Abstract
White MA and Burns NR (2021). The risk of being culpable for or involved in a road crash after using cannabis: A systematic review and meta-analyses. Drug Science, Policy and Law, 7, 1-20. https://doi.org/10.1177/20503245211055381
Jeff Brubacher was the corresponding author for one (Brubacher et al., 2019) of the 13 papers that we covered in our review of the relationship between the prior use of cannabis and the risk of being culpable for a road crash (White and Burns, 2021). Jeff identified an error in the value we selected to represent Brubacher et al’s adjusted cannabis-culpability odds ratio (OR) in our cannabis-culpability meta-analysis and meta-regression. We selected the value of 1.07 (0.98-1.20) that Brubacher et al. provided in their Table 3 for the “Model with THC in ng/ml”. Unfortunately, that value was for the slope of the (non-significant) linear relationship between the concentration of THC (in ng/ml) and the risk of being culpable for a crash, and not for the relationship of interest: the effect of the presence vs absence of THC on the risk of being culpable for a crash. Surprisingly, the value that we sought for the dichotomous cannabis variable was not provided anywhere in Brubacher et al’s paper.
However, Jeff has now generously run the adjusted model for THC > 0.00 ng/ml, which delivered a cannabis-culpability OR of 1.18 (0.80-1.76). So, the cannabis-culpability OR of 1.07 (0.98-1.20) that we reported for Brubacher et al. in our Table 1 should have been 1.18 (0.80-1.76). All of the other information for Brubacher et al’s study that we reported in our Tables 1 and 2 remains correct.
We have re-run our cannabis-culpability meta-analysis and meta-regression, and the results are provided below.
The cannabis-culpability ORs from the 13 studies included in our meta-analysis are given in Corrected Figure 2. The meta-analytic summary OR for the corrected model is 1.41 (1.14-1.74), which is only marginally higher than the uncorrected value of 1.37 (1.10-1.69).
The 13 effect sizes remain heterogeneous Q (12) = 21.9, p = .038; I2 = 45.3%, with moderate-to-high heterogeneity.
Forest plot for the 13 cannabis-culpability studies. The size of each square is proportional to the relative weight that the study contributed to the summary OR, which is indicated by the diamond. The horizontal bars and the width of diamond represent the 95% confidence intervals. Q (12) = 21.9; p = .038; I2 = 45.3%.
A meta-regression assessed our directional bias score as a moderator of the reported cannabis-culpability ORs. For our corrected meta-regression model, the directional bias score was significantly associated with the cannabis-culpability ORs (Q (1) = 8.22, p = .004), and explained 79.6% of the heterogeneity. The intercept of the corrected model (i.e., for a directional bias score of zero) expressed as an OR was 0.75 (0.48-1.19) (see Corrected Figure 3). That value is only marginally higher than the value of 0.68 (0.45-1.05) for the uncorrected model.
Bubble plot of observed ORs for the 13 cannabis-culpability studies against our directional bias score. Bubble size is proportional to the study weight and the solid line is the fitted meta-regression model. The value of the OR that corresponds to a zero bias is 0.75 (0.48-1.19).
We conclude that the changes required to accommodate our error are marginal, and of no consequence for our overall review.
MICHAEL WHITE & NICK BURNS
School of Psychology, University of Adelaide, South Australia
E-mail: