|
Abstract
|
n a recent work 1, we argued that the estimated road traffic crash (RTC) risk associated with cannabis intoxication had been overestimated in the earlier meta‐reviews of Asbridge et al. 2 and Li et al. 3, each based on nine studies. In a replication study we showed that these were biased by a set of methodological issues that—when corrected—resulted in significantly lower estimates. We also showed that these revised estimates were in line with results from a larger set of 21 studies that we were able to identify.
In their Commentary to our paper, Gjerde & Mørland point to two further methodological issues that we overlooked and which need to be corrected 4:
We interpreted our results as estimates of the increase in crash risk associated with intoxicated use—but many of the underlying epidemiological studies scored case and control counts as positive using criteria that would also be satisfied by drivers with recent (or regular) cannabis use who were neither intoxicated nor impaired while driving. As a consequence, our estimates are biased downwards, and the actual risk associated with acute intoxication will be higher.
Three of the 21 studies used in the main analysis used different methods to assess cases and controls without calibrating the methods to ensure that results were comparable.
These are valid and potentially important issues that neither we nor the earlier meta‐reviews had addressed adequately, and we are grateful to Gjerde & Mørland for bringing this to our attention. The conclusions they draw from this, however, are most probably wrong: they suggest that acute cannabis intoxication is associated with an increased crash risk OR in the range of 2–14, but risk increases in this range appear largely inconsistent with the low average risk of cannabis‐positive drivers in general. Analyzing plausible upper bounds, we find that the average OR of acutely intoxicated drivers is unlikely to be substantially above 2.
Of the issues raised, the second turned out to be inconsequential in practice: removing these three studies from our sample and re‐estimating the pooled effects yields an odds ratio (OR) of 1.35 (1.12–1.61), rather than the original 1.36 (1.15, 1.61). The first issue, however, means that our study should have used the term ‘intoxication and recent use’ rather than ‘intoxication’ or ‘acute intoxication’. The average risk increase for ‘intoxication and recent use’ will sometimes be the relevant rate; for instance, when assessing the overall harm of tetrahydrocannabinol (THC)‐positive drivers identified in statistics based on similar low thresholds. The re‐interpretation of our estimates, however, raises the question of what the RTC risk increase associated with acute cannabis intoxication actually is.
Gjerde & Mørland argue that the risks when driving under acute cannabis intoxication are best assessed by considering four studies that have estimated the OR of drivers with THC levels above 5 ng/ml, a level that, in their words, ‘may indicate acute cannabis intoxication’. Three of these estimates are from studies included in our meta‐analysis, while the fourth is based on a subsample of one of the others and should not be viewed as an independent estimate (see Supporting information to our original paper). The three remaining OR estimates are 2.12, 6.6 and 14.32, defining an OR range of 2–14 that Gjerde & Mørland describe as ‘representative’ of the risks of acute cannabis intoxication. This range, however, is most probably strongly inflated and misleading.
First, the lower end of the range is overestimated, as two of the estimates are from culpability studies which need to be adjusted for baseline culpability rates 1. After adjustment, the three OR estimates are 1.6, 5.0 and 14.3.
Secondly, the upper end of the range, based on an estimate in Kuypers...
Keywords: Cannabis impaired driving
Language: en |