Canada’s post-legalization pharmacovigilance system collected 698 cannabis adverse reaction reports over six years, with hallucinations, headache, and nausea among the most common events, predominantly reported by older medical users of extract products. However, without knowing how many people actually use legal cannabis, these numbers describe the reporters, not the true frequency of harm in the broader population.

When Canada legalized cannabis in 2018, it became one of the first major jurisdictions to create a national post-market surveillance system for a product used by millions. Six years later, the first comprehensive published account of what that system has captured is now available. For clinicians prescribing or counseling medical cannabis users, for regulators designing risk communications, and for patients trying to understand what adverse events have been flagged, this baseline dataset marks the starting point for evidence-informed cannabis safety monitoring. The patterns it reveals, particularly the overrepresentation of older adults, extract products, and events like hallucinations, demand clinical attention even as the system’s fundamental inability to quantify risk demands intellectual caution.

When Canada legalized cannabis for non-medical purposes in October 2018, the federal government extended its existing Canada Vigilance Program to capture adverse reaction reports for legal cannabis products. This study, authored entirely by Health Canada staff, represents the first published descriptive analysis of the spontaneous adverse reaction reports accumulated over six years. The reporting system functions differently from controlled research: licence holders must report serious events, while consumers, patients, and healthcare providers submit reports voluntarily. The system is designed as a signal detection tool for identifying patterns that may warrant further investigation, not as a mechanism for calculating how often adverse events occur among cannabis users.

Among the 698 cases, the mean reporter age was 56 years, with 54.6% male, and 67.5% self-reporting medical use, most commonly for pain management. Cannabis extracts accounted for 68.8% of implicated products. The five most frequently reported adverse events were hallucination, headache, nausea, dizziness, and dyspnea. Of the cases, 62.3% were classified as “serious,” though 58.6% of those fell into the broad “other medically important condition” category rather than life-threatening events or hospitalizations. Causality assessments, available only for 2022 to 2024, rated most events as “possible,” the second-lowest affirmative tier in the WHO-UMC framework. The authors explicitly state that statistical testing is not appropriate for spontaneous reporting data and that additional scientific investigations are required to establish cause-and-effect relationships.

Canada’s Cannabis Safety Reporting System: Six Years of Signals Without a Denominator

When Canada federally legalized cannabis in 2018, it inherited a pharmacovigilance framework built for pharmaceuticals and adapted it for a product used by millions recreationally. Six years later, the system has collected 698 adverse reaction reports, a number that sounds reassuringly small until you realize we have no idea what fraction of all adverse events it represents. This paper from Health Canada honestly describes what was reported to a voluntary surveillance system. It does not, and cannot, tell us how common adverse events are among the millions of Canadians who use legal cannabis. I find the authors’ intellectual honesty commendable: they decline to run inappropriate statistical tests, they explicitly warn readers not to infer causal relationships from case reports, and they restrict their causality analysis to a three-year window where methodology was consistent. That restraint is itself a contribution to the field. But it also means that any headline claiming “62% of cannabis reactions are serious” would be misleading in the extreme. That figure describes the reporting system’s intake, not the cannabis experience broadly. Counting the number of calls to a car complaint hotline tells you about frustrated callers, not about the overall safety record of the vehicle. You need to know how many cars are on the road to calculate a breakdown rate. The same logic applies here: 698 reports over six years, without knowing the denominator of total users, tells us only about who picks up the phone.

The pharmacologically interesting findings are the signal patterns by product class. Hallucinations clustered with THC-predominant products, which is consistent with established cannabinoid pharmacology. Dyspnea appearing prominently among a cohort where nearly 70% of cases involved extract products, many of them inhalable, warrants clinical attention. But these observations are hypothesis-generating correlations within a biased reporter sample, not controlled comparisons. The reporter demographics themselves are revealing: a mean age of 56 years, two-thirds self-reported medical use, with pain as the leading indication. This does not mean medical cannabis is more dangerous than recreational cannabis. Medical users are older, have more complex health profiles, are more likely to interact with the healthcare system, and are therefore more motivated to report adverse events. Saying this reporting system captures the cannabis safety landscape is like evaluating a restaurant’s food quality based solely on its complaint line. People who had a pleasant meal rarely call to report it. The sample is systematically skewed toward those who had problems and were inclined to describe them formally.

As a clinician, I find this paper most useful as a counseling prompt for a specific patient subgroup: older adults using cannabis extracts for pain. This population appears to be the primary reporter cohort, and the events they experience, including hallucinations and respiratory symptoms, are clinically meaningful even if their population frequency remains unknown. I would tell these patients that Canada’s safety system has flagged certain patterns worth knowing about, that extract products and higher THC concentrations may carry particular risks for older users, and that we cannot yet quantify those risks precisely. For colleagues, I would emphasize that this is not a risk study; it is a surveillance inventory, and the “possible” causality rating is the weakest affirmative assessment available, meaning cannabis could not be ruled out but other explanations remain plausible. The value of this paper lies not in what the numbers tell us about cannabis risk, but in what they reveal about the structure and limitations of the surveillance system itself. Health Canada has built something functional and transparent: a regulatory smoke detector for a newly legalized market. But policymakers, clinicians, and patients who want to understand how likely a harmful event actually is when someone uses a legal cannabis extract will need a different kind of study entirely, one with a denominator, a control group, and ideally, an active rather than passive surveillance design. Until that evidence exists, the message from this dataset is appropriately humble: signals detected, incidence unknown.

