Table 2.
Studies related to the legalization of medical cannabis: design, main results, and limitations.
| Article | Statistical Analysis | Main Result | Limitations |
|---|---|---|---|
| Jones et al. (2019) [22] | Regression analysis of interrupted time series. | Before legalization, decreased positive drivers by 0.5%/year (95% CI: −1.0/year, 0.0/year). After legalization, a significant increase in the trend of cannabis-positive drivers of 0.6%/year (95% CI: 0.0/year, 0.8/year). |
Differences between and within states in drug testing protocols. Testing positive for THC does not imply recent use. No control group. |
| Wadsworth and Hammond (2019) [23] | Chi-squared tests, nominal logistic regression, and multinomial logistic regression models. | There are differences in driving after cannabis use between countries. England: less likely to drive after drinking than Canada (p < 0.001) and the U.S. (p < 0.001). Canada: less likely than the U.S. (p < 0.001). | Self-report measures. Social desirability. Biases in a recall. General legislation of the country is taken and differences between different districts/cities are not taken into account. Non-random subject selection. |
| Steinemann et al. (2018) [24] | t-tests and Chi-squared tests. | Significant increase (p < 0.001) in the number of drivers testing positive for THC after legalization in 2000 in fatal accidents. | THC+ does not imply recent consumption. A THC level cannot be extrapolated to an accident risk level. There is no discrimination between acute, heavy, and chronic use. |
| Sevigny (2018) [25] | Generalized linear model (GLM) with a binomial distribution and logit link function. Sensitivity analyses. | There is no relationship between the MCL and the number of fatal traffic accidents OR [95% CI] 1.05 [0.93, 1.19]. The implementation of medical cannabis dispensaries does correlate with a higher number of accidents involving cannabis OR [95% CI] 1.14 [1.02, 1.29] (p < 0.01). | THC+ does not imply recent use. Differences between states in drug testing protocols, as well as changes in trends after legalization. Not all crash victims are drug tested. Levels of cannabis use are not systematically quantified. |
| Santaella-Tenorio et al. (2017) [26] | Multilevel regression models with state-level random intercepts. | States with MCL laws have lower rates of traffic fatalities than states without MCL laws (26.3% lower; 95% confidence interval [CI] = 13.9%, 36.9%). A 10.8% reduction in fatalities after MCL (95% CI = 9.0%, 12.5%; % reduction = [1 − exp(−0.114)] − 100). Following the legalization of cannabis, there was a reduction in fatal traffic accidents in 7 states. | Causal relationships cannot be established. Local aspects that may have an influence are not taken into account. Short post-legalization periods. THC tests are not performed. |
| Masten and Guenzburger (2014) [27] | Time series analyses. Auto-regressive integrated moving average analysis. |
MCL is only associated with an increase in fatal traffic accidents in California, Hawaii, and Washington (out of a total of 12 states) after adjusting for the frequency of drug testing by state and the prevalence of cannabinoids in drivers in states without legalization laws. | They only study fatal traffic accidents. State differences in drug testing protocols and changes in trends after legalization. Not all crash victims are tested for drugs. THC+ does not imply recent use. |
| Salomonsen-Sautel et al. (2014) [28] | Linear regression analysis. Estimated generalized least squares (EGLS) methods. |
Significant trend change (2.16 (0.45), p < 0.0001) in Colorado (not in the rest of the states) after legalization, with an increase in THC-positives among drivers involved in fatal accidents. |
They only study fatal traffic accidents. THC+ does not imply recent use. Levels of marijuana use are not systematically quantified. |
| Anderson et al. (2013) [29] | Linear regression analysis by ordinary least squares (OLS) estimates | Legalization leads to a significant decrease in fatal traffic accidents, although when state-specific time trends are included, the decrease is not significant (p = 0.139). Legalization leads to a reduction in crashes where the driver tests positive for alcohol (R = −0.141, p < 0.01) and in cases where the alcohol level is above 0.10 (R = 0.168, p < 0.05). | Not specified |