Intended for healthcare professionals

CCBYNC Open access
Research

Balancing risks and benefits of cannabis use: umbrella review of meta-analyses of randomised controlled trials and observational studies

BMJ 2023; 382 doi: https://doi.org/10.1136/bmj-2022-072348 (Published 30 August 2023) Cite this as: BMJ 2023;382:e072348

Linked Editorial

Benefits and risks of cannabinoids

  1. Marco Solmi, associate professor1234567,
  2. Marco De Toffol, psychiatrist8,
  3. Jong Yeob Kim, public health doctor9,
  4. Min Je Choi, psychiatrist9,
  5. Brendon Stubbs, advanced fellow1011,
  6. Trevor Thompson, associate professor of clinical research12,
  7. Joseph Firth, research fellow1314,
  8. Alessandro Miola, psychiatrist15,
  9. Giovanni Croatto, psychiatrist16,
  10. Francesca Baggio, psychiatrist16,
  11. Silvia Michelon, psychiatrist17,
  12. Luca Ballan, psychiatrist17,
  13. Björn Gerdle, professor18,
  14. Francesco Monaco, psychiatrist1920,
  15. Pierluigi Simonato, psychiatrist21,
  16. Paolo Scocco, psychiatrist22,
  17. Valdo Ricca, Professor23,
  18. Giovanni Castellini, associate professor23,
  19. Michele Fornaro, assistant professor24,
  20. Andrea Murru, researcher25,
  21. Eduard Vieta, professor25,
  22. Paolo Fusar-Poli, professor526,
  23. Corrado Barbui, professor27,
  24. John P A Ioannidis, professor282930,
  25. Andrè F Carvalho, senior researcher31,
  26. Joaquim Radua, associate professor32,
  27. Christoph U Correll, professor73334,
  28. Samuele Cortese, professor635363738,
  29. Robin M Murray, professor39,
  30. David Castle, professor4041,
  31. Jae Il Shin, professor4243,
  32. Elena Dragioti, associate professor1844
  1. 1Department of Psychiatry, University of Ottawa, Ontario, ON, Canada
  2. 2On Track: The Champlain First Episode Psychosis Program, Department of Mental Health, The Ottawa Hospital, Ontario, ON, Canada
  3. 3Ottawa Hospital Research Institute, Clinical Epidemiology Program, University of Ottawa, Ottawa, ON, Canada
  4. 4School of Epidemiology and Public Health, Faculty of Medicine, University of Ottawa, Ottawa, ON, Canada
  5. 5Early Psychosis: Interventions and Clinical detection Lab, Institute of Psychiatry, Psychology and Neuroscience, Department of Psychosis Studies, King's College London, London, UK
  6. 6Centre for Innovation in Mental Health-Developmental Lab, School of Psychology, University of Southampton, and NHS Trust, Southampton, UK
  7. 7Department of Child and Adolescent Psychiatry, Charité Universitätsmedizin, Berlin, Germany
  8. 8Psychiatry Unit, Veris Delli Ponti Scorrano Hospital, Department of Mental Health, ASL Lecce, Lecce, Italy
  9. 9Yonsei University College of Medicine, Seoul, South Korea
  10. 10Department of Psychological Medicine, Institute of Psychiatry, Psychology and Neuroscience, King's College London, London, UK
  11. 11Physiotherapy Department, South London and Maudsley NHS Foundation Trust, London, UK
  12. 12Centre of Chronic Illness and Ageing, University of Greenwich, London, UK
  13. 13Division of Psychology and Mental Health, University of Manchester, Manchester Academic Health Science Centre, Manchester, UK
  14. 14Greater Manchester Mental Health NHS Foundation Trust, Manchester Academic Health Science Centre, Manchester, UK
  15. 15Neurosciences Department, Padua Neuroscience Center, University of Padua, Italy
  16. 16Mental Health Department, AULSS 3 Serenissima, Mestre, Venice, Italy
  17. 17Department of Mental Health, AULSS 7 Pedemontana Veneto, Italy
  18. 18Pain and Rehabilitation Centre, Department of Health, Medicine and Caring Sciences, Linköping University, Linköping, Sweden
  19. 19Department of Mental Health, Asl Salerno, Salerno, Italy
  20. 20European Biomedical Research Institute of Salerno, Salerno, Italy
  21. 21Department of Clinical, Pharmaceutical and Biological Sciences, School of Life and Medical Sciences, University of Hertfordshire, Hatfield, UK
  22. 22Mental Health Department, ULSS 6 Euganea, Padova, Italy
  23. 23Psychiatry Unit, Department of Health Sciences, University of Florence, Florence, Italy
  24. 24Section of Psychiatry, Department of Neuroscience, University School of Medicine Federico II, Naples, Italy
  25. 25Institute of Neuroscience, Hospital Clinic, University of Barcelona, IDIBAPS, CIBERSAM, Barcelona, Catalonia, Spain
  26. 26Department of Brain and Behavioural Sciences, University of Pavia, Pavia, Italy
  27. 27WHO Collaborating Centre for Research and Training in Mental Health and Service Evaluation, Department of Neuroscience, Biomedicine and Movement Sciences, Section of Psychiatry, University of Verona, Verona, Italy
  28. 28Meta-Research Innovation Center at Stanford, Stanford University, Stanford, CA, USA
  29. 29Meta-Research Innovation Center Berlin, Berlin Institute of Health, Charité Universitätsmedizin, Berlin, Germany
  30. 30Departments of Medicine, of Epidemiology and Population Health, of Biomedical Data Science, and of Statistics, Stanford University, Stanford, CA, USA
  31. 31IMPACT - The Institute for Mental and Physical Health and Clinical Translation, School of Medicine, Barwon Health, Deakin University, Geelong, VIC, Australia
  32. 32Institut d'Investigacions Biomediques August Pi i Sunyer, CIBERSAM, Instituto de Salud Carlos III, University of Barcelona, Barcelona, Spain
  33. 33Department of Psychiatry, Zucker Hillside Hospital, Northwell Health, Glen Oaks, NY, USA
  34. 34Department of Psychiatry and Molecular Medicine, Donald and Barbara Zucker School of Medicine at Hofstra/Northwell, Hempstead, NY, USA
  35. 35Clinical and Experimental Sciences (Central Nervous System and Psychiatry), Faculty of Medicine, University of Southampton, Southampton, UK
  36. 36Solent NHS Trust, Southampton, UK
  37. 37Division of Psychiatry and Applied Psychology, School of Medicine, University of Nottingham, Nottingham, UK
  38. 38Hassenfeld Children's Hospital at NYU Langone, New York University Child Study Center, New York City, New York, NY, USA
  39. 39Psychosis Studies, Institute of Psychiatry, Psychology and Neuroscience, King's College of London, London, UK
  40. 40Department of Psychiatry, University of Tasmania, Sandy Bay, TAS, Australia
  41. 41Co-Director, Centre for Mental Health Service Innovation, Department of Health, Tasmania, Australia
  42. 42Department of Pediatrics, Yonsei University College of Medicine, Seoul, South Korea
  43. 43Severance Underwood Meta-research Center, Institute of Convergence Science, Yonsei University, Seoul, South Korea
  44. 44Research Laboratory Psychology of Patients, Families and Health Professionals, Department of Nursing, School of Health Sciences, University of Ioannina, Ioannina, Greece
  1. Correspondence to: M Solmi msolmi{at}toh.ca
  • Accepted 27 June 2023

Abstract

Objective To systematically assess credibility and certainty of associations between cannabis, cannabinoids, and cannabis based medicines and human health, from observational studies and randomised controlled trials (RCTs).

