As cannabis legalization spreads and cultural perceptions of its safety solidify, a critical assumption deserves scrutiny: that smoking cannabis is meaningfully safer for the lungs than tobacco. A large retrospective analysis now challenges that assumption with some of the most statistically robust population-level evidence to date, carrying implications for anyone who views cannabis as a relatively benign recreational or therapeutic habit.
Drawing on two decades of electronic health records from 67 U.S. tertiary healthcare systems, researchers constructed propensity-score-matched cohorts of 149,632 individuals with cannabis use disorder and an equal-sized comparison group without it. After controlling for demographic factors and known lung cancer risk factors, those meeting criteria for cannabis use disorder faced a relative risk of 3.87 (95% CI 3.43–4.38) for developing lung or bronchus cancer — nearly a fourfold elevation. Crucially, elevated risk persisted at both one- and five-year follow-up intervals and spanned all major histologic subtypes examined: small cell carcinoma (RR 2.70), squamous cell carcinoma (RR 2.90), and adenocarcinoma (RR 2.54).
The breadth across histologic subtypes is particularly notable. Squamous cell and small cell carcinomas are classically tobacco-linked malignancies driven by proximal airway exposure to combustion byproducts — their elevation here is biologically coherent. But the adenocarcinoma signal is more surprising, since that subtype is more associated with peripheral lung tissue and, increasingly, non-smokers. This could suggest either residual confounding from mixed tobacco-cannabis use (a well-known methodological challenge in this research domain), or a distinct carcinogenic mechanism from cannabis combustion products such as polycyclic aromatic hydrocarbons, which are present at comparably high or higher concentrations than in tobacco smoke.
The retrospective design and reliance on a clinical diagnosis of cannabis use disorder — which likely represents heavy, chronic users rather than occasional consumers — limits generalizability. Propensity matching reduces but cannot eliminate confounding, especially for unmeasured variables like concurrent tobacco use, occupational exposures, or genetic predisposition. Still, with nearly 300,000 matched subjects and consistent effect sizes across cancer subtypes, this study is incrementally more than routine — it warrants serious attention from clinicians and public health communicators alike.