Introduction

The emergence of Cannabis sativa as a therapeutic can be dated back to 2700 BC, with use becoming widely adopted in the USA by the 19th century.1 2 Since then, both recreational and medicinal cannabis have undergone a series of proscription and later decriminalisation processes globally. As of the 21st century, cannabis for medicinal purposes has been legalised in many countries, including the USA, the UK, Australia, Canada, Israel and the Netherlands.3

The Cannabis plant contains over 500 different compounds. Of these, 113 are recognised as cannabinoids, where they function as cannabinoid receptors for biological effect.4 Notably, cannabidiol (CBD) and delta-9-tetrahydrocannabinol (THC) are the two main active constituents of over 100 medicinal cannabis products available worldwide. Cannabis-based medicine (CBM) is the products containing cannabinoids that are used for a clear therapeutic purpose, rather than recreational purposes. CBM may be obtained on prescription or otherwise and is used for symptomatic control of intractable chronic diseases.2 These include, but are not restricted to, spasticity in multiple sclerosis (MS), epilepsy, neuropathic pain, cancer-related pain, as well as chemotherapy-induced nausea and vomiting.2 Emerging evidence is expanding the therapeutic usage of CBM to include psychiatric disorders such as anxiety and post-traumatic stress disorder, sleep disorders, fibromyalgia and Parkinson’s disease.2 5–7

The growing evidence base and media attention have triggered a shift in public paradigms towards acceptance of cannabis as a medicine.8 The increasing community demand for CBMs is apparent in the uptrend of prescription approvals. Over the last 7 years in Australia, there were 949,732 patients who were newly prescribed a specific medicinal cannabis product, biannually via the TGA’s Authorised Prescriber System.9 This uptrend in prescribing rates is further reflected by a percentage increase of 402% in new prescriptions in the 6-month period ending January 2022, compared with the 6-month period ending January 2023 (online supplemental material 1).10

However, unlike conventional medications, public demand rather than preclinical studies for quality control, have driven increasing clinical uptake.11 12 Given a history of legislative resistance and restrictions in conducting clinical trials with CBMs, gaps remain in the literature surrounding side effects and adverse reactions. Notably, there is limited safety evidence on CBMs for vulnerable populations commonly excluded from clinical trials, such as pregnant women, children and patients with complex comorbidities.13–15 Additionally, the illicit drug market, over-the-counter availability and unregulated product commercialisation have created a landscape of products that vary in formulation, strength, route of administration and quality.16–18 As such, growing use necessitates prescriber and consumer vigilance on side effects and adverse reactions.

The gap between available safety evidence and clinical use warrants rigorous surveillance for postmarketing signal detection of adverse events associated with CBMs. This need has been addressed in various ways by different countries. To our knowledge, no other research has comprehensively described and evaluated the postmarketing surveillance systems which have been established to monitor the adverse effects of CBM. Therefore, the objective of this research is to provide an overview of current methods for real-world monitoring of the side effects and adverse events associated with CBM use. Using a systematic search of peer-reviewed databases and grey literature, this review aims to answer the following question: What are the systems in place internationally to monitor side effects and adverse events of cannabis use as a medicine?

MethodsSearch strategy

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) reporting guidelines were used as a methodological framework to inform the approach to a systematic search of literature.19 20 Five main keyword elements were identified using the Population, Intervention, Comparison, Outcome (PICO) framework, and subsequently used to guide the development of search terms and inclusion/exclusion criteria (online supplemental material 2).

Two separate categories of searches were conducted, each with search terms from a different combination of keyword elements (online supplemental material 3). For the first category, the search query combined terms relating to elements of medical usage, cannabis, monitoring systems and side effects or adverse events. The second category included cannabis-related terms, as well as terms relating to pharmacovigilance, monitoring systems and medical usage. The first category aimed to increase the specificity of our search, whereas the second category focused on search sensitivity, incorporating the more loosely defined concept of pharmacovigilance, without specific mention of side effects and adverse events. Both categories were used to create searches on 23 June 2023, identical across Scopus and Ovid Embase for peer-reviewed publications, with added MESH terms in the latter (online supplemental material 4).

Category 2 search terms were further used for a grey literature search to supplement our literature database search and maximise the scope of our results. The grey literature search composed of extracting the first 1000 titles of a Google Scholar search using Category 2 search terms. An identical search for grey literature was applied to Mednar, a medically focused search engine, to include deep web searches that were not indexed by standard search engines.21–23 Search terms across different keyword elements were combined with Boolean operator “AND” while terms within a keyword element were combined with the Boolean operator “OR”.

Supplementary articles were identified in the references of retrieved papers. All searches were limited to papers published between January 2015 and June 2023; the period in which cannabis legalisation occurred in multiple countries worldwide, triggering the need for widespread monitoring systems.24–26 The search was manually filtered to papers published in the English language, to yield the results shown in the PRISMA flow chart (online supplemental material 5).