Design Umbrella review.

Data sources PubMed, PsychInfo, Embase, up to 9 February 2022.

Eligibility criteria for selecting studies Systematic reviews with meta-analyses of observational studies and RCTs that have reported on the efficacy and safety of cannabis, cannabinoids, or cannabis based medicines were included. Credibility was graded according to convincing, highly suggestive, suggestive, weak, or not significant (observational evidence), and by GRADE (Grading of Recommendations, Assessment, Development and Evaluations) (RCTs). Quality was assessed with AMSTAR 2 (A Measurement Tool to Assess Systematic Reviews 2). Sensitivity analyses were conducted.

Results 101 meta-analyses were included (observational=50, RCTs=51) (AMSTAR 2 high 33, moderate 31, low 32, or critically low 5). From RCTs supported by high to moderate certainty, cannabis based medicines increased adverse events related to the central nervous system (equivalent odds ratio 2.84 (95% confidence interval 2.16 to 3.73)), psychological effects (3.07 (1.79 to 5.26)), and vision (3.00 (1.79 to 5.03)) in people with mixed conditions (GRADE=high), improved nausea/vomit, pain, spasticity, but increased psychiatric, gastrointestinal adverse events, and somnolence among others (GRADE=moderate). Cannabidiol improved 50% reduction of seizures (0.59 (0.38 to 0.92)) and seizure events (0.59 (0.36 to 0.96)) (GRADE=high), but increased pneumonia, gastrointestinal adverse events, and somnolence (GRADE=moderate). For chronic pain, cannabis based medicines or cannabinoids reduced pain by 30% (0.59 (0.37 to 0.93), GRADE=high), across different conditions (n=7), but increased psychological distress. For epilepsy, cannabidiol increased risk of diarrhoea (2.25 (1.33 to 3.81)), had no effect on sleep disruption (GRADE=high), reduced seizures across different populations and measures (n=7), improved global impression (n=2), quality of life, and increased risk of somnolence (GRADE=moderate). In the general population, cannabis worsened positive psychotic symptoms (5.21 (3.36 to 8.01)) and total psychiatric symptoms (7.49 (5.31 to 10.42)) (GRADE=high), negative psychotic symptoms, and cognition (n=11) (GRADE=moderate). In healthy people, cannabinoids improved pain threshold (0.74 (0.59 to 0.91)), unpleasantness (0.60 (0.41 to 0.88)) (GRADE=high). For inflammatory bowel disease, cannabinoids improved quality of life (0.34 (0.22 to 0.53) (GRADE=high). For multiple sclerosis, cannabinoids improved spasticity, pain, but increased risk of dizziness, dry mouth, nausea, somnolence (GRADE=moderate). For cancer, cannabinoids improved sleep disruption, but had gastrointestinal adverse events (n=2) (GRADE=moderate). Cannabis based medicines, cannabis, and cannabinoids resulted in poor tolerability across various conditions (GRADE=moderate). Evidence was convincing from observational studies (main and sensitivity analyses) in pregnant women, small for gestational age (1.61 (1.41 to 1.83)), low birth weight (1.43 (1.27 to 1.62)); in drivers, car crash (1.27 (1.21 to 1.34)); and in the general population, psychosis (1.71 (1.47 to 2.00)). Harmful effects were noted for additional neonatal outcomes, outcomes related to car crash, outcomes in the general population including psychotic symptoms, suicide attempt, depression, and mania, and impaired cognition in healthy cannabis users (all suggestive to highly suggestive).

Conclusions Convincing or converging evidence supports avoidance of cannabis during adolescence and early adulthood, in people prone to or with mental health disorders, in pregnancy and before and while driving. Cannabidiol is effective in people with epilepsy. Cannabis based medicines are effective in people with multiple sclerosis, chronic pain, inflammatory bowel disease, and in palliative medicine but not without adverse events.

Study registration PROSPERO CRD42018093045.

Funding None.

Introduction

Cannabis contains over 100 cannabinoids, of which Δ9-tetrahydrocannabinol and cannabidiol are the most clinically relevant. Tetrahydrocannabinol is a partial agonist at CB1 and binds CB2 receptors. CB1 is widely expressed by central and peripheral neurones but also by immune cells and other type of cells in the brain and in the periphery, and when it binds with tetrahydrocannabinol, a so-called high is induced, which is responsible for potential misuse. CB2 receptors are also expressed by neurons, but less than CB1, and are most abundantly expressed in immune cells.123 Cannabidiol, however, does not produce the high and thus does not carry the same potential for substance misuse.4 Furthermore, cannabidiol does not seem to promote psychosis inducing effects.5 Cannabis use can evolve into cannabis use disorder, broadly defined as an inability to quit cannabis use, continuous use despite harmful consequences (eg, cannabinoid hyperemesis syndrome6), or functional impairment.78

According to the Global Burden of Disease 2019 study, more than 23.8 million people have cannabis use disorder globally,9 and cannabis use ranks third worldwide among consumed substances of misuse, after alcohol and tobacco.10111213 Cannabis use disorder is more common in men and high income countries. The prevalence of cannabis use disorder in the USA has been estimated to be around 6.3% in a lifetime and 2.5% for 12 months, and in Europe, around 15% of people aged 15-35 years reported cannabis use in the previous year.14 Of those using cannabis, one in three developed problems related to cannabis use that impaired functioning,13 and 10% used cannabis on a daily basis.15 Cannabis use disorder can affect up to 50% of people who use cannabis daily.16

In Europe, over the past decade, self-reported use of cannabis within the past month has increased by almost 25% in people aged 15-34 years, and more than 80% in people who are 55-64 years.17 Cannabis or products containing tetrahydrocannabinol (cannabinoids) are widely available and have increasingly high tetrahydrocannabinol content.18 For instance, in Europe, tetrahydrocannabinol content increased from 6.9% to 10.6% from 2010 to 2019.17 Evidence has suggested that cannabis may be harmful, for mental1920 and physical health,21 as well as driving safety,22 across observational studies but also in experimental settings.23 Conversely, more than a decade ago, cannabidiol was proposed as a candidate drug for the treatment of neurological disorders such as treatment-resistant childhood epilepsy. Furthermore, it has been proposed that this substance might be useful for anxiety and sleep disorders, and even as an adjuvant treatment for psychosis.24 Moreover, cannabis based medications (ie, medications that contain cannabis components) have been investigated as putative treatments for several different conditions and symptoms.23

The multifarious nature of cannabis’s main active components, contrasting evidence from observational studies reporting detrimental effects of cannabis, and therapeutic findings of cannabis based medicines from interventional studies, is reflected in different legislative approaches. Thus, in most countries cannabis use is illegal, but in a small and growing number of countries and states cannabis is legally sold without the need for a medical prescription.252627

Publication of meta-analyses investigating the effects of cannabinoids on health and other outcomes have substantially increased. However, most meta-analytical findings synthesised data from observational studies and are prone to several sources of bias.2829 To date, no umbrella review has systematically evaluated the evidence around cannabis, cannabinoids and cannabis nased medicines and health outcomes in humans from meta-analyses encompassing both observational studies and randomised controlled trials. Thus, this work aimed to systematically evaluate the breadth, quality, credibility, and certainty of associations between cannabis, cannabinoids, cannabis based medicines, and human health. We aimed to use established quantitative criteria, account for several sources of bias,303132 and identify converging findings from different study designs.

