Cannabis and driving: what the evidence and Irish law actually say

The most-cited number is from a 2012 meta-analysis by Asbridge, Hayden and Cartwright, published in the BMJ. They pooled nine observational studies covering 49,411 participants and found that acute cannabis use before driving nearly doubled the odds of a collision: a pooled odds ratio of 1.92, with a 95% confidence interval running from 1.35 to 2.73. The effect was larger, not smaller, in the higher-quality designs. Case-control studies gave an odds ratio of 2.79 (95% CI 1.23 to 6.33) and studies of fatal collisions gave 2.10 (95% CI 1.31 to 3.36), against 1.65 (95% CI 1.11 to 2.46) for culpability studies. Their own conclusion was that acute cannabis is associated with roughly a doubling of crash risk, in their words "especially for fatal collisions."

That figure has been challenged, and the challenge is worth taking seriously rather than hiding. In 2016 Rogeberg and Elvik published a more conservative re-analysis in Addiction. Their argument was that Asbridge, and a parallel meta-analysis by Li in 2012, had overstated the risk by pooling unadjusted odds ratios rather than ones adjusted for confounders such as age, sex and alcohol. Re-analysing 28 estimates from 21 studies, their updated pooled figure was much lower: a random-effects odds ratio of 1.36 (95% CI 1.15 to 1.61), and a meta-regression sensitivity model that came in lower still at 1.22 (95% CI 1.10 to 1.36). They described the effect as "a statistically significant risk increase of low-to-moderate magnitude," roughly a 20 to 36 per cent increase in odds rather than a doubling. This 1.2 to 1.4 band is the best-supported modern point estimate for cannabis on its own.

Why the gap between roughly 1.9 and roughly 1.3? It turns on whether you pool the raw odds ratios or the adjusted ones, and how you weight the studies. This is a genuine live methodological dispute, not settled fact. Later work, including White and Burns in 2021 and Rogeberg's own published exchanges, keeps landing in the low-to-moderate range without fully resolving it. Our reading is that the conservative, adjusted 1.2 to 1.4 is the more defensible figure for cannabis alone, while Asbridge's roughly 1.9 remains the most-cited number and may better capture the serious and fatal end of outcomes. Either way, both point the same direction: up.

The reason adjustment matters so much shows up starkly in a large US case-control study. Compton and Berning, in a 2015 NHTSA research note (the Virginia Beach "Drug and Alcohol Crash Risk" study), found a raw, unadjusted crash odds ratio for THC-positive drivers of 1.25, which was significantly elevated. But once they controlled for age, gender, race and ethnicity, and above all for alcohol, that elevation largely disappeared: the adjusted THC odds ratio fell to 1.00 (95% CI 0.83 to 1.22), no significant increase in crash involvement. Their finding was that after adjusting for demographics and alcohol, "there was no indication that marijuana use... [was] associated with an elevated crash risk." The paper attributes most of the raw signal to confounding: THC users are disproportionately young men who also drink. Alcohol, by contrast, showed a strong, dose-dependent effect that survived every adjustment, climbing steeply with blood-alcohol concentration.

One combination is not ambiguous at all. Cannabis plus alcohol together is substantially worse than either alone. Chihuri, Li and Chen, in a 2017 US case-control study (1,944 fatally injured drivers against 7,719 roadside-survey controls), reported adjusted odds ratios for fatal-crash involvement of 1.54 for marijuana alone (95% CI 1.16 to 2.03), 16.33 for alcohol alone (95% CI 14.23 to 18.75), and 25.09 for both together (95% CI 17.97 to 35.03). That combined figure exceeds the sum of the individual effects, a positive interaction: in their words, "the combined effect of marijuana and alcohol... was greater than the sum of the net effects." Controlled experiments echo this, with alcohol and THC together producing significantly more lane-weaving than either substance on its own.

In 2020 Arkell and colleagues published in JAMA a double-blind, placebo-controlled crossover trial with a genuine 100-kilometre on-road driving test conducted on a public highway. Twenty-six occasional cannabis users (mean age 23.2, sixteen women, of whom 22 completed all eight tests) drove after vaporising THC-dominant cannabis, a THC/CBD-equivalent blend, CBD-dominant cannabis, or placebo. The primary outcome was the standard deviation of lateral position, or SDLP, the standard laboratory measure of lane weaving.