All results from search queries were uploaded into an Excel spreadsheet for duplicate removal. Title and abstract screening were performed, subject to inclusion and exclusion criteria. Records were excluded primarily based on relevance and the a priori decision to exclude records with no mention of pharmacovigilance nor a monitoring system in the title and abstract. Secondary Google searches were performed for primary sources such as reporting forms, where more specific information on databases was required. Small scale surveys were not considered a formal monitoring system and subsequently excluded. All full-text articles identified for inclusion following the screening process were evaluated independently by two reviewers. RQW reviewed all titles, and YAB provided a second review. There was discrepancy between the reviewer assessments in <5% of articles, which were subsequently resolved among the reviewers, thus not requiring a third reviewer. Any points of contention were discussed in meetings and subsequently resolved for a full list of titles for data extraction.

The PRISMA flow chart (online supplemental material 5) outlines the full search strategy and results.

Patient and public involvement

Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this review.

Results

Of the 3939 records identified in the initial peer-reviewed database search, 1004 duplicates were removed, with an additional 2889 records excluded following title and abstract screening. The screening process yielded 46 potentially relevant full-texts from the peer-reviewed database search, with 3 unable to retrieved, leaving 43 full texts for inclusion. The grey literature search identified 1127 records. Following duplicate removal and title and abstract screening, 12 full texts were identified. Of these, one record was unable to be retrieved. Subsequently, 11 additional records from the grey literature search were identified for full text review, leaving 54 records for full text review. Of the 54 full texts assessed for eligibility, 5 were excluded on the basis of relevance and 2 were excluded for recreational cannabis use as the study population. Seven additional papers were identified through citations, yielding a total of 54 included records (online supplemental material 6).

Monitoring systems identified by our search were either registries or databases. Although used interchangeably, there are several distinguishing characteristics between the two (online supplemental material 7). There were a total of 7 regulatory authority databases and 17 registries captured within our search (table 1). Of these registries, eight were smaller registries briefly mentioned in articles, without readily available data and were not analysed in depth (table 1).

Table 1

Monitoring systems captured by search with readily available data for analysis*

Monitoring system characteristicsPrimary purpose

There were two distinct purposes for establishment of monitoring systems. Some systems were created as broader forms of postmarketing surveillance to inform safety and regulation. These include all aforementioned databases and all registries with the exception of five. These registries were created as data collection for observational studies, with postmarketing surveillance as a secondary aim of the research. As such, they are categorised as research registries (table 2).

Table 2

Primary purposes of monitoring systems for side effects and adverse events associated with CBM usage

Duration of data collection

Of the sixteen monitoring systems that are included within this review, fifteen are still actively engaging in data collection at the time of review. All systems with postmarketing surveillance as a primary outcome, provide ongoing data collection at the time of our literature search.

Research registries, due to the longitudinal nature of observational studies, are also ongoing forms of monitoring, with two exceptions (table 3). The Quebec Cannabis Registry, established in 2015, ceased data collection in 2018. Serious adverse events were reported to the Canada Vigilance Database for evaluation.27–30 The Registry collected data in the UK from 2012 to 2015, however, data collection in Germany and Switzerland remains ongoing.31–33

Table 3

Elements assessed by each monitoring system for side effects and adverse events associated with CBM usage

Level of detail assessed

Most systems collect information on formulation and dosage of the CBM, however, this is more common in regulatory databases. Route of administration and concomitant medications are frequently accounted for. Differences in details exist in assessing patient demographics. Age and sex of consumers are commonly collected; however, comorbidities and pregnancy status are not routinely reported (table 3).

Mode of monitoring: spontaneous or mandatory?

Where data sources elect to participate in data collection, the system is considered to adopt a spontaneous reporting protocol. Therefore, all research registries, as well as the ToxiC registry, are spontaneous reporting systems (table 3). The Italian Medicines Agency (AIFA) registry mandates reporting from patients. The German Pain E-registry collects information from 200 pain centres across Germany and fulfils the obligatory requirements of physicians to document patients under treatment for chronic pain.31 34–37

Reporter nature

Monitoring systems collect data from combinations of the following subgroups of individuals: patients, healthcare professionals and/or manufacturers. Three registries (table 4) use patient-reported outcomes. Four registries accept reports from healthcare professionals alone. The ToxiC registry,38–41 collecting data from medical records in participating hospitals, is included in this category. Larger databases encourage reports from patients, healthcare professionals, as well as manufacturers and producers of CBMs.

Table 4

Nature of reporters to each monitoring system for the side effects and adverse events associated with CBM usage

Specificity to cannabis versus other pharmaceuticals

Six of the sixteen monitoring systems captured by the search are specific to CBM monitoring. Databases offer assessments of adverse events associated with regulated and/or unregulated products within a region, rather than specific CBM monitoring.