Methods

Searches and inclusion criteria

We conducted an umbrella review of meta-analyses of observational studies(ie, case-control and cohort studies) and randomised controlled trials that reported on any outcome associated with cannabis and cannabinoids use in humans. We followed an a-priori protocol (PROSPERO CRD42018093045). We adhered to PRIOR and PRISMA 2020 guidelines (adapting PRISMA to the abstract of an umbrella review; supplementary tables 1-2).3334 Two of the authors independently screened literature that was retrieved systematically by searching PubMed, Embase, and PsycINFO from database inception up to 9 February 2022, without language restrictions, and extracted data into a spreadsheet. The search key is available in the supplementary methods. We also manually searched the Cochrane Library. When two or more meta-analyses examined the same association, we selected only the one that included the largest number of studies. We excluded systematic reviews without a meta-analysis, meta-analyses of risk factors for cannabinoids use, meta-analyses of cross-sectional studies only, pooled analyses of studies identified without a systematic search, and individual studies.

Outcomes

The co-primary outcomes were the efficacy and safety of cannabinoids on target symptoms (eg seizures in epilepsy) in meta-analyses of randomised controlled trials. The secondary outcomes were any outcome reported in the meta-analyses of observational studies.

Data extraction and quality assessment

Extracted information from meta-analyses and individual studies included in meta-analyses were the bibliographic identifiers of the publication (ie, PubMed-Indexed for Medline or the digital object identifier), first author, year of publication, design of included studies (ie, cohort, case-control, randomised controlled trial), number of included studies in the meta-analysis, specific population under investigation (ie, general population, pregnant women, or people with medical disorders), the exposure and comparison definitions (eg author defined marijuana use v no use or heavy use of cannabis v no use), the outcomes, and their effect size and dispersion measure (when adjusted and unadjusted effect sizes were reported, we selected the adjusted ones). We also extracted what factors analyses were adjusted for. The methodological quality of each included meta-analysis was assessed by two independent investigators using A Measurement Tool to Assess Systematic Reviews version 2 (AMSTAR 2).35

Data analysis

For each association from observational studies (ie, between exposure to cannabis or cannabinoids and outcomes), we extracted the effect sizes of individual studies reported in each meta-analysis, recalculating the pooled effect sizes and 95% confidence intervals, using random effects models. Specifically, we re-analysed each eligible association under the random effects model with DerSimonian and Laird method if included studies were equal or more than 10,36 and Hartung, Knapp, Sidik, and Jonkman if less than 10.37 We transformed the initial effect sizes or modified the direction of associations presented by the original authors to present comparable estimates (ie, equivalent odds ratio; supplementary methods).38 Heterogeneity was tested with the I2 and Tau statistics.39 I2 measures the proportion of the total variability due to heterogeneity, Tau measures true heterogeneity as an absolute measure of heterogeneity, instead. Moreover, 95% prediction intervals for the summary random effect sizes were computed to estimate the possible range in which the effect sizes of future studies were anticipated to fall.40 We calculated prediction intervals using both the estimated between-study heterogeneity variance given from tau2 as well as the standard error of the pooled effect. We then examined small study effect bias (ie, whether smaller studies generated larger effect sizes compared with larger studies).38414243444546 Small study effect was deemed present when both the Egger regression asymmetry test indicated publication bias (P value ≤0.10), and the random effects summary effect size was larger than the effect size of the largest study contributing to that association.42444546 Finally, we evaluated significance bias using an updated method to detect the publication selection of statistically significant findings based on observable excess statistical significance.4748 We computed the test of excess statistical significance and the proportion of statistical significance, which have adequate control for type I errors and high statistical power. The presence of excess significance bias for individual meta-analyses was considered if either excess statistical significance or proportion of statistical significance were greater than 1.645.47

All analyses were conducted in Stata/SE, version 17.0.

Assessment of the credibility of evidence

In accordance with previous umbrella reviews,49505152 eligible associations from observational studies were classified into five levels according to the strength of the evidence of potential environmental risk or protective factors: convincing (class I), highly suggestive (class II), suggestive (class III), weak (class IV), and not significant. Briefly, credibility of evidence from observational studies is rated on the basis of the number of events developing the outcome of interest, P value of the association, small study effect, excess of significance bias, prediction intervals, statistical significance of the largest study, and heterogeneity. The specific criteria are exhaustively reported in the supplementary methods. We used sensitivity analyses on all levels of evidence, removing the criterion of more than 1000 cases, and on adjusted estimates and cohort studies on class I and II evidence only (supplementary methods).

We classified evidence from meta-analyses of randomised controlled trials, updating a previously proposed framework, classifying certainty of evidence as high, moderate, low, or very low,53 based on GRADE (Grading of Recommendations, Assessment, Development and Evaluations).32 GRADE is a transparent framework that is widely used to develop and present evidence synthesis, providing a set of explicit criteria across different domains to assess level of evidence, and making clinical practice recommendations. As recommended by GRADE, the level of evidence was determined by risk of bias, inconsistency, indirectness, imprecision, and publication bias (supplementary methods).

Patient and public involvement

This study was author funded and we did not involve patients and the public in this work, but we will apply for funding to involve them in the knowledge translation of present findings. Knowledge translation activities will include, but will not be limited to, dissemination of findings via personal and institutional social media, education of health professional trainees, and continuous medical education activities for health professionals. We will involve patient and public representatives in creating a plain language summary of findings to be distributed to the clinical population with mental health disorders, and pregnant women, informing policy makers across different countries with written communications.

Results

Literature search

Starting from 6657 records after duplicate removal, we excluded 5941 studies at title and abstract screening stage, and 599 at full-text level, resulting in 101 publications included. Studies identified by manual search had already been identified from the systematic search. The list of studies excluded after full-text assessment, with reason for exclusion, is reported in supplementary table 3, and the article selection flow is reported in figure 1.33 Of the 101 articles, 50 were meta-analyses of observational studies (215 meta-analytical associations),212254555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101 and 51 were meta-analyses of randomised controlled trials2374102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151 (364 meta-analytical associations) Of note, one meta-analysis reported on both observational and randomised controlled trials (table 1, table 2, supplementary material 2).74

Fig 1
Fig 1

Study selection flow. References of excluded studies after full text assessment available in supplementary table 3. *One meta-analysis included both observational and randomised controlled trials

Table 1

Characteristics of included meta-analyses of observational studies, or non-randomised studies

View this table:
Table 2

Characteristics of included meta-analyses of randomised controlled trials

View this table:

Meta-analyses of randomised controlled trials

The eligible meta-analyses of randomised controlled trials were published between 2008 and 2022. The quality of included meta-analyses according to AMSTAR 2 was high in 20 meta-analyses, moderate in seven, low in 21, and critically low in four (table 2). The median number of studies included in meta-analyses was five (interquartile range 3-9, range 2-42) and the median number of participants was 540 (251-1276, 37-4243).

Cannabidiol was specifically evaluated in seven meta-analyses, while others considered different combinations of cannabis, cannabinoids, tetrahydrocannabinol, and cannabis-based medicines including nabiximols, dronabinol, nabilone, levonantradol, and CT3. Overall, 364 unique meta-analytical associations were identified reporting on acceptability or tolerability of physical adverse events (n=213), psychiatric or psychological related outcomes (n=54), pain related outcomes (n=39), cognitive related (n=20), euphoria or feeling high (n=5), quality of life (n=5), and other various outcomes (n=28). Supplementary table 4 (associations with low or very low certainty) shows the summary effects of the unique meta-analyses or associations for randomised controlled trials.

