CHALLENGES AND OPPORTUNITIES FOR FURTHER RESEARCH INVOLVING THE CANNABIS-TO-OPIOID USE TRANSITION

The recent meta-analysis by Wilson et al. [1] found that initial cannabis use was associated with nearly threefold greater odds for subsequent opioid use and opioid use disorders. Here, we propose a few questions to stimulate ongoing discussion around the implications of this work.

First, the component studies lacked detailed descriptions of cannabis use, opioid use and time-lines for substance use (early versus later). Specifically, the frequency and intentions for cannabis use were often lacking, further highlighting the blurred distinction between medicinal and recreational cannabis use. Consequently, the interpretations of the overall associations are quite limited, preventing a more nuanced and contextualized understanding of the underlying relationships.

Secondly, between-study heterogeneity might have affected the authors’ results and interpretation. For example, the authors’ meta-analyses combined life-time and past-year cannabis or opioid use and pooled opioid use disorder, abuse and dependence. Combining these outcomes in the same analysis might have yielded statistically invalid results, as these outcomes are temporally, nosologically and epidemiologically distinct. Given the low study yield and heterogeneous designs (cross-sectional, longitudinal and twin studies), it is unclear if proceeding with meta-analysis was justified. While the authors included E-values to estimate unmeasured confounding, these methods may not have been appropriate because the bias factor may not have been normally distributed across studies [2].

Although the authors used a weighted technique to combine study results, they took some liberties by pooling odds, hazard and rate ratios with variable adjustment for covariates (e.g. age, sex, psychiatric comorbidity, concurrent substance use). For example, some subgroup analyses found no significant effect after covariate adjustment (e.g. cannabis use disorder), yet some subgroup analyses were underpowered, often using just one or two studies. The issues of low study yield and heterogeneous designs suggest that meta-analysis was perhaps not justified.

A few other minor issues stemmed from the meta-analysis’ inclusion of only six non-randomized studies. Although further studies may have been identified had the authors included non-English-language studies, consultation with content experts in the field, grey literature or a search of ongoing trial or study registries, the authors should still be commended for their thorough search strategy. Consequently, while the meta-analyses had low I2 estimates, suggesting low heterogeneity across studies, this may be related to the small study yield. In addition, using the ROBINS-I framework to assess the risk of bias in non-randomized intervention studies may have been an issue, given that the included studies were largely observational designs rather than intervention-based studies [3]. Finally, investigating the risk of publication bias could have been a useful asset worth discussing, given that there were only six eligible studies.

Overall, the recent meta-analysis by Wilson et al. [1] has several strengths, and is a helpful synthesis of the available literature on the transition from cannabis to opioid use and opioid-related disorders. However, additional rigorous and unbiased analyses are needed to more clearly understand the direction of the relationship between cannabis use and opioid-related harms in light of the important public health implications of cannabis use during the ongoing opioid overdose crisis.