At 40 to 100 minutes after use, lane weaving was significantly worse than placebo for THC-dominant cannabis, by +2.33 cm (95% CI 0.80 to 3.86; P<.001), and for the THC/CBD blend, by +2.83 cm (95% CI 1.28 to 4.39; P<.001). CBD-dominant cannabis produced no such effect (−0.05 cm; 95% CI −1.49 to 1.39; P>.99), behaving just like placebo. To put the size in context, the paper notes that a 2.4 cm rise in SDLP is what you would typically see from a driver at a blood-alcohol concentration of 0.05 per cent. So the THC impairment at 40 to 100 minutes was modest and comparable to a driver just at the drink-driving limit, not catastrophic, but real.

Here is the single most practical point, and it is where the science and the law meet. How long you are impaired, how long you are detectable, and how long you are legally exposed are three different lengths of time. They are not the same, and in frequent users they can be wildly different.

The impairment window is short. From inhaled cannabis, behavioural driving impairment peaks in the first hour and largely resolves within about three to four hours, which is consistent with Arkell's null result at four to five. That is the window in which you actually feel it and actually drive worse.

The detection window is longer, sometimes far longer. THC can remain measurable in blood and oral fluid well after the impairment has gone. In occasional users that gap is a matter of hours; in frequent or chronic users it is extreme. Bergamaschi and colleagues (2013, Clinical Chemistry) monitored 30 chronic daily smokers in a locked ward and found THC still detectable in blood in some subjects after up to about 30 days of continuous abstinence. They warned explicitly that a positive blood THC "even at concentrations >2 ng/mL" does not prove recent use or impairment in a frequent user. Karschner and colleagues (2009) similarly found detectable blood THC after about seven days of abstinence in heavy users. A later systematic review confirmed the pattern: frequent users carry residual THC without any matching acute impairment.

The legal-exposure window in Ireland is set by a specified blood concentration, not by whether you feel impaired. Because the law is presence-based, you are exposed for as long as your blood THC sits at or above the limit, which can outlast the impairment by hours in an occasional user and by days in a frequent one. And the arithmetic can run the other way too. THC leaves the blood faster than the brain effect resolves, so impairment can be present after blood THC has already dropped. No single biomarker cleanly tracks "currently impaired."

The presence-based offence and the exact THC figure

Ireland has a statutory presence-based offence with a specified blood concentration for THC. It was created by section 8 of the Road Traffic Act 2016, which inserted a new subsection 4(1A) and a new Schedule into the Road Traffic Act 2010. The offence is that a person shall not drive, or attempt to drive, a mechanically propelled vehicle in a public place while there is present in the body a quantity of a specified drug such that, within 3 hours after so driving, the concentration of that drug in the blood is equal to or greater than the level set out in the Schedule. Note the precise wording: the statute does not require the drug to be measured at the roadside; the offence is committed if, within three hours of driving, the blood concentration is at or above the level.

So the widely quoted 1 ng/ml THC figure is correct, and there is a separate 5 ng/ml limit for the inactive carboxy-THC metabolite. Either one at or above its level is enough. This per-se offence applies only to those three drugs, cannabis, cocaine and heroin; every other drug is dealt with under the impairment offence below. One point of accuracy: the Schedule's "heroin" entry is actually 6-acetylmorphine, a specific short-lived heroin metabolite, not general morphine or codeine.

How the roadside test works

Gardai use a preliminary oral-fluid (saliva) test, either at a Mandatory Intoxicant Testing checkpoint or at a Garda station, using a swab device. Collection takes about a minute and the result comes in about eight minutes. The device screens for cannabis, cocaine, opiates, benzodiazepines and, on the current device, amphetamines including methamphetamine and MDMA. If the oral-fluid test is positive for cannabis or cocaine, the driver is arrested and taken to a Garda station, where an evidential blood specimen is taken and sent to the Medical Bureau of Road Safety for analysis; for these two drugs the Garda does not need to prove impairment. If the test is positive for benzodiazepines or opiates, arrest follows only if the Garda also forms the opinion that the driver is impaired. Refusing to provide an oral-fluid sample is itself an offence, punishable on conviction by a fine of up to €5,000 or up to six months' imprisonment, or both.