Summary of associations

Based on the GRADE approach, 14 statistically significant meta-analytical associations (3.8%) met the high certainty criteria, 92 (25.3%) moderate certainty, 200 associations (55.0%) met the low certainty, and 58 associations (16.9%) met the very low certainty. The table detailing the classification of the level of evidence is presented in the supplementary material (supplementary table 4). In the following sections, we principally described the associations with high and moderate GRADE by subgroup of populations.

GRADE of evidence of cannabinoids and outcomes

Mixed chronic pain conditions

Among the 34 associations in this population, cannabis-based medicines or cannabinoids reduced pain by 30% (equivalent odds ratio 0.59 (95% confidence interval 0.37 to 0.93)), but for pain relief no effect emerged (equivalent odds ratio not calculable, mean difference −0.09 (95% confidence interval −0.30 to 0.10)) with high certainty. An additional seven beneficial effects were supported by moderate certainty, including analgesic efficacy (n=5), pain reduction (n=1), and change in pain scores (n=1), yet no effect emerged on patient global impression much or very much improved (n=1), and 50% pain reduction (n=1) (fig 2 supplementary table 4). Two other associations with harmful effects were supported by moderate certainty, including psychological distress (n=1) and withdrawals due to adverse events (n=1). Low (n=17) or very low (n=3) certainty were found for the remaining associations (supplementary table 5).

Fig 2
Fig 2

Moderate and high certainty evidence according to Grading of Recommendations, Assessment, Development and Evaluations (GRADE), from randomised controlled trials on outcomes of cannabis based medications in people with chronic pain, multiple sclerosis, inflammatory bowel disease, and cancer. Only associations for which an eOR was available are displayed. Results are displayed in descending order of level of evidence and effect size. CBM=cannabis based medications eOR=equivalent odds ratio; H=high; M=moderate

Multiple sclerosis and paraplegia

None of the 18 associations in this population was supported by high certainty. Two beneficial effects of cannabis based medicines were supported by moderate certainty, including pain reduction (n=1), and spasticity (subjective; n=1) (fig 2, supplementary table 4). An additional four harmful effects were supported by moderate certainty, including dizziness (n=1), dry mouth (n=1), nausea (n=1) and somnolence (n=1) (fig 2, supplementary table 4). Low (n=10) or very low (n=2) certainty were found for the remaining associations (supplementary table 4).

Inflammatory bowel or Crohn’s disease

Among the three associations in this population one between cannabinoids and better quality of life (fig 2, supplementary table 4) presented high certainty (equivalent odds ratio 0.34 (95% confidence interval 0.22 to 0.53)). Low (n=1) or very low (n=1) certainty were found for the remaining associations (supplementary table 5).

Cancer

None of the 60 associations in this population was supported by high certainty. A beneficial effect emerged on sleep disturbances (n=1), as well as an increased risk of adverse events of gastrointestinal disorders (n=1), nervous system disorders (n=1), serious adverse events (n=1), tolerability (n=1), nausea (n=1), and no effect on daily breakthrough opioid dosage (n=1), constipation (n=1), pain (n=4), risk of psychiatric disorder (n=1), vomiting (n=1), or withdrawal due to adverse events (n=1), with moderate certainty (fig 2, supplementary table 4).

Epilepsy

Among the 46 associations in this population one between cannabidiol and diarrhoea presented high certainty with harmful effects (equivalent odds ratio 2.25 (95% confidence interval 1.33 to 3.81)), and no effect on sleep disruption (equivalent odds ratio not calculable, mean difference −0.29 (95% confidence interval −0.88 to 0.30)). Moderate certainty emerged for seven harmful effects, namely any adverse event (n=2), decreased appetite (n=1), diarrhoea (n=1), serious adverse events (n=1), somnolence (n=1), treatment related adverse events (n=1), as well as for 10 beneficial effects: seizures reduction (n=7), global impression improvement (n=2), and quality of life (n=1). No effect was noted, with moderate certainty, for gastrointestinal side effects (n=1), quality of life in children (n=1), status epilepticus (n=1), upper respiratory infection (n=1), vomiting (n=1), pyrexia (n=1). Low (n=16) or very low (n=3) certainty were reported for the remaining associations (fig 3, supplementary table 4).

Fig 3
Fig 3

Moderate and high certainty evidence according to Grading of Recommendations, Assessment, Development and Evaluations (GRADE), from randomised controlled trials on outcomes of cannabis based medications (CBD) in people with epilepsy. Results are displayed in descending order of level of evidence and effect size; only associations for which an eOR was available are displayed. eOR=equivalent odds ratio; H=high; M=moderate

Mixed conditions

Among the 140 associations in this population three between cannabis-based medicines and various adverse events (supplementary figure 1, supplementary table 4) presented high certainty with harmful effects (equivalent odds ratio 2.84 (95% confidence interval 2.16 to 3.73) for central nervous system adverse events; 3.07 (1.79 to 5.26) for psychological adverse events, and 3.00 (1.79 to 5.03) for vision related adverse events). Moderate certainty supported a beneficial effect on nausea or vomit reduction (n=1), pain reduction (n=3), spasticity reduction (global impression of change) (n=1), an increased risk of feeling high (n=1), gastrointestinal adverse events (n=2), gastrointestinal disorder (non-serious; n=1), emerging psychiatric disorder (n=1), somnolence (n=1), and withdrawal due to adverse events (n=1), while no associations were reported with application site discomfort (n=1), cardiac adverse events (n=1), headache (n=1), musculoskeletal and connective disorder (n=1) and musculoskeletal adverse events (n=1), quality of sleep (n=1), renal urinary disorder (n=1), respiratory disorder (n=1), spasticity reduction (n=1), or weakness (n=1).

Two other beneficial effects of cannabidiol were noted with high certainty, on seizures (equivalent odds ratio 0.59 (95% confidence interval 0.38 to 0.92) for 50% seizure reduction and 0.59 (0.36 to 0.96) for seizure events; supplementary figure 1, supplementary table 4), but moderate evidence supported an increased risk of pneumonia (n=1), somnolence (n=1), gastrointestinal hyperactivity events (n=1), and withdrawal due to adverse events (n=1) (supplementary figure 1, supplementary table 4).

Low (cannabis-based medicines, n=63; cannabis, n=28) or very low (cannabis based medicines, n=9; cannabidiol, n=12) certainty were found for the remaining associations (supplementary table 5).

General population

Among the 23 associations in this population, two between cannabis and emerging psychiatric symptoms presented at high certainty with harmful effects (positive psychotic symptom severity, equivalent odds ratio 5.21 (3.36 to 8.01) and total psychiatric symptoms, equivalent odds ratio 7.49 (5.31 to 10.42)). An additional 12 harmful effects were supported by moderate certainty, including negative symptom severity (n=1), and cognitive outcomes (n=11) (supplementary figure 2, supplementary table 5). Low (n=7) or very low (n=2) certainty were found for the remaining associations (supplementary table 4).

Healthy people

Among the three associations in this population two between cannabinoids and pain outcomes presented high certainty with beneficial effects (equivalent odds ratio 0.74 (95% confidence interval 0.59 to 0.91) for pain threshold and 0.60 (0.41 to 0.88) for pain unpleasantness). Low (n=1) certainty was noted for the remaining association (supplementary table 4).

