch cohort of individuals with FEP (n = 551). The regression found that cannabis use did not significantly predict suicidal ideation as measured by an item on the Community Assessment of Psychic Experiences (CAPE). Cross-sectional studies were categorized as moderate (Fridman et al., Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023) and strong (Heuschen et al., Reference Heuschen, Bolhuis, Zantvoord, Bockting, Denys, Lok and Schirmbeck2024) in quality and together suggest a non-significant relationship between cannabis use and suicidal ideation.
A similar pattern of non-significant results was found across all longitudinal studies. Ricci et al. (Reference Ricci, Di Muzio, Ceci, Di Carlo, Mancusi, Piro and Maina2024) compared differences in suicidal ideation between Italian patients with FEP (n = 57) grouped by self-reported cannabis, spice, and non-cannabis use. After combining results across multiple time-points, Ricci and colleagues found a non-significant relationship between cannabis use and suicidal ideation. Furthermore, employing similar research designs, both Salagre et al. (Reference Salagre, Grande, Jiménez, Mezquida, Cuesta and Llorente2021) and Sicotte et al. (Reference Sicotte, Iyer, Lacourse, Séguin and Abdel-Baki2023) compared cannabis use over time in groups categorised by trajectory of suicidal ideation (i.e., non vs. worsening, and low vs. persistent, respectively). Both studies found non-significant relationships between cannabis use and suicidal ideation in both Canadian (n = 382) and Spanish (n = 334) samples with FEP. Lastly, in a study of US patients with FEP (n = 1101), Phalen et al. (Reference Phalen, Jones, Davis, Sarpal, Dickerson, Vatza and Bennett2024) reported relationships between cannabis use and both clinician-rated and self-reported suicidal ideation. When combined, they found a non-significant relationship between cannabis use and suicidal ideation, though this just failed to reach significance. All longitudinal studies were rated as moderate in quality, except for Phalen colleagues (2024), which was rated as strong.
Collectively, the meta-analysis showed that cannabis did not significantly increase the odds of suicidal ideation in people with a diagnosis of schizophrenia, but this just failed to reach significance (k = 6, OR = 1.16, 95% CI = 0.97 – 1.38) (See Figure 5). Between sample heterogeneity was low and non-significant (I 2 = 0.6%, p = 0.41). The summary effect was further reduced after the removal of studies holding significant weighting (Phalen et al., Reference Phalen, Jones, Davis, Sarpal, Dickerson, Vatza and Bennett2024) (k = 5, OR = 1.04, 95% CI = 0.77 – 1.40).
Figure 5. Forest plot for the meta-analysis examining odds ratios between cannabis use and suicidal ideation.
Discussion
This review found that cannabis use is significantly associated with some, but not all, suicide-related outcomes in people with a diagnosis of schizophrenia. Specifically, while cannabis use was not associated with odds of suicide death or suicidal ideation, it was associated with risks of death and odds of attempted suicide. Summary effects remained significant in most sub-groups (i.e., in cross-sectional studies, in studies rated as strong, moderate or weak in quality, in studies comprising male-only, female-only, and gender combined samples and in studies sampling individuals with chronic schizophrenia), but just failed to reach significance in longitudinal studies of attempted suicide and studies investigating attempted suicide in first episode samples. While there was evidence of moderate homogeneity, there was no evidence of publication bias and no missing studies. Neither publication year, sample age or PANSS score significantly predicted the summary estimates.
Our findings advance existing work on the relationship between cannabis use and suicide-related outcomes in people with a diagnosis of schizophrenia. Existing reviews have concluded that research supporting the link between cannabis and suicide in this group is inconclusive (Ricci et al., Reference Ricci, Cristofori, Passarello, Paggi, Cavallo, Ceci and Maina2023b; Serafini et al., Reference Serafini, Pompili, Innamorati, Rihmer, Sher and Girardi2012). However, by grouping studies by suicide outcome, we found evidence for a significant link between cannabis use, risks of suicide, and odds of attempted suicide specifically. These findings lend support for individual studies reporting positive associations between cannabis use and both suicide death and attempted suicide (Koubaa et al., Reference Koubaa, Aden and Barrimi2023; Lahteenvuo et al., Reference Lähteenvuo, Batalla, Luykx, Mittendorfer-Rutz, Tanskanen, Tiihonen and Taipale2021; Waterreus et al., Reference Waterreus, Di Prinzio, Badcock, Martin-Iverson, Jablensky and Morgan2018) and suggest that cannabis use should be considered a risk factor for adverse outcomes (WHO, 2016). Interestingly, we did not find a significant link between cannabis use and either odds of suicide death or suicidal ideation, especially following the removal of studies holding significant weighting. This lends partial support for anecdotal qualitative accounts on the potential buffering effects of cannabis use on hopelessness, boredom, and anxiety (Asher & Gask, Reference Asher and Gask2010; Parshotam & Joubert, Reference Parshotam and Joubert2015), experiences known to precipitate suicidal thinking (Ben-Zeev et al., Reference Ben-Zeev, Young and Depp2012; Chong et al., Reference Chong, Wahab, Muthukrishnan, Tan, Ch’ng and Yoong2020). Furthermore, the discrepancy in significance between cannabis use and suicide death as measured by reporting of ORs or HRs could suggest that while cannabis might not confer a significantly greater odds of suicide death, it might influence the speed at which suicide deaths occur in people with a diagnosis of schizophrenia.
