Table 4.

Studies related to the legalization of recreational cannabis: design, main results, and limitations.

Article	Statistical Analysis	Main Results	Limitations
Callaghan et al. (2021) [30]	Interrupted time-series analysis. Seasonal autoregressive integrated moving average models	There is no significant effect of increased injury visits among young drivers and all drivers before and after legalization. Significant differences (95% CI −26.32; 84.19; p = 0.30) are observed for all drivers in Ontario in emergency department visits.	Data are only collected from two provinces in Canada and from cases that have resulted in moderate to severe injuries.
Tefft and Arnold (2021) [31]	Logistic regression and marginal standardization.	Increase in the proportion of drivers who tested positive for THC from 9.3% before and 19.1% after legalization (APR: 2.3, 95% CI: 1.3, 4.1) and in the concentration of THC (APR: 4.7, 95% CI: 1.5, 15.1).	There is a significant number of drivers for which there is no drug test. Bias in the results by not taking into account other variables. Data are not compared with other states that have not legalized RC.
Windle et al. (2021) [32]	Poisson regression, meta-analyzed estimates, and DerSimonian and Laird random-effects models.	Increase in fatal traffic accidents (IRR 1.15, 95% CI 1.06–1.26) and deaths (IRR 1.16, 95% CI 1.06–1.27) in the first year after legalization.	Observational study. Jurisdiction differed among states.
Lensch et al. (2020) [33]	Chi-square tests. APR and 95% CI.	Higher incidence of use in states that legalized cannabis in the previous 30 days (APR: 1.34; 95% CI: 1.19, 1.51) and in the previous 12 months (APR: 1.16; 95% CI: 1.06, 1.28). Higher protective behaviors in states that have legalized RC.	Cross-sectional study. Non-representative sample of the general population.
Nazif-Muñoz et al. (2020) [34]	Interrupted time-series analysis. Extension autoregressive integrated moving average.	Significant increase in the light motor vehicle driver fatality rate (CI = 11.6, 93.3, p = 0.012). Significant increase in automobile driver mortality in Montevideo (CI = 0.01, 0.11, p = 0.025) but not in rural areas.	Prevalence of cannabis use in traffic accidents. Accidents with injuries are not considered. Possible biases in the coding of accidents.
Rotermann (2020) [35]	t-test statistics.	Stability in the number of cases of driving after having consumed before and after legalization, being more frequent in men than in women (p < 0.05). In general, decrease in the number of cases of traveling in a vehicle whose driver had consumed.	Self-report data. The type of design does not allow for causal inferences. The study is limited to surveyed households only.
Borst et al. (2020) [36]	Multivariate logistic regression. Linear regression. Binomial logistic regressions. Pearson χ2. Time-series regression analysis	A 7.6-percentage point increase of THC+ cases in accidents after legalization. The THC+ group used fewer protective measures while driving (8.5% vs. 14.3%, p < 0.001) and suffered more serious injuries (8.4 ± 9.4 vs. 9.0 ± 9.9, p < 0.001) than the non-consumption group.	There is no legal threshold for driving under the influence of cannabis. Time in which cannabinoids are maintained in the blood. No toxicological screening of all patients. Variability in detection rates between institutions.
Hansen et al. (2020) [37]	Synthetic control approach. Permutation testing of the ratio of mean squared error ratios	Control states show the same increases in the number of accidents per billion vehicle miles traveled in Colorado (p = 0.361) and Washington (p = 0.404).	Levels of marijuana use not quantified and variation among states. Only fatal traffic accidents. No causality can be asserted.
Santaella et al. (2020) [38]	Ecological study used a synthetic control approach. Mean square prediction error.	Increase (p = 0.047) in traffic fatalities in Colorado but not in Washington following legalization of RC.	Other variables associated with legalization. Time of legalization in Washington. No data on injuries. Not included if the driver tested positive for THC at the time of the accident.
Delling et al. (2019) [3]	Linear, logistical, and multinomial models.	An increase (p < 0.05) in the number of traffic accidents, alcohol abuse, overdose injuries and a decrease in chronic pain admissions is observed following the legalization of RC versus control states.	The post-legalization period is short (two years). They only compare with two control states. Colorado legalized medical marijuana prior to 2012.
Aydelotte et al. (2017) [39]	Random effects multivariate regression for panel data. Difference-in-differences approach.	Despite an increase, there is no significant association between legalization and changes in the rate of fatal traffic accidents (ADDC = +0.2 fatalities/billion vehicle miles traveled; 95% CI = –0.4, +0.9) in the first 3 years of legalization.	They only analyze two states in which the substance has been legalized. They do not study adjacent control states. They study fatal traffic crashes as a whole, not those in which cannabis is involved.
Aydelotte et al. (2019) [40]	Retrospective longitudinal cohort study. Difference–indifference analyses.	Significant increase in the number of fatal traffic accidents in states that legalized RC after opening dispensaries to buy recreational marijuana (CI: +0.4 to +3.7, p = 0.020) compared to control states. This increase was not significant before the dispensaries opened (CI: −0.6 to 2.1, p = 0.087).	Do not make any statistical adjustments for multiple analyses. Biases in the selection of control states. Missing more states where recreational marijuana use has been legalized. Only study fatal traffic accidents.
Lane and Hall (2019) [41]	Interrupted time-series design. Generalized least squares regression models.	The general trend is an increase in the mortality rate both in states with legalization laws and in neighboring states. There is a trend of increased mortality (p < 0.001) at the beginning and a decrease in the subsequent months (p < 0.001). At 6 months these effects are greater (step: 1.36, p = 0.006; trend: −0.07, p < 0.001).	They do not differentiate between fatal traffic accidents in which marijuana is involved and those in which it is not. They do not consider other types of accidents. They do not analyze economic factors or current policies of neighboring states regarding cannabis.