Affiliations

Four registries are affiliated with independent ownership (table 5).

Table 5

Features of monitoring systems for monitoring of side effect and adverse events associated with CBM usage*

Causality assessment

The strength of the causal relationship between CBM usage and the observed adverse event is considered in five monitoring systems (table 5). The Quebec Cannabis Registry conducts causality assessments on reports, however, the mode of assessment is not described in literature captured by our search.28–30 42 43 The Australian Emyria database does not implement a formal causality assessment, however, possible causal relationships are guided by clinicians’ medical judgement.44

Discussion

As of June 2023, there remain no robust and rigorous monitoring systems globally for collection of postmarketing safety data to accompany the international expansion in CBM uptake. The existence of several regulatory databases and multiple smaller registries globally, some of which have limited published data, demonstrates heterogeneity in postmarketing surveillance of CBM. The literature included in this review did not identify a quality assessment process for the data collected within the fifteen monitoring systems that remain active at the time of this review.

In some countries, the monitoring of CBM-related adverse events is embedded within the national regulatory framework for pharmaceuticals. The Italian Phytovigilance database, coordinated by the Italian National Institute of Health, collects reports on suspected adverse events associated with plant ingredient preparations and food supplements in Italy.45 Similarly, Canada Vigilance Adverse Reaction Online Database (CVAR) evaluates reports of suspected adverse reactions related to heath products with marketing authority within Canada.42 FDA Adverse Event Reporting System (FAERS) in the USA and Database of Adverse Event Notifications (DAEN) in Australia follow similar frameworks.46 47 Eudravigilance European Database of Suspected Adverse Drug Reaction Reports (EDSADR), collecting data on suspected adverse reactions to authorised medicines or products undergoing trials in the European Economic Area, receives reports from National Medicine Regulatory Authorities and Marketing Authorisation Holders that are submitted by patients and healthcare professionals.48 Prior to Brexit, the UK data collected from the Yellow Card System, managed by Medicines and Healthcare Products Regulatory Agency (MHRA), was integrated into EDSADR, as the centralised database the European Economic Area. Post Brexit, MHRA directly manages data collected via the Yellow Card System, independent of EDSADR.49 Many of the national databases input individual case reports into Vigibase (online supplemental material 8), the database established by the WHO Programme for International Drug Monitoring, which bears the greatest resemblance to a centralised monitoring system for CBM-related adverse events.32 42 50 51

Other countries have established registries, either for observational studies or for regulatory purposes. Although some of these systems provide data targeted to cannabis, many rely on spontaneous reporting and are, therefore, subject to selection bias from affiliations.

Mandatory versus spontaneous reporting systems

The AIFA e-registry, to our knowledge, remains the only completely mandatory reporting system captured by our search, obligating patients to submit any side effects and adverse events experienced. The Italian Medicines Agency, under a reimbursement scheme called the Managed Entry Agreement, established an e-registry for all patients commencing on Sativex, to identify ‘non-responders’ for subsequent reimbursement and discontinuation of treatment.52 EDSADR mandates Marketing Authorisation Holders and National Competent Authorities to submit reports of adverse events received from patients and healthcare professionals. However, patients and healthcare professionals are not required to report adverse events.48 CVAR and TGA follow a similar reporting structure to EDSADR.42 46 Similarly, German pharmacists are obliged to report encountered suspected adverse reactions to the Drug Commission of the German Medical Association, however, consumers are not required to report side effects and adverse events.53

The ad hoc nature of reporting requirements in many databases and registries risks a variety of reporting biases.54 55 These include under-reporting, notoriety bias and the preference to only report severe, usually rare, adverse events.54 Spontaneous adverse drug reaction reporting typically peaks following the second year of marketing, then subsequent declines, unaccompanied by changes in drug usage or adverse event incidence.56 57 Under-reporting can be secondary to complacency, where adverse events are (incorrectly) believed to have been already well documented following marketing. Uncertainty surrounding causal relationships is a further contributing factor to under-reporting in spontaneous systems. Additionally, fear of medicolegal consequences, alongside overall clinician indifference is known to further discourage consistent reporting.55 58 59

In November 2017, the change from spontaneous to mandatory submission of suspected ADRs from Marketing Authorisation Holders and National Competent Authorities to the Eudravigilance database, resulted in a significant increase in the number of reports collected.60 Therefore, mandatory reporting framework appear to mitigate issues of under-reporting. Additional interventions such as financial incentives, training on ADR selection and existing reporting systems, as well as continuous feedback on safety signals identified may lessen other reporting biases intrinsic to spontaneous reporting systems.