This study sits at the very beginning of the cannabis pharmacovigilance research arc. It establishes a descriptive baseline for what Canada’s surveillance system captures, but it does not advance our understanding of cannabis risk in any quantitative sense. It is analogous to the first published summaries of pharmaceutical adverse event databases: useful for identifying what kinds of signals are appearing, essential for regulatory planning, but not a substitute for the pharmacoepidemiological studies that must follow. Clinicians should view these findings as a catalog of what has been reported, not as a risk profile for their patient population.

The overrepresentation of cannabis extracts (68.8% of cases) and of hallucination as a leading event warrants heightened counseling for patients using inhalable or ingestible extract products, particularly those who are older, medically complex, or new to cannabis. Clinicians should also note that many reporters were using cannabis concurrently with other medications, raising the possibility of pharmacokinetic interactions, especially with drugs metabolized by cytochrome P450 enzymes. The most actionable recommendation from this data is to ask older medical cannabis patients specifically about hallucinations, respiratory symptoms, and dizziness, and to document these conversations in the clinical record.

This is a descriptive cross-sectional analysis of spontaneous adverse reaction reports from a national pharmacovigilance database. In the evidence hierarchy, it ranks below observational cohort studies and far below randomized controlled trials. Its evidentiary authority is limited to characterizing what was reported to the surveillance system, and the single most important inference constraint is the absence of a denominator: without knowing the total number of cannabis users, no frequency, probability, or incidence rate can be calculated from this data.

This study adds a Canadian national dataset to the small but growing international pharmacovigilance literature on legal cannabis. Its findings are broadly consistent with adverse event profiles reported in other post-market surveillance systems, including the U.S. FDA’s FAERS database and European national reporting systems, where hallucinations, gastrointestinal symptoms, and respiratory events also appear among frequently reported cannabis-associated events. The demographic skew toward older medical users mirrors patterns observed in U.S. state-level cannabis adverse event tracking, reinforcing the interpretation that surveillance systems disproportionately capture medical rather than recreational user experiences. What this study uniquely contributes is a longitudinal six-year dataset from a jurisdiction with full federal legalization, providing a structured baseline that future studies can reference as Canada’s cannabis market and regulatory landscape continue to evolve.

The most consequential analytic choice was the decision, appropriate though it was, not to perform any formal statistical testing. Had the authors attempted to compute crude reporting rates using even rough denominator estimates from Statistics Canada’s National Cannabis Survey, they could have produced preliminary incidence rate approximations, which would have been methodologically imperfect but clinically informative. Additionally, restricting causality assessment to the 2022 to 2024 period means that almost half the surveillance timeline contributes only descriptive data without any formal assessment of whether cannabis plausibly contributed to the reported events. Had earlier causality data been available or had the authors applied retrospective reassessment to earlier cases, the overall causality profile might have shifted. Neither of these alternative approaches would overcome the fundamental denominator problem, but they could have offered a more granular picture of reporting trends over time.

The most likely and most consequential misreading is treating the finding that 62.3% of cases were classified as “serious” as evidence that cannabis commonly causes serious harm. This figure describes the composition of reports received by the surveillance system, not the probability of a serious event among cannabis users. Voluntary reporting systems systematically over-capture serious events because people with minor or transient symptoms rarely file formal reports. Similarly, the predominance of medical users (67.5%) does not indicate that medical cannabis is inherently more dangerous; it reflects the fact that medical users are more connected to healthcare systems, more aware of reporting mechanisms, and more likely to attribute symptoms to a specific product. Interpreting these proportions as population-level risk estimates would be a fundamental methodological error.

This study provides the first published descriptive baseline of Canada’s cannabis pharmacovigilance dataset, establishing what kinds of adverse events are being reported, who is reporting them, and which product types are involved. It does not establish how common these events are, whether cannabis caused them, or how Canada’s legal cannabis safety profile compares to pharmaceutical alternatives. For clinical practice today, its most actionable contribution is the signal that older medical cannabis users of extract products warrant heightened adverse event counseling and monitoring.

Does this study prove that legal cannabis is dangerous?

No. This study describes 698 adverse reaction reports submitted to a voluntary surveillance system over six years. Without knowing how many total people used legal cannabis during that period, we cannot determine how common these events are. The data identifies patterns worth investigating further, but it does not quantify risk.

Should I be worried about hallucinations from cannabis?

Hallucination appeared among the most frequently reported adverse events in this dataset, and it clustered with THC-predominant products. This is consistent with what is already known about THC’s psychoactive effects, particularly at higher doses or in people with lower tolerance. If you are older, new to cannabis, or using high-THC extract products, discussing this possibility with your healthcare provider before starting or adjusting use is a reasonable precaution.

Are cannabis extracts more dangerous than dried flower?

The data cannot answer this question. Cannabis extracts accounted for 68.8% of reports, but this could reflect their popularity in the marketplace, differences in how they are consumed, or the tendency of extract users to be more engaged with the reporting system. Without data on how many people use each product type, we cannot compare their relative safety.

Does the “possible” causality rating mean cannabis probably caused the adverse events?

Not quite. In the WHO-UMC causality framework, “possible” is the second-lowest affirmative tier. It means that cannabis cannot be ruled out as a contributing factor, but other explanations, including underlying health conditions and concurrent medications, also remain plausible. Think of it as a detective saying a suspect “cannot be ruled out” rather than a conviction.

Have thoughts on this? Share it: