Are Stoned Motorists Driving Amok in Post-Legalization Colorado?

"Marijuana-related" crashes are not necessarily related to marijuana.


National Advanced Driving Simulator

The latest report on marijuana legalization in Colorado from the Rocky Mountain High Intensity Drug Trafficking Area (RMHIDTA), published this month, features all the usual tricks aimed at portraying that policy change as a disastrous mistake, including a misleading presentation of data on traffic fatalities. RMHIDTA, a federally supported anti-drug agency that is determined to paint legalization in a negative light (which it is legally required to do, as NORML's Paul Armentano points out), wants people to believe Colorado's roads are full of dangerously stoned drivers, who are maiming and killing people left and right. But the numbers it uses to create that impression are not an accurate measure of the harm caused by marijuana-impaired motorists.

"Marijuana-related traffic deaths when a driver tested positive for marijuana more than doubled from 55 deaths in 2013 to 123 deaths in 2016," the report says. "Marijuana-related traffic deaths increased 66 percent in the four-year average (2013-2016) since Colorado legalized recreational marijuana compared to the four-year average (2009-2012) prior to legalization." But contrary to what you might think, "marijuana-related" does not necessarily mean related to marijuana. "This report will cite datasets with terms such as 'marijuana-related' or 'tested positive for marijuana,'" a note at the end of the introduction says. "That does not necessarily prove that marijuana was the cause of the incident."

In the case of car crashes, drivers who "test positive for marijuana" include people whose blood contains inactive metabolites or THC levels too low to cause impairment. RMHIDTA muddies this point with a quote from a Denver Post story published last August: "In 2016, of the 115 drivers in fatal wrecks who tested positive for marijuana use, 71 were found to have Delta 9 tetrahydrocannabinol, or THC, the psychoactive ingredient in marijuana, in their blood, indicating use within hours, according to state data. Of those, 63 percent were over 5 nanograms per milliliter, the state's limit for driving." Another way of putting that: Most of the drivers (61 percent) either had no active THC in their blood or had less than five nanograms per milliliter, which is the level at which Colorado juries may infer that a driver was impaired.

That inference, which is rebuttable, will often be mistaken, as many people are perfectly capable of driving safely at THC blood levels far above Colorado's arbitrary cutoff. In other words, even the minority of "marijuana-positive" drivers who exceeded the five-nanogram THC threshold were not necessarily impaired at the time of the crash. Because of wide variation in how people respond to marijuana, there is no scientific basis for a rule that equates any particular THC blood level with impairment.

The increase in "marijuana-positive" drivers since 2013 may reflect a genuine threat to public safety, or it may reflect general increases in cannabis consumption. As the number of Coloradans who use marijuana and the frequency with which they use it go up, so will the percentage of drivers whose blood contains traces of marijuana. That includes drivers involved in fatal crashes, even if marijuana played no role in those accidents.

A new review of the research on marijuana and driving by University of Adelaide psychologist Michael White provides further reason to be cautious about making the assumption that RMHIDTA wants us to make. White's 145-page report focuses on 11 epidemiological studies that used active THC as a measure of marijuana exposure, excluding studies that treated drivers with inactive metabolites in their blood as if they were under the influence. After taking into account various sources of bias and confounding, he concludes "there is no good evidence" to support the hypothesis that marijuana use increases the risk of a car crash. "If cannabis does increase the risk of crashing," White says, "the increase is unlikely to be more than about 30%."

By comparison, research indicates that a blood alcohol concentration of 0.10 percent quintuples the risk of a car crash—an increase of 400 percent. White's upper estimate for the crash risk associated with cannabis consumption is based on a 2016 meta-analysis by Ole Rogeberg and Rune Elvik. He says Rogeberg and Elvik "expos[ed] serious over-estimation biases in two earlier meta-analyses," although he notes that they erred by describing the subjects in the studies they included as showing evidence of "acute cannabis intoxication," since some tested positive only for nonpsychoactive cannabinoids and therefore might have used marijuana days or weeks before.

White also considers experimental studies of marijuana and driving, which involve lab tests, simulators, or driving on closed courses, to see if they provide supplementary evidence that cannabis intoxication contributes to crashes. While those studies do indicate that marijuana affects driving-related skills, he says, the impact on crash risk is not clear. White concludes that "modest decrements in the level of driving-related skills that are sometimes found in the laboratories that have studied the effects of cannabis on human performance are of little relevance to road safety."

White also questions the widespread belief that cannabis compounds alcohol-related impairment, saying the associations found in epidemiological studies can be explained by higher levels of drinking among people who combine pot and alcohol. Looking at the laboratory studies, he finds "the evidence that cannabis exacerbates the impairing effects of alcohol is rather weak." In particular, White doubts that "standard deviation of lateral position" (SDLP, a measure of weaving within a lane) has "much relevance to road safety as a measure of the exacerbating effect of cannabis on the impairing effects of alcohol."

Nor is SDLP necessarily a good indicator of marijuana's impact on crash risk independent of alcohol. A 2015 simulator study, for example, found that a THC blood level of 13.1 nanograms per milliliter had an impact on SDLP similar to a blood alcohol concentration of 0.08 percent, the current DUI threshold in every state. "It is widely acknowledged by researchers in the field that users of cannabis are generally aware of any possible drug-related impairments," White writes. "It is likely that cannabis users deploy their attention to the main safety-related driving tasks at the expense of keeping strictly in the centre of their lane."

You may or may not agree with White's conclusions, but his discussion of the methodological obstacles to measuring marijuana's impact on road safety clarifies why the subject remains unsettled and contentious. His subtle and sophisticated analysis is a welcome rejoinder to the simpleminded propaganda typified by RMHIDTA's reports.