Substitution of prescription opioid use by medical cannabis is not public health relevant, as it is restricted to substitution of a small subset of less potent opioids such as codeine. In addition, it substitutes one substance with abuse potential by another, and there are better alternatives available. Liang and colleagues 1 have presented a time–series analysis based on 21 years of state-level data of Medicaid records, in which they found that medical cannabis legalization was associated with substantial reductions in the number of Schedule III opioid prescriptions in the United States (−30%), their dosage (30%) and related Medicaid spending (−29%). Prescription opioids in Schedule III are characterized by lower abuse potential compared to opioids in Schedule II, with abuse of these opioids potentially leading to moderate or low physical dependence and/or high psychological dependence 2. In the analyses of Liang and colleagues 1, the Schedule III prescriptions made up only approximately 5% of all opioid prescriptions for the last year of analysis (2014), with codeine prescriptions comprising 99% of all Schedule III prescriptions. Codeine is used in the United States to treat mild to moderate pain and as a cough suppressant. Cannabis has been in use for both of these indications, and thus a substitution effect may be plausible. What would be the public health implications of a partial substitution? First, the overall substitution was not found to be affecting a large portion of the overall opioid prescription use. Secondly, however, codeine has an abuse potential. Although this abuse potential was evaluated to be lower compared to Schedule II opioids, abuse has been documented with both prescribed and over-the-counter codeine 3. Furthermore, codeine use may contribute as a gateway to other opioids, both prescription and illicit opioids, with high overdose potential 4, 5. Overall, the United States has, by far, globally the highest prescription opioid use per capita 6, and probably also the highest overall opioid use per capita in the world (i.e. combining prescription and illicit use), at least if indirectly inferred by opioid use disorders 7. The overall culture of medically overusing opioids has contributed markedly to this situation 8. What about cannabis? Its scheduling currently seems to be in flux, with many US states starting medical marijuana programs 9. However, while its consequences are not as detrimental as those of other drugs, in particular in terms of mortality 10, they are far from benign 11. Finally, the effectiveness of both codeine and cannabis for relieving moderate chronic pain seems to be limited, with no Cochrane Reviews supporting such a claim as primary drug of choice (e.g. 12, 13). Perhaps this judgement about effectiveness is premature for cannabis, with research to collect better evidence only now commencing, but it reflects the current evidence base. However, from a public health point of view, given the abuse potentials of both drugs involved, perhaps solutions should be sought with non-pharmacological therapies or with medications with less abuse potential than opioids or cannabis. The history of opioids in pain medication in North America should be a lesson. Despite limited evidence for effectiveness, prescription opioids were seen as a wonder drug, especially for pain management, with rapidly accelerating use to the point where they contributed markedly to a high level of opioid addiction on the population level and a major public health crisis 14, and decreasing life expectancies 15, 16. As a result, new national guidelines for opioid pain prescription stress non-pharmacological therapies and non-opioid medication as first-line treatment 17. We should not repeat this history with yet another wonder drug installed before proper evidence for effectiveness and well-documented unintentional consequences. Given the current evidence, there is no good reason not to start pain management with non-pharmacological therapies, but there are good reasons to limit the use of drugs with high abuse potential to the highest degree possible. None. The author acknowledges funding from the Canadian Institutes of Health Research, Institute of Neurosciences, Mental Health and Addiction (CRISM Ontario Node grant no. SMN-13950).