A new modelling study claims recreational cannabis in Canada may be linked to roughly 28,000 to 30,000 premature deaths each year, but these are not directly observed deaths. They are statistical estimates derived from previously published observational data, aggregated across multiple disease categories, and should be interpreted with considerable caution before informing clinical decisions or policy.

Canada’s 2018 Cannabis Act represented one of the most significant drug policy experiments in modern history, and honest evaluation of its population-level health consequences is essential for evidence-based governance worldwide. When a study claims tens of thousands of annual deaths attributable to a legal substance, those numbers inevitably shape political debate, clinical attitudes, and patient anxiety. Whether they withstand methodological scrutiny determines whether they advance or undermine the cause of sound public health policy. Clinicians and patients alike need to understand what these figures actually represent before drawing conclusions.

Since Canada legalized recreational cannabis in 2018, roughly 22% of Canadians aged 15 and older have reported use, raising legitimate questions about downstream health effects at the population level. This study, authored by researchers from Western University’s engineering and business faculties, attempts to quantify those effects by applying population-attributable fraction (PAF) calculations to Canadian mortality data across six outcome categories: cardiovascular disease, neurocognitive disorders including dementia and Alzheimer’s disease, lung and colon cancer, injury mortality, suicide, and opioid poisoning. The study also introduces a parallel land-use analysis, modelling what would happen if over two million square metres of cannabis cultivation land were redirected to nutrient-dense food production.

The headline finding is an aggregated estimate of approximately 28,000 to 30,000 premature deaths per year attributable to recreational cannabis use, with cardiovascular disease and dementia accounting for the largest shares. On the agricultural side, the authors estimate counterfactual crop substitution could feed more than 3,600 people annually. However, these figures are model outputs, not directly observed outcomes. The underlying risk ratios are drawn from observational studies with heterogeneous confounding control, the PAF methodology assumes a causal relationship that is not established for many of the included outcomes, and the land-use scenario is hypothetical with no modelling of economic or agronomic feasibility. The authors themselves acknowledge these limitations and call for more rigorous, longitudinal research, but the framing of the paper strongly implies a policy-advocacy orientation that risks overstating the evidentiary base.

I appreciate the ambition of trying to quantify the public health footprint of cannabis legalization, because that question genuinely matters. But the gap between what this study can actually show and what its numbers imply is enormous. Population-attributable fractions are only as reliable as the causal assumptions baked into them, and for outcomes like dementia and cardiovascular disease, the epidemiological evidence linking cannabis to causation (rather than correlation) remains contested and incomplete. When you sum PAF estimates across six different disease categories, each with its own confounding issues, the resulting aggregate number inherits every weakness of every input. It looks precise. It is not.

In my practice, I talk with patients about real cardiovascular and cognitive risks that may accompany heavy or chronic cannabis use, particularly for patients with preexisting conditions. I do not cite a single mortality number, because no single number exists that is reliable enough to anchor that conversation. What I emphasize is individualized risk assessment: age, frequency of use, route of administration, comorbidities, and medication interactions. That approach respects the patient more than a headline ever could.

This study sits at the modelling and synthesis stage of the research arc, which is valuable for hypothesis generation and policy discourse but insufficient for establishing causal certainty. Clinicians should note that the cardiovascular risk estimates cited in the paper draw on studies like those by Desai and colleagues, which found elevated hazard ratios for heart failure and stroke among cannabis users but could not fully disentangle concurrent tobacco use, socioeconomic confounders, or polysubstance exposure. Similarly, the dementia risk estimate (a 72% increase) comes from a single cohort study with specific demographic characteristics that may not generalize uniformly to all cannabis-using populations.

From a pharmacological standpoint, clinicians working with cannabis patients should remain attentive to the plausible cardiovascular and neurocognitive signals in the literature without treating this paper’s aggregated mortality figure as established clinical fact. The most relevant drug-interaction concern in this context involves patients on anticoagulants, antihypertensives, or CNS-active medications, where THC’s effects on heart rate, blood pressure variability, and cognitive load may compound risk. The actionable recommendation is straightforward: continue screening cannabis-using patients for cardiovascular risk factors, counsel those with preexisting conditions about the uncertain but plausible risk signal, and avoid citing this specific mortality figure in patient discussions until higher-quality evidence either confirms or refutes it.