Mental health disorders, dementia, Alzheimer’s, and Parkinson’s disease

None of the 37 associations in various neuropsychiatric populations (ie, psychiatric disorders, dementia, Alzheimer’s disease, Parkinson’s disease, opioid use disorder, and cannabis use disorder) was supported by either high or moderate certainty. Low (n=26) or very low (n=11) certainty was found for all the associations (supplementary table 4).

Meta-analyses of observational studies

The eligible meta-analyses of observational studies were published between 2002 and 2022. The quality of included meta-analyses according to AMSTAR 2 was high in 13 meta-analyses, moderate in 24, low in 12, and critically low in one (table 1). The median number of individual studies included in the meta-analyses was 6 (interquartile range 4-13, range 2-69), the median number of participants was 1063 (526-4414, 44-5 962 412), and the median number of cases was 814 (447-2078, 126-8060).

The meta-analyses of observational studies reported a wide range of meta-analytical associations between cannabinoids and related health outcomes (supplementary table 5): cognitive, neuropsychological (n=81), brain function, volume (n=38), maternal and neonatal (n=12), psychosis symptoms and relapse (n=15), cancer (n=14), motor vehicle accidents (n=7), suicide (n=6), depression (n=4), behavioural inhibition (n=5), adherence to antipsychotic treatment (n=4), liver fibrosis (n=3), physical dating violence (n=2), and others (n=24). The 215 meta-analytical associations included 878 individual estimates from individual studies: 375 were derived from cohort studies, 493 from case-control studies, and 10 from mixed study designs.

Summary of associations

Of the 215 examined meta-analytical associations, 109 (51%) had a nominally statistically significant effect (P≤0.05) under the random-effects models, but only 14 of those (7%) reached a P value of 10−6 or less. Only 15 meta-analytical associations (7%) had more than 1000 cases and none had more than 20 000 participants for continuous outcomes. Sixty-eight meta-analytical associations (32%) exhibited large heterogeneity (I2 >50%), and only 12 of them (6%) had a 95% prediction interval that excluded the null value. Additionally, small study effects were found for 13 meta-analytical associations (6%) and excess significance bias was found for 15 (7%).

Only two associations (1%) showed a convincing level of evidence (class I), and one (<1%) showed highly suggestive evidence (class II). Of the remaining associations, four (2%) showed suggestive evidence (class III), 102 (47%) weak evidence (class IV), and 106 (49%) had no evidence (not significant). The table detailing the classification of the level of evidence is presented in the supplementary material (supplementary table 5). In the following sections, we principally described the associations with the highest classes (I-convincing, II-highly suggestive, III-suggestive) of the evidence in the main and general sensitivity analysis by subgroup of populations.

Credibility of evidence of associations between cannabinoids and outcomes

Pregnant women

Among the 19 associations in this population only two outcomes (fig 4, supplementary table 5) presented convincing evidence with harmful effects of cannabinoids (marijuana use and low birth weight, equivalent odds ratio 1.43 (95% confidence interval 1.27 to 1.62)) and marijuana and small for gestational age (1.61 (1.41 to 1.83); both unadjusted estimates). Class III evidence emerged for two other associations with harmful effects: one between marijuana use and preterm delivery (1.32 (1.14 to 1.54)) and one between marijuana and neonatal intensive care unit admission (1.41 (1.15 to 1.71); both unadjusted estimates). After removing the criterion of number of cases of more than 1000 in the sensitivity analysis, no change was reported in the level of class I and III evidence, however, one additional association was upgraded from weak (class IV) to suggestive evidence (class III; mean birth weight, unadjusted; supplementary table 5). No evidence was found for the remaining associations (supplementary table 5).

Fig 4
Fig 4

Observational meta-analytical associations between cannabis and outcomes in pregnant women, drivers, and people with psychosis supported by convincing, highly suggestive, or suggestive evidence in main or sensitivity analysis. Results are displayed in descending order of level of evidence and effect size; only associations for which an eOR was available are displayed. n=cases; N=population; CE=class of evidence (convincing (I), highly suggestive (II), suggestive (III), weak (IV)); CES=class of evidence after removing the n>1000 cases criterion; eOR=equivalent odds ratio; NR=not reported

The association between marijuana use and low birth weight was downgraded to no evidence using only adjusted estimates or cohort studies. The association between marijuana and small for gestational age remained at the same level (ie. convincing) using only cohort studies (adjusted sensitivity analysis not possible). The association with preterm delivery remained suggestive in analyses of only cohort studies, but the level was downgraded to no evidence with use of only adjusted estimates (supplementary table 5).

Drivers

None of the seven associations in this population was supported by convincing or highly suggestive evidence (class I and II) (fig 4, supplementary table 5). Evidence was weak (class IV) for the seven associations between cannabis use and driving outcomes with harmful effects (supplementary table 4). In the sensitivity analysis, after removing the criterion of number of studies as more than 1000, two associations were upgraded from weak (class IV) to convincing evidence (class I) for tetrahydrocannabinol and harmful effects of car crash and culpability (adjusted estimates). Two other associations between cannabis use and car death after car crash (unadjusted) and unfavourable traffic events related to cars (unadjusted) were upgraded from weak (class IV) to highly suggestive evidence (class II). Two additional associations between cannabis use and car collision and car injury (both unadjusted estimates) were upgraded from weak (class IV) to suggestive evidence (class III) (fig 4, supplementary table 5).

Psychosis

None of the 50 associations in this population was supported by convincing or highly suggestive evidence (class I and II) (fig 4, supplementary table 5). Weak evidence (class IV) was available for 13 associations with harmful effects, whereas no evidence was found for the remaining associations (supplementary table 5). After removing the criterion of more than 1000 cases in the sensitivity analysis, five associations of cannabinoids with harmful effects (ie, working memory, psychosis relapse, premorbid IQ (unadjusted), poor adherence to antipsychotics (two associations adjusted)) were upgraded from weak (class IV) to suggestive evidence (class III).

General population

Among the 119 associations in this population, only one between cannabis and psychosis (fig 5, supplementary table 5) presented highly suggestive evidence with harmful effects of cannabinoids in adolescents (equivalent odds ratio 1.71 (95% confidence interval 1.47 to 2.00); no information on adjustments). Evidence was suggestive (class III) for two other associations with harmful effects: one between heavy use of cannabis and suicide attempt (3.20 (1.72 to 5.94)) and one between most frequent use of cannabis and psychotic symptoms (2.18 (1.45 to 3.27); both adjusted estimates). Weak or no evidence were found for the remaining associations (supplementary table 5). After removing the criterion of more than 1000 cases in the sensitivity analysis, the level of class II and III evidence did not change. However, one additional association with harmful effects between tetrahydrocannabinol and increased heart rate (unadjusted) was upgraded from weak (class IV) to highly suggestive evidence (class II). Additionally, eight associations with harmful effects were upgraded from weak (class IV) to suggestive evidence (class III) including mania symptoms (adjusted), depression (adjusted), testicular cancer (three associations), orbitofrontal cortex volume (medial, total), and physical dating violence (supplementary table 5). The association with cannabis and psychosis also remained highly suggestive (table 2), but the level of evidence was upgraded to convincing when only cohort studies were included (adjusted sensitivity analysis not possible).

Fig 5
Fig 5

Observational meta-analytical associations between cannabis and outcomes in the general population and healthy people supported by convincing, highly suggestive, or suggestive evidence in main or sensitivity analysis excluding 1000 cases criterion. Results are displayed in descending order of level of evidence and effect size; only associations for which an eOR was available are displayed. n=cases; N=population; CE=class of evidence (convincing (I), highly suggestive (II), suggestive (III), weak (IV)); CES=class of evidence after removing the n>1000 cases criterion; eOR=equivalent odds ratio; NR=not reported

Healthy people who use cannabis

None of the eight associations in healthy people who use cannabis was supported by convincing or highly suggestive evidence (class I and II) (fig 5, supplementary table 5). Only weak evidence (class IV) was noted for eight associations between cannabis use and cognitive outcomes with harmful effects (supplementary table 5). After removing the criterion of more than 1000 cases in the sensitivity analysis, four associations with harmful effects (ie, visual immediate recall, prospective memory, verbal learning, and verbal delayed recall) were upgraded from weak (class IV) to highly suggestive evidence (class II). Additionally, three associations (ie, verbal immediate recall, verbal recognition, and working memory) were upgraded from weak (class IV) to suggestive evidence (class III) (fig 5, supplementary table 5).

Other populations

Across people with cannabis use disorder, insomnia, chronic pain, mixed conditions, hepatitis C virus or non-alcoholic fatty liver disease, and central nervous system malignant disease, none of the 12 associations was supported by convincing or highly suggestive evidence (class I and II, supplementary table 5). Weak evidence (class IV) was noted for five associations between cannabis use with harmful effects, whereas no evidence was found for the remaining associations (supplementary table 5). After removing criterion of more than 1000 cases in the sensitivity analysis, three associations of cannabinoids with beneficial effects (namely sleep quality or quantity improvement, pain relief, and hepatic steatosis (all unadjusted)) were upgraded from weak (class IV) to suggestive evidence (class III).

Other details of cannabis use, adjustment of analyses, and quality of individual studies

Details on type of cannabis, route of administration, use, variables that analyses were adjusted for, and quality or risk of bias of individual studies included in eligible meta-analyses are reported in supplementary material 3.

Of the 512 individual studies included in the eligible meta-analyses, 325 were observational studies and 187 were randomised controlled trials. Among the 325 observational studies (cohort n=160, cross-sectional n=97, case-control n=68), 211 reported on cannabis, 108 on marijuana, two on dronabinol, two on nabilone, one on cannabidiol, and one on tetrahydrocannabinol and cannabidiol. Of these, 312 focused on recreational use of cannabinoids, 12 on medical use, and one on both; 292 studies did not report the route of administration, which was inhaled in 28 studies and oral in five studies. Overall analyses were unadjusted in 79 studies and adjusted or matched in the remaining studies. The median Newcastle-Ottawa score of case-control and cohort studies was 7 (interquartile range 7-9).

Among the 187 randomised controlled trials, 64 reported on tetrahydrocannabinol, 32 on nabilone, 26 on nabiximols, 22 on cannabis, 18 on cannabidiol, and the remaining on various combinations of cannabis-based medicines, or other individual cannabis based medicines. Of these, 186 focused on medical use of cannabinoids, and one on recreational use; the route of administration was oral in 121, oral spray in 29, inhaled in 21, intravenous in six, intramuscular in four, oral and inhaled in three, and transdermal in two studies. The risk of bias was high in 79 randomised controlled trials, unclear in 55, low in 48, and moderate in five.

Discussion

Principal findings

This umbrella review grades the credibility and certainty of evidence on the effect of cannabinoid use, encompassing observational and interventional evidence.

Regarding harmful outcomes, among all meta-analytical associations supported by at least suggestive evidence in observational studies and moderate certainty in randomised controlled trials, converging evidence supports an increased risk of psychosis associated with cannabinoids in the general population. Specifically, cannabis use was associated with psychosis in adolescents (highly suggestive credibility, convincing certainty in main sensitivity analyses) and adults (suggestive credibility, suggestive certainty), and with psychosis relapse in people with a psychotic disorder (weak credibility, suggestive certainty). Use of cannabinoids in adult non-clinical and clinical populations was associated with positive (high certainty) and negative (moderate certainty) psychotic symptoms in randomised controlled trials.

Evidence from observational studies (weak credibility, suggestive certainty) and randomised controlled trials (high credibility, moderate certainty) show an association between cannabis and general psychiatric symptoms, including depression and mania, as well as detrimental effects on prospective memory, verbal delayed recall, verbal learning, and visual immediate recall (weak credibility, highly suggestive in observational evidence, moderate certainty in randomised controlled trials). Across different clinical and non-clinical populations, observational evidence suggests an association between cannabis use and motor vehicle accidents (weak credibility, convincing certainty). Additionally, evidence from randomised controlled trials shows an association with somnolence (cannabinoids (moderate certainty) and cannabidiol (high certainty)),103 and cannabis based medicines and visual impairment (high certainty), disorientation, dizziness, sedation, and vertigo (moderate certainty), among others.

These associations are of particular concern given the epidemiology and age pattern of cannabis use disorders, and the population attributable fraction of cannabis for schizophrenia, which is almost 10%.152 According to the Global Burden of Disease 2019, cannabis use disorders are associated with 690 000 (95% uncertainty interval 421 000-1 080 000) disability adjusted life years per 100 000 individuals globally.9 Prevalence and disability related to cannabis start to be measurable at ages 10-14 years (11 900 disability adjusted life years), peak at ages 20-24 years (163 000 disability adjusted life years), then gradually decrease.912153 The age pattern of cannabis use disorders coincide with the peak age at onset of mental health disorders. According to the largest meta-analysis on the age at onset of mental disorders published to date, which pooled 192 studies and 708 561 individuals, around 34.6% of mental health disorders have onset by age 14 years, 48.4% by 18 years, and 62.5% by 25 years; the age that any mental health disorder onset peaks is at 14.5 years.154 For cannabis use disorders, 66% of people will have onset by age 25 years, with age of peak onset 20.5 years. Of note, age at peak onset of schizophrenia spectrum disorders is also in the early 20s, with a slightly lower proportion of people with onset by 25 years (47.8%). In addition to the association between cannabis and psychosis, cannabis is also associated with a worse outcome after onset, including poorer cognition,87 lower adherence to antipsychotics,56 and higher risk of relapse.85 In other words, use of cannabis when no psychotic disorder has already occurred increases the risk of its onset, and using cannabis after its onset, worsens clinical outcomes. Mood disorders also have their peak of onset close to that for cannabis use, which is of concern given the associations shown in this work between cannabis and depression, mania, and suicide attempt. Moreover, high tetrahydrocannabinol content cannabis could serve as a so-called gateway to other substances, particularly in younger people: this effect has been shown in humans155 and animal models,156157 strengthening the recommendation to avoid cannabis use in adolescents and young adulthood.