It is the MBRS blood certificate, showing a result at or above the Schedule level, that grounds the per-se prosecution, not the roadside swab. The MBRS is the national forensic laboratory for this work, ISO 17025 accredited, and it analyses the evidential blood using chromatography with tandem mass spectrometry. Drug analysis typically takes around three months, and can take up to five.

The separate impairment offence, which needs no limit

Independent of the per-se limits, section 4(1) of the Road Traffic Act 2010 makes it an offence to "drive or attempt to drive a mechanically propelled vehicle in a public place while he or she is under the influence of an intoxicant to such an extent as to be incapable of having proper control of the vehicle." "Intoxicant" here includes any drug, illicit, prescription or over-the-counter, or alcohol, or a combination. This route requires proof of impairment but no threshold concentration. That means a driver can be convicted even below 1 ng/ml THC, or for a drug that is not on the Schedule at all, if actually impaired. Gardai can also carry out a roadside impairment (field) test. Citizens Information confirms that it is illegal to drive under the influence of drugs, including prescription drugs, where driving is impaired to the extent of not having proper control.

The penalties

Both offences are contraventions of section 4, so the same criminal penalty applies. Section 4(5) provides that a person is "liable on summary conviction to a fine not exceeding €5,000 or to imprisonment for a term not exceeding 6 months or to both." That is a statutory maximum set by the court, not a fixed penalty. The disqualification periods, however, differ by offence.

So for a first per-se cannabis conviction you are looking at a fine of up to €5,000 and/or up to six months' imprisonment, plus a minimum one-year driving ban, without the prosecution needing to prove you were impaired at all. A prior drink-driving offence counts as a prior for the "second or subsequent" escalation. The disqualification minima here are taken from Citizens Information and are consistent with the Road Traffic Act 2010 as amended; if you need a black-letter citation for a court, verify the ban lengths against the disqualification provisions of the Act as amended.

Prescribed medicine and medical cannabis

There is a specific statutory defence, but it is narrow and applies only to the per-se cannabis offence. Section 4(1B) of the Road Traffic Act 2010, inserted by the 2016 Act, provides that the per-se offence does not apply in respect of the drug at reference number 1 or 2 in the Schedule, that is THC and its carboxy metabolite and not cocaine or heroin, where the person holds a certificate in the prescribed form, a "medical exemption certificate," indicating that at the relevant time THC had been lawfully prescribed for them, signed by the prescribing doctor. The prescribed form is in force under S.I. No. 158/2017. Knowingly signing a false certificate is itself an offence.

Everything above is individual-level evidence: case-control, culpability and experimental studies of drivers. State-level or ecological work asks something else entirely, whether legalising cannabis changes a region's overall crash or fatality rate. Fowles and Loeb (2021), in the Journal of Transport & Health, is one example: a panel of the US states across 2010 to 2016, stress-tested against many model specifications using Bayesian sturdy-values. Two things in it matter for us, and they pull in different directions. Their recreational-legalisation dummy on its own was neither statistically significant nor sturdy, so simply changing the law showed no robust, detectable effect on the state death rate. But their marijuana-use variable is a different story: with controls in place, Fowles and Loeb report a positive, harmful association and conclude, in their own words, that there is a "life-taking relationship between marijuana use and vehicle crashes." They reach that conclusion by adopting an optimistic statistical prior; under wider priors the marijuana coefficient becomes fragile, while the seat-belt variable stays sturdy across every prior. So the paper does not exonerate cannabis, and we are not citing it to. We use it for one narrow purpose: to show how far a population-level result can swing with the analyst's choices, and how a raw correlation can point the wrong way.

So the honest summary is this: population-level legalisation effects are genuinely complex and contested. Fowles and Loeb's own conclusion is that marijuana use carries a harmful, "life-taking" association with fatality rates; what their analysis also shows is that this conclusion is prior-dependent, that the legalisation dummy itself is not robust, and that sturdier fundamentals such as seat-belt use hold up across every specification. Other ecological studies land in different places again, from a tripling of THC-related fatalities after legalisation in Hawaii to apparent substitution effects pointing the other way. None of it is, or can be, evidence about your individual risk at the wheel, which the individual-level studies settle in one direction: up. We have written a separate statistics-literacy guide dedicated to that ecological paper and to why the boring, well-measured fundamentals in it prove sturdier than the frightening headline variable; if you want the population-level story in depth, read it there rather than here.