Cannabis edibles impaired simulated driving below the legal blood limits

Most of what is known about cannabis and driving comes from smoked cannabis, and most of the laws written to police it use blood THC concentration as the test. A randomised crossover trial published in JAMA Network Open on 31 August tested edibles instead, and found the two facts sitting awkwardly together: the driving impairment appeared, the blood concentrations that would trigger enforcement mostly did not [s1].

The trial

The study was a randomised, double-blind, placebo-controlled crossover trial at a single tertiary care academic hospital in Toronto, Ontario, run between September 2024 and April 2025 [s1]. Participants were healthy adults aged 19 to 45 who reported at least weekly cannabis use and monthly use of edibles [s1].

Oral delta-9-THC was given in soft-chew gummies containing 0 or 2 mg each, to reach total doses of 0, 2, 10 and 20 mg [s1]. The primary outcome was standard deviation of lateral position — the lane-keeping measure that is the workhorse of driving-simulator research — assessed at two and five hours after exposure in the intention-to-treat population [s1]. Secondary outcomes were standard deviation of speed, reaction time, self-reported willingness to drive and perceived driving ability, and whole-blood THC and metabolite levels [s1].

Of 85 potential participants, 52 were enrolled as eligible, 10 withdrew before receiving THC and two were removed from analysis, leaving 40 — 20 female and 20 male, mean age 30.4 years (SD 6.9) [s1].

What it found

Relative to placebo, lateral position variability increased significantly at 20 mg (3.3 cm; 95% CI, 2.4 to 4.3; P < .001) and at 10 mg (2.0 cm; 95% CI, 1.0 to 2.9; P < .001), but not at 2 mg (0.7 cm; 95% CI, −0.3 to 1.6; P = .16) [s1].

Secondary driving measures moved in the same direction. Reaction time was higher at 20 mg (difference in means 0.034 s; 95% CI, 0.018 to 0.050; P < .001), and speed variability increased at 20 mg (0.469 km/h; 95% CI, 0.154 to 0.784; P = .004) [s1].

Participants noticed. Willingness to drive fell dose-dependently, significantly so at 20 mg (difference in means −2.113; 95% CI, 1.166–1.821; P < .001), as did participant-perceived driving ability (−1.694; 95% CI, 1.422–1.966; P < .001) [s1].

The measurement that matters for law enforcement is the last one. Whole-blood THC concentrations peaked at 3.4 ng/mL at the 20 mg dose and 1.6 ng/mL at 10 mg — near or below the per se thresholds commonly used for roadside enforcement, which the authors identify as 2 or 5 ng/mL [s1].

Why that is a policy problem, not just a finding

Per se limits work by substituting a measurable proxy for impairment, on the assumption that the two track each other. For alcohol they broadly do. For inhaled THC the relationship is already poor, because blood concentrations spike and fall much faster than the subjective and cognitive effects.

Edibles make the mismatch worse in a specific way: absorption is slower and blood levels are lower for a given psychoactive effect. This trial documents that gap directly. A 10 mg edible dose degraded lane-keeping to a statistically significant degree while producing a peak blood THC of 1.6 ng/mL [s1] — below both thresholds the authors name. A driver in that state would test clean under a 2 ng/mL rule and be measurably impaired in the simulator.

The authors' conclusion is exactly that: THC dose-dependently impaired simulated driving at blood THC concentrations below per se thresholds commonly used for roadside enforcement [s1].

The size of the effect, in context

The lateral-position effects here are consistent with the broader experimental literature. A 2022 systematic review and meta-analysis of 57 experimental driving studies covering 1,725 participants found cannabis alone associated with impaired lateral control — Hedges' g = 0.331 (95% CI, 0.212–0.451) for lateral position variability and g = 0.198 (95% CI, 0.001–0.395) for lane excursions — along with decreased driving speed (g = −0.176; 95% CI, −0.298 to −0.053) [s2].

That review also found that the effects of cannabis on driving performance were similar in magnitude to those of low blood alcohol concentrations, and that combining cannabis with alcohol was more detrimental than either alone [s2]. The new trial did not test the combination.

Limits worth stating

Forty participants is a small sample even for a crossover design, and the population was deliberately experienced: at least weekly cannabis use and monthly edible use [s1]. Tolerance cuts both ways here — it may understate impairment in naive users, or overstate it if experienced users compensate more.

Simulated driving is not driving. The measures used are validated proxies for on-road performance and are what regulators have to work with, but a 3.3 cm increase in lane-position variability is not a crash [s1].

And the trial tested a single commercial product format at four doses in one session [s1]. It does not establish a safe dose, a safe interval, or a threshold that should replace the ones now in law. What it establishes is that the current thresholds are measuring something other than what they are meant to measure, at least for edibles.

The trial is registered as NCT06595576 [s1].

Sources

• [s1] "Dose-Dependent Effects of Cannabis Edibles on Simulated Driving Performance: A Randomized Clinical Trial," JAMA Network Open, 31 August 2026. https://doi.org/10.1001/jamanetworkopen.2026.31306
• [s2] "The effects of cannabis and alcohol on driving performance and driver behaviour: a systematic review and meta-analysis," Addiction, 26 January 2022. https://doi.org/10.1111/add.15770