Causality assessment

Although postmarketing surveillance is more likely to identify a strong causal relationship between CBMs and adverse events, rather than certain proof of causality, there is variability in the level of causality assessment accepted by each monitoring system.61 The Italian Phytovigilance database and CVAR both use the WHO-UMC causality classification system, whereas many other registries and databases such as the FAERS and DAEN do not implement a formal causality assessment process. Within these systems, reports are broadly classified as monitoring ‘suspected’ adverse events. Factors known to help determine the strength of a causal relationship include the temporal relationship between the commencement of CBM and adverse event onset, as well as response to ceasing the CBM and subsequent readministration.57 The FAERS reporting form provides an opportunity for details on response to dechallenge and rechallenge with the drug agent, however, these fields are not mandatory for report submission. VigiBase, collecting case reports from national pharmacovigilance centres internationally, notes discrepancies between reports and does not validate causality claims.32 43 50 62 Eudravigilance similarly accepts reports of varying strengths of causality.48 A standardised method of determining causality across monitoring systems may improve the strength of safety data derived from reports of adverse events.

Diversity in report quality and detail

With no formal assessment of the quality of data collection within the monitoring systems, there are further inconsistencies in the level of detail collected by each monitoring system. Some databases do not specify the dosage or route of administration of the CBM, decreasing the utility of available data in determining accurate safety data. Information collected on patient demographics also varies significantly between systems, especially when considering a patient’s comorbidities, pregnancy status and concomitant medications. As potential confounding factors in determining the cause of the adverse reaction, variations in these demographic details impacts the interpretation of safety data collected.63 64

Variability in quality of reports poses an additional problem in centralised databases such as VigiBase, where data are derived from various sources. Between healthcare professionals, consumers and manufacturers from over 150 countries,43 50 62 65 differences exist in terminology, coding practices and reporting standards across these different regions and healthcare systems. Standard terminology for drug reactions and medicinal products is codified using the Medical Dictionary for Regulatory Activities (MedDRA) and WHODrug classifications respectively.66 However, data at a national level may be mistranslated and inexact when transferred to the standardised VigiBase terminology. For instance, CBD-dominant cannabis products may be coded as CBD according to the standardised WHODrug system, when their THC content would qualify them to be coded as C. Sativa whole extracts. The current MedDRA classification for severity of adverse reactions also provides limited descriptors for cannabis-related adverse events, such as ‘cannabis hyperemesis syndrome’ or ‘cannabis dependence’ and ‘withdrawal’.42 Such discrepancies limit data comparability and complicate assessment of adverse events associated with CBM usage.

The diversity of data sources reporting to a centralised system such as VigiBase predisposes to potential duplication reports. Although efforts are made to identify duplicates, slight differences in nomenclature of related reports may allow duplicates to bypass the algorithm employed by VigiBase.66 As such, the development of one central reporting system with standardised nomenclature and formatting may help with the imprecision in adapting multiple data sources into a standardised framework.

Affiliations

The affiliation of certain registries with independent ownership presents risk of selection bias when evaluating results. The Registry, a multicentre observational research registry collecting data from the UK, Germany and Switzerland, is sponsored by GW Pharmaceuticals, the manufacturers of the THC: CBD oromucosal spray (Nabiximol). Prescribers were identified and invited to participate in data collection by GW Pharmaceuticals, and nominally compensated for completing case report forms.33 UK Medical Cannabis Registry is maintained by Curaleaf Clinic, inviting patients from a private healthcare setting not representative of the broader population of CBM consumers.44 67–85 Similarly, the Australian Emyria Clinical e-Registry sources participation from Emerald Clinics, a network of clinics specialising in use of currently unregistered medicines and commercialisation of collected clinical evidence with Spectrum Therapeutics, the medical division of a cannabis company known as Canopy Growth.44 86 Of note, patients from these registries are a specific subset of CBM consumers, and these registries are aligned with various companies that have interests outside that of the accumulation of real-world safety data.

Accessibility to reporting and safety data

Accessibility to reporting forms, time constraints and awareness of existing reporting schemes have been forwarded as factors limiting participation in monitoring systems.87–89 Additionally, there exists a delay between onset and recognition of the adverse drug reaction, and another lag between reports being input into national pharmacovigilance centres and successful transfer into a central monitoring system such as VigiBase.66 Between access issues to reporting forms, as well as access to safety data published by databases, current monitoring systems are difficult to incorporate into the busy workflow of clinical practice. As such, monitoring of CBM adverse events, from reporting to publishing of safety data, requires a streamlined approach to parallel speed of CBM uptake.

Limitations

Information on details collected in smaller registries is limited by the availability of published data, as often the original reporting form was inaccessible via a secondary Google Search. Additionally, the discipline of pharmacovigilance posits a difference between the definition of side effects, adverse events and adverse reactions.90 However, these terms were used interchangeably in papers, and therefore, adverse events were assumed to encompass side effects.