Can Cannabis Edibles and 0.05% Breath Alcohol Impair Driving When Field Sobriety Tests Are Negative? A 25-Person Trial

What was the precise question tested?

The investigators asked whether oral cannabis combined with alcohol produced different driving and behavioral outcomes from each substance administered separately. They also examined whether familiar field tests, a tablet-based driving-related assessment, and measured biological analytes tracked changes in simulated driving. The key distinction is between an experimental difference in performance and a conclusion about an actual motorist. Simulators recreate selected driving demands; they do not reproduce every hazard, social consequence, or compensatory response associated with driving on a public highway.

The sequence of administration matters. Earlier literature often involved smoked cannabis or alcohol administered first. Here participants swallowed an edible THC brownie and consumed alcohol approximately an hour later. This schedule puts delayed oral absorption alongside alcohol exposure at a defined interval. A result about this particular combination should not be presented as an equivalent measured effect for smoking, vaping, topical products, or unknown formulations.

Who were the 25 independent participants?

Sixty volunteers entered screening, 30 were randomized, and 25 completed the seven treatment sessions. The completer group included 15 men and 10 women, with a mean age of 25.6 years (SD 4.9). They were healthy adults with prior cannabis–alcohol co-use and a recent history of binge drinking; reported cannabis use averaged 0.6 (SD 0.7) times per week. Recruitment deliberately limited habitual cannabis frequency to reduce possible tolerance, which also limits extrapolation to people who use several times daily.

These were not 175 independent test subjects. The same 25 completers undertook seven experimental conditions, with at least a week between sessions. Multiple simulator drives, blood draws, breath measurements, or clue observations within a visit remain repeated observations of one person. The analysis therefore relies on within-participant contrasts rather than treating every visit or sample as a new human volunteer.

Study volunteers were medically screened and familiarized with the driving simulator. They completed practice assessments until performance stabilized under the protocol. That step reduced one source of learning effects but could not remove every learning, expectation, order, fatigue, or laboratory-context influence that might affect real-world generalizability.

How were THC and alcohol actually administered?

The double-blind, double-dummy crossover used brownies containing nominally 0, 10, or 25 milligrams of delta-9-tetrahydrocannabinol (THC). Brownies were prepared from cannabis material specified at 18% THC or a placebo preparation. Participants received a standardized low-fat breakfast before dosing. After swallowing the brownie, they began consuming three four-ounce drinks at 45 minutes, completing the drinks over roughly 15 minutes. Alcohol content was individualized to aim for a breath alcohol result of 0, 0.05%, or 0.08% BrAC, depending on condition.

The seven conditions were complete placebo; THC 10 mg alone; THC 25 mg alone; alcohol targeting BrAC 0.05% alone; THC 10 mg plus BrAC 0.05%; THC 25 mg plus BrAC 0.05%; and alcohol targeting BrAC 0.08% alone. There was no experimental arm combining 10 mg or 25 mg THC with a BrAC 0.08% target. A forensic reconstruction must not interpolate an untested eighth or ninth treatment as though it had been observed.

The dose in a brownie is an administered mass, not a measured blood concentration. Similarly, the 0.05% and 0.08% labels are breath-alcohol treatment targets, not automatically legally collected venous whole-blood ethanol values. The trial performed serial testing of actual BrAC; it did not report an independently sampled forensic whole-blood ethanol gas-chromatography result for each condition. An expert cannot silently exchange those matrices or claim an observed concentration exactly matched the nominal target throughout the day.

What did the simulator measure?

Driving was evaluated using a STISIM Drive M4000-R simulator. The reported global drive score (GDS) combined nine variables after normalizing participants’ changes to their own baseline performance. On this particular scale, a higher change means worse simulated driving performance. It is not a blood drug concentration, an observed crash, a probability of causing injury, or a recognized legal impairment percentage.

The researchers also evaluated standard deviation of lateral position (SDLP) during car-following and divided-attention tasks. Lateral-position changes are expressed in centimeters, unlike the GDS’s standardized scale. This distinction is critical: a difference of 1.0 in GDS does not mean the car moved one centimeter, and a change in one subtask cannot be assumed for every other driving subtask.

Simulated driving was repeated over a session extending to 7.5 hours after edible dosing. Consequently, the table’s peak change reflects a selected summary of repeated results, not the drug’s exact concentration or every driver’s value at a single universal minute. Time-to-peak, variability among people, and testing context remain relevant.

What were the seven observed peak driving and field-test results?

The following table is a compact forensic reorganization of the original measured experimental outcomes. GDS and field-test clues are separate instruments; the numbers show the mean and standard deviation of peak change within the study, not the number of impaired motorists, the number of crashes, or a legal diagnostic accuracy estimate. Because the same people completed all arms, columns are paired repeated observations rather than independent cohorts.

The strongest mean GDS change appeared in the high-dose edible plus 0.05% BrAC condition. This supports an experimentally observed interaction in functional burden for the tested regimen, but it does not permit the simple claim that every combined-dose individual was more impaired than every participant under alcohol alone. The standard deviations and paired analyses are essential to the inference.

Can 25 mg THC plus 0.05% BrAC exceed alcohol alone at 0.08%?

In the measured GDS outcome, yes: the mean peak change was 2.5 (SD 1.7) for 25 mg THC plus alcohol targeting 0.05% BrAC, versus 1.6 (SD 1.6) for alcohol alone targeting 0.08%, with a planned comparison of p = .02. That is a statistically significant group-level simulator contrast in the original experiment. It cannot be converted into an across-jurisdiction statement that a particular co-using defendant necessarily satisfied a criminal standard, nor into an exact multiplier of collision risk.

The lower-dose edible with alcohol targeting 0.05% yielded mean peak GDS 1.6 (SD 1.4), numerically similar to the 0.08%-only mean 1.6 (SD 1.6). “Numerically similar” should not be upgraded to a formal equivalence claim without a prospectively specified equivalence margin and appropriate analysis. The combined-condition effects must be interpreted using the authors’ actual planned contrasts and testing strategy.

By comparison, 25 mg THC alone had GDS 1.5 (SD 1.3), whereas 10 mg alone had GDS 0.7 (SD 1.1). The 10 mg-alone comparison with placebo missed the conventional significance threshold (p = .06); it therefore is not a proven null effect or proof the dose can be safely used