Several potential mechanisms could underlie the summary effects observed in this meta-analysis. Firstly, cannabis, and specifically THC, could directly affect serotonin and neurotransmitter function to produce symptoms of depression and / or other mental health conditions known to increase suicidal risk and attempts (Pedersen, Reference Pedersen2008). Moreover, the role of cannabinoid-1 receptor disruption and dysregulation of the endocannabinoid system following cannabis use have also been implicated in the emergence of suicidal behavior (Ceccarini et al., Reference Ceccarini, Kuepper, Kemels, van Os, Henquet and Van Laere2015; Volkow et al., Reference Volkow, Hampson and Baler2017). Secondly, cannabis use might be indirectly related to suicide via other mechanisms. It is widely known that cannabis intoxication produces changes in behavioral, cognitive, and motor functions, including impaired problem-solving and impulsivity (Trull et al., Reference Trull, Wycoff, Lane, Carpenter and Brown2016). Cannabis use has also been associated with antipsychotic medication treatment resistance (Reid & Bhattacharyya, Reference Reid and Bhattacharyya2019), non-compliance (Patel et al., Reference Patel, Sreeram, Vadukapuram and Baweja2020), and symptomatic relapse (Bioque et al., Reference Bioque, Mezquida, Amoretti, García-Rizo, López-Ilundain, Diaz-Caneja and Bernardo2022; Hindley et al., Reference Hindley, Beck, Borgan, Ginestet, McCutcheon, Kleinloog and Howes2020), all of which correlate with suicide attempts (Huang et al., Reference Huang, Fox, Ribeiro and Franklin2018). However, future studies employing repeated, longitudinal measures of psychosis, medication adherence, cognition, impulsivity, cannabis use, and suicidality are required to examine temporal relationships between these variables and permit more nuanced conclusions regarding underlying mediators and moderators. Lastly, it could be that cannabis use is related to suicide and attempted suicide in combination with other substances. Epidemiological data suggests cannabis might be a gateway drug that enhances vulnerability to the rewarding effects of other substances (Anthony, Reference Anthony2012), including alcohol (Ceccarini et al., Reference Ceccarini, Hompes, Verhaeghen, Casteels, Peuskens, Bormans and Van Laere2014), itself correlated with suicidal outcomes (Mulligan et al., Reference Mulligan, Varese, Harris and Haddock2024b). Therefore, future research should attempt to disentangle the relative effects of individual substances on suicide-related outcomes to inform a greater understanding of drug-specific mechanisms.
To our knowledge, this is the first ever attempt to quantify the relationship between cannabis use and suicide-related outcomes in people with a diagnosis of schizophrenia and explore potential sources of heterogeneity. We also included unpublished data from existing sources (Fekih-Romdhane et al., Reference Fekih-Romdhane, Abassi, Ghrissi, Loch, Cherif, Damak and Cheour2023; Fridman et al., Reference Fridman, Bloemhof-Bris, Weizman, Kessler, Porat, Ivry and Shelef2023; Golay et al., Reference Golay, Reitzel and Conus2023; Lahteenvuo et al., Reference Lähteenvuo, Batalla, Luykx, Mittendorfer-Rutz, Tanskanen, Tiihonen and Taipale2021). However, our findings should be considered in the context of the following limitations. Firstly, the summary effects for suicide death and suicidal ideation, and sub-group analyses for attempted suicide, were based on a small number of pooled studies. Whilst our searches were exhaustive, the small number of eligible studies means some reported summary effects may lack precision and should therefore be interpreted cautiously. Secondly, a binary measure of cannabis use (i.e., absence or presence of CUD) was chosen as the main predictor across all analyses as this was reported by most primary studies. The amount of THC could not be determined or examined. This is notable as there is evidence that the onset and frequency of cannabis use as well as cannabis potency may be more important predictors of adverse outcomes, including suicide, in people with psychosis (Di Forti et al., Reference Di Forti, Sallis, Allegri, Trotta, Ferraro, Stilo and Murray2014; Quattrone et al., Reference Quattrone, Ferraro, Tripoli, La Cascia, Quigley, Quattrone and Di Forti2021). There are also differences in the route of cannabis exposure (i.e., inhalation vs. ingestion) and clinical outcome (Monte et al., Reference Monte, Shelton, Mills, Saben, Hopkinson, Sonn and Abbott2019), which could explain between study differences in reported summary effects. Finally, it is unclear whether or how the legality of THC cannabis may lead to biases in data collection and participant selection. Therefore, the representativeness of the samples included in this meta-analysis could be questionable.
In conclusion, cannabis use is significantly associated with some, but not all, suicide-related outcomes in people with a diagnosis of schizophrenia. Specifically, summary effects were significant for suicide risks and attempted suicide, which persisted across most sub-groups. However, there was no significant relationship with suicide odds or suicidal ideation. While further research is needed to examine potential mechanisms of significant effects, clinicians should inquire about cannabis use, inform patients and carers about its effects, and support those who wish to abstain as part of routine clinical practice, psychoeducation and suicide-risk prevention. This is important given the increasing prevalence and continued legalization and de-criminalization of cannabis use globally (UNODC, 2023, 2013; Wang et al., Reference Wang, Qin, Xing, Zhu and Jia2024) and international public health priorities regarding the prevention of suicide in people with a diagnosis of schizophrenia.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725000236.
Data Sharing
Please contact the corresponding author if you would like to see any data that are not included in the Article or the appendix.
Acknowledgments
This research was supported by the NIHR Manchester Biomedical Research Centre (NIHR203308). The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.
Author contribution
LM, FV, KH, and GH designed the study. LM and KH conducted the literature search, assessed studies for eligibility, extracted all data, and rated study quality. LM analyzed the data and interpreted them together with FV and GH. LM, FV, KH, and GH critically reviewed the report for important intellectual content and approved the final submitted version. LM had final responsibility for the decision to submit for publication.
Competing interests
We declare no competing interests.
Open access
