Stoned driving

It's hard to get a good read on the risks of driving while stoned. It matters because it's one of the less crazy objections to marijuana legalisation: if penalising driving while stoned is hard, and if it is risky, and if consumption increases under legalisation, that could be a cost of legalisation.

Existing estimates are a bit of a problem. Studies finding associations can fail to account properly for alcohol use, or for that cannabis in the system does not necessarily indicate recent use. There are blood tests that can more reliably indicate acute intoxication, and saliva tests, but urine tests don't say much about recent use.

Ole Rogeberg and Rune Elvik look back at a couple of meta-studies and tries to make sense of things. They go back and properly account for how different studies measured drug use, what other confounding factors were included, and whether there were any dose-response relationship established.

They find current metastudies overestimate the effects of cannabis on crash risk. They note that the relative risk of driving at 0.05 BAC is 3.6. Controlling for alcohol use reduces the relative risk of cannabis use from about 1.7 to about 1.2.

They conclude:

A comprehensive review of the literature on acute cannabis intoxication and road traffic crashes finds that acute intoxication is related to a statistically significant risk increase of low to moderate magnitude. Higher estimates from earlier meta-reviews were found to be driven largely by methodological issues - in particular, the use of counts data without adjustment for known confounders. Correcting for these issues, the pooled estimates from these reviews were in line with the results from the updated and more extensive review.

How big is a relative risk of either 1.7 or 1.2?
A few "driving while" risks: 

• Driving while using a cell phone, whether hands-free or not, is about four times riskier than baseline.
• Driving while using a cell phone, in simulator, is about as risky as driving while at .08, whether hands-free or not.
• But note that a really rather nice regression discontinuity design shows no real effect, and a potential upper-bound risk estimate around 3 times baseline. Ungated discussion here.
• If you're getting six hours of sleep per night, you're driving at the equivalent of .05. If you've been awake for 24 hours, you're comparable to someone at .10.
• Should we ban new parents from driving on basis of likely sleep deprivation?
• Older adults with cataracts (but still allowed to drive) were 2.5 times as likely to have had a history of at-fault crashes than those without cataracts. 
• Old people and young people are risky. Compared to 30-59 year olds (baseline), those over 70 have twice the risk of fatal accident involvement, as do 19 year olds. A 16 year old with passengers is 4.72 times as risky as a 30-59 year old with passengers; a 20-24 year old with passengers is 2.54 times as risky.
• Drivers over the age of 85 had 10.62 times the baseline risk of multiple-vehicle accidents at intersections and 3.74 times baseline risk elsewhere. 
• Drivers in the .05-.079 range are 5.5 times baseline risk; those above .08 are much riskier (15.5 times baseline). Same study has no increased risk for cannabis, but combinations of alcohol and other drugs were very risky.
• People who are left-handed have 2.35 times baseline accident risk.
• Not including any alcohol-related accidents, time of day matters a lot. Driving at 4 in the morning is 5.7 times as risky for accidents as driving at 10-11 am; they chalk it up to sleepiness. Maybe we should just have curfews.

It is frustrating to watch America legalise cannabis, state-by-state, and the UK debate whether legalisation or decriminalisation is best for all drugs, while NZ remains stuck in the stone age.

A comprehensive review of the literature on acute canna-

bis intoxication and road traffic crashes finds that acute

intoxication is related to a statistically significant risk

increase of low to moderate magnitude. Higher estimates

from earlier meta-reviews were found to be driven largely

by methodological issues—in particular, the use of counts

data without adjustment for known confounders.

Correcting for these issues, the pooled estimates from these

reviews were in line with the results from the updated and

more extensive review. Remaining selection effects

discussed in the ‘Alternative interpretations’ section may

complicate causal interpretations of the pooled estimates.

A comprehensive review of the literature on acute canna-

bis intoxication and road traffic crashes finds that acute

intoxication is related to a statistically significant risk

increase of low to moderate magnitude. Higher estimates

from earlier meta-reviews were found to be driven largely

by methodological issues—in particular, the use of counts

data without adjustment for known confounders.

Correcting for these issues, the pooled estimates from these

reviews were in line with the results from the updated and

more extensive review. Remaining selection effects

discussed in the ‘Alternative interpretations’ section may

complicate causal interpretations of the pooled estimates.