Evidence suggests detrimental effects on cognition, an association with motor vehicle accidents, together with the age pattern of cannabis use (disorder), and related burden, which raise two additional matters. Firstly, given the adverse effects of cannabis on verbal delayed recall, verbal learning, visual immediate recall, and mental health, negative effects on scholastic or academic performance are reasonably expected, particularly in people who heavily use. Secondly, psychiatric symptoms such as suicide ideation and attempt, mania, and poor cognition, among other adverse events (eg, somnolence, disorientation, dizziness, sedation, vertigo, and visual impairment) might mediate the association between cannabis and increased risk of motor vehicle accidents. According to the DRUID project (driving under the influence of drugs, alcohol, and medicines in Europe), tetrahydrocannabinol ((0.5-2.2), measured as tetrahydrocannabinol or carboxy-tetrahydrocannabinol, in oral fluid or blood) is the second most frequent compound detected in seriously injured drivers, after alcohol (14.1-30.2%), then cocaine and amphetamines.158

Numerous observational associations indicated harmful outcomes, but they were either isolated without converging evidence from different study designs, supported by weak evidence only, or downgraded to not significant. Downgrading applied to the association between cannabis and low birth weight, and preterm delivery,100 which might be mediated by smoking.

Regarding the therapeutic potential of cannabis-based medicines, cannabidiol was beneficial in reducing seizures in certain forms of epilepsy in children and adults, including Lennox-Gastaut syndrome, Dravet syndrome, or other types of epilepsy. Cannabis based medicines were beneficial for pain and spasticity in multiple sclerosis, as well as for chronic pain in various conditions, and in palliative care, yet not without adverse events. However, cannabidiol and other cannabis-based medicines were associated with lower acceptability and tolerability than placebo in children and adults, and cannabis based medicines were also associated with psychiatric adverse events, as stated previously. These findings must be put into a clinical perspective to be fully appreciated and compared with available alternatives. Regarding epilepsy, established anticonvulsants are not free from adverse events, including sedation, weight gain, cognitive impairment, and psychiatric symptoms.159160161 Regarding chronic pain, excessive use of prescribed opioid medications has contributed to the opioid crisis, indicating the need for novel pharmacological and non-pharmacological treatment options for chronic pain162 to reduce prescribed opioid medications abuse. Regarding multiple sclerosis, botulinum toxin seems to be the only pharmacological alternative to cannabis based medicines for spasticity.110163 Finally, the clinical populations included in eligible meta-analyses had treatment resistant or chronic conditions or were being treated in the context of palliative care and ongoing chemotherapy, and other treatment options had not proven effective. Thus, cannabis-based medicines could be reasonable options for chronic pain in different conditions, muscle spasticity in multiple sclerosis, and for nausea and vomiting in mixed clinical populations, and for sleep in people with cancer. Importantly, in patients with chronic pain, evaluation of the clinical effects considering the whole clinical presentation (several of the included reviews question the clinical value), the effects of prolonged use of cannabinoids still needs to be tested because current findings only come from short term randomised controlled trials. Also, active comparisons between cannabidiol and available options for epilepsy, as well as between cannabis-based medicines and other pain medications, other treatments for muscle spasticity in multiple sclerosis, or treatments for sleep in persons with cancer are needed, with a focus on both efficacy and safety, to inform future guidelines.

Overall, a mismatch is manifest between the legislation ruling cannabinoids versus alcohol use, considering both the well-known harms of alcohol on physical and mental health, in any age group,164 and the epidemiological figures. According to Global Burden of Disease 2019, alcohol use disorders were associated with 17 000 000 (95% uncertainty interval 13 500 000-21 500 000) disability adjusted life years per 100 000 individuals,9 roughly 25 times higher than for cannabis. Also, disability related to cannabis was largely limited to individuals aged 10-24 years, whereas alcohol is associated with disability from early stages of life, increasing continuously to 2 120 000 disability adjusted life years at age 35-39 years, and very slowly decreasing to less than 200 000 disability adjusted life years only after age 80 years.9 If cannabis use prevalence increased in the younger portion of the population due to large scale legalisation, whether the gap described previously would diminish is unclear. Moreover, to the best of our knowledge alcohol has no role as a medical treatment, whereas our research shows that cannabinoids can have beneficial effects in specific clinical conditions. The (scientific) reasoning behind extreme or ideological legislative approaches, namely complete legalisation and commercialization of cannabis even in young adults versus complete prohibition, and the different legislative requirements between cannabis and alcohol in disclosing to consumers the associated risks remains unclear.9

Strengths and limitations

The main strength of this work is that we pool evidence from different sources of evidence and deliberately consider convergence of results from different study designs. Also, this umbrella review is the first to pool observational and interventional studies on the effects of cannabinoids on humans.

Our results should be interpreted with caution. Firstly, the evidence from observational studies has ecological validity with regards to the type of cannabis available in the legal or illegal market on a large scale. However, tetrahydrocannabinol content and other cannabinoids on which no meta-analytical evidence was included can vary considerably among legal and illegally sold products. At the individual level, this variation can mean the difference between harmful or neutral or beneficial effects. Moreover, evidence from more than a decade ago, might not be representative of the cannabis that can be purchased nowadays illegally and legally, which is rich in tetrahydrocannabinol. This means that findings of this work might be underestimating harmful effects of cannabis. Also, the clinical effect of tetrahydrocannabinol on GABA and glutamate signalling via partial agonism on CB1 receptors depends on the concentration and distribution of CB1 receptors in the brain of each individual.10 As such, not all individuals will experience the same effects of cannabis on their mental health and cognition. Nonetheless, a crossover trial of 64 volunteers found that short term detrimental effects of 10 mg of inhaled tetrahydrocannabinol on psychological measures and cognition was not influenced by the co-administration of up to 30 mg of cannabidiol, potentially mitigating potential concerns with a role of tetrahydrocannabinol or cannabidiol ratio as a confounder of findings of this work.165 However, this trial was limited by a very short follow-up (90 min) and high loss to follow-up (28%). Furthermore, cannabidiol products that contain either no tetrahydrocannabinol, or subclinical amounts, are unlikely to result in psychological or cognitive impairment. Similarly, cannabis use disorder seems to have similar rates of people who use recreationally versus medically.166 Secondly, another reason to be cautious is that umbrella reviews neglect evidence from individual cohort studies or randomised controlled trials that have not been previously pooled in meta-analyses. However, individual studies need replication, are frequently exploratory, and need to be pooled in systematic reviews (and ideally meta-analyses) so that a comprehensive understanding of a given association or intervention can be appraised. Hence, any evidence that was not included in this umbrella review, even if potentially relevant, could be exploratory or preliminary. Thirdly, confounding factors could drive associations in observational evidence. However, we have applied stringent criteria, as confirmed by downgrading convincing evidence to non-significant on the association between cannabis and pregnancy outcomes. The quantitative criteria we applied to grade evidence from observational evidence accounted for selection and publication bias, excess of significance driven by small studies with larger effect sizes than the largest study in the meta-analysis, or marginal statistical significance driven by large sample sizes. Also, we have discussed findings from observational evidence in the context of converging evidence from different sources of evidence. For instance, observational studies might be affected by confounding factors, whereas randomised controlled trials might not be representative of the real-world population and affected by selection bias instead. We believe that converging evidence from these two study designs strengthens the ecological and methodological credibility of our findings. Fourthly, excess of significance bias testing might have been underpowered in meta-analyses with few studies, which could arguably apply to all meta-analyses included in this umbrella review, yet, a specific threshold of number of studies to set adequate power of excess of significance bias has not been established. Fifthly, we could have included meta-analyses based on their quality instead of the number of studies. However, that would have introduced a selection bias, leaving out a large portion of evidence. Sixthly, to harmonise effect sizes, we calculated the equivalent odds ratio as a measure of strength of the association. Yet, the harmonisation comes at the cost of losing information on time-to-event analyses and of course any association should be considered more in depth considering the frequency of each outcome, and the follow-up duration and time to event occurrence in each of the included studies. Additionally, the number of cases over the overall population of included studies does not reflect the prevalence of outcomes of interest. For instance, the global prevalence of preterm birth is 11.3%167 versus 9.9% reported in this work and varies across regions and countries’ income levels. The global prevalence of small for gestational age is 27%167 versus 9.1% reported here, with large variations across regions. The prevalence of testicular cancer is 0.04% versus 29.5% reported in the included meta-analysis of case-control studies. Lastly, and most importantly, the results of this work aim to inform future guidelines. These guidelines should account for additional aspects such as cost-effectiveness considerations, clinical relevancies (eg, numbers needed to treat for benefit), long term effects of cannabinoids on which evidence is lacking, and stakeholders, including patients and family members perspectives.

Future research assessing use of cannabis should clearly report what type of cannabis that patients used, how cannabis was administered, the content of tetrahydrocannabinol and cannabidiol, and the amount of cannabis consumed. The dose of exposure to different cannabinoids is needed to infer any causal relation between cannabis and outcomes.

Conclusions

Convincing or converging evidence supports that cannabis use is associated with poor mental health and cognition, increased the risk of car crashes, and can have detrimental effects on offspring if used during pregnancy. Cannabis use should be avoided in adolescents and young adults (when neurodevelopment is still occurring), when most mental health disorders have onset and cognition is paramount for optimising academic performance and learning, as well as in pregnant women and drivers. Conversely, cannabidiol could be considered a potential beneficial treatment option in epilepsy across age groups to reduce seizures. Cannabis based medicines could also be considered for chronic pain across different conditions, such as multiple sclerosis, spasticity in multiple sclerosis, for nausea and vomiting in people with mixed conditions and for sleep in cancer. However, clinical relevance must be considered before a possible incorporation into clinical guidelines; for example, including numbers needed to treat for benefit, risk to benefit ratios, comparative efficacy and safety with existing treatment options, and development of patient information concerning potential adverse events. Cannabidiol appears to be safe regarding psychiatric symptoms, but more research needs to be conducted before this drug can be recommended for the treatment of any psychiatric disorder. The remaining associations between cannabis and health outcomes are not supported by converging or convincing evidence.

Law and public health policy makers and researchers should consider this evidence synthesis when making policy decisions on cannabinoids use regulation, and when planning a future epidemiological or experimental research agenda, with particular attention to the tetrahydrocannabinol content of cannabinoids. Future guidelines are needed to translate current findings into clinical practice, while involving stakeholders.

What is already known on this topic

  • Observational evidence reported that cannabinoids were associated with numerous health outcomes and have been tested for several conditions in randomised controlled trials

  • Credibility and coherence of findings from different sources of evidence on the same outcomes have not been assessed to date

What this study adds

  • Most outcomes associated with cannabinoids are supported by weak evidence (observational studies), low to very low certainty (randomised controlled trials), or are not significant (observational studies, randomised controlled trials)

  • Convincing or converging evidence recommends avoiding cannabis during adolescence and early adulthood in people prone to have or have mental health disorder, who are pregnant, and while driving

  • Cannabidiol is effective for epilepsy, notably in children, while other cannabinoids can be effective in use for multiple sclerosis, chronic pain, inflammatory bowel disease, and palliative care

Ethics statements

Ethical approval

Ethical approval for this study was not required.

Data availability statement

The whole dataset is available from authors on request.

Footnotes

  • Contributors: MS, MDT, and JYK are joint first authors. JIS and MS contributed equally as corresponding authors. ED, MS, and JIS designed and supervised the study. FB, LB, MJC, GC, MDT, ED, JYK, SM, AM, FM, and MS screened the literature and extracted the data. MDT, ED, and MS drafted the manuscript. ED conducted the analyses, and MS had full access to data. All authors approved the project design, and critically reviewed, contributed to the final version of this work, and approve it. ED, MS, and MDT are guarantors of this work. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

  • Funding: None.

  • Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/disclosure-of-interest/ and declare: MS received honoraria/has been a consultant for AbbVie, Angelini, Lundbeck, Otsuka. DC has received grant monies for research from Eli Lilly, Janssen Cilag, Roche, Allergen, Bristol-Myers Squibb, Pfizer, Lundbeck, Astra Zeneca, Hospira; Travel Support and Honoraria for Talks and Consultancy from Eli Lilly, Bristol-Myers Squibb, Astra Zeneca, Lundbeck, Janssen Cilag, Pfizer, Organon, Sanofi-Aventis, Wyeth, Hospira, Servier, Seqirus; and is a current or past Advisory Board Member for Lu AA21004: Lundbeck; Varenicline: Pfizer; Asenapine: Lundbeck; Aripiprazole LAI: Lundbeck; Lisdexamfetamine: Shire; Lurasidone: Servier; Brexpiprazole: Lundbeck; Treatment Resistant Depression: LivaNova; Cariprazine: Seqirus. He is founder of the Optimal Health Program, currently operating as Optimal Health Australia; and is part owner of Clarity Healthcare. He is on the scientific advisory of The Mental Health Foundation of Australia. He does not knowingly have stocks or shares in any pharmaceutical company. EV has received grants and served as consultant, advisor or CME speaker for the following entities: AB-Biotics, AbbVie, Angelini, Biogen, Boehringer-Ingelheim, Celon Pharma, Dainippon Sumitomo Pharma, Ferrer, Gedeon Richter, GH Research, Glaxo-Smith Kline, Janssen, Lundbeck, Novartis, Orion Corporation, Organon, Otsuka, Sage, Sanofi-Aventis, Sunovion, Takeda, and Viatris, outside of the submitted work. CUC has been a consultant or advisor to or have received honoraria from: AbbVie, Acadia, Alkermes, Allergan, Angelini, Aristo, Boehringer-Ingelheim, Cardio Diagnostics, Cerevel, CNX Therapeutics, Compass Pathways, Darnitsa, Gedeon Richter, Hikma, Holmusk, IntraCellular Therapies, Janssen/Johnson & Johnson, Karuna, LB Pharma, Lundbeck, MedAvante-ProPhase, MedInCell, Merck, Mindpax, Mitsubishi Tanabe Pharma, Mylan, Neurocrine, Newron, Noven, Otsuka, Pharmabrain, PPD Biotech, Recordati, Relmada, Reviva, Rovi, Seqirus, SK Life Science, Sunovion, Sun Pharma, Supernus, Takeda, Teva, and Viatris. He provided expert testimony for Janssen and Otsuka. He served on a Data Safety Monitoring Board for Lundbeck, Relmada, Reviva, Rovi, Supernus, and Teva. He has received grant support from Janssen and Takeda. He received royalties from UpToDate and is also a stock option holder of Cardio Diagnostics, Mindpax, and LB Pharma.

  • All authors declare manuscript is an honest, accurate, and transparent account of the study being reported; no important aspects of the study have been omitted; and that any discrepancies from the study as originally planned (and, if relevant, registered) have been explained in published protocol.

  • Dissemination to participants and related patient and public communities: Our dissemination of findings will be shared via personal and institutional social media, the education of health professional trainees, continuous medical education activities for health professionals, and co-produced plain language summary of findings to be distributed to the clinical population with mental health disorders and to pregnant women. We will inform policy makers across different countries with written communications.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

http://creativecommons.org/licenses/by-nc/4.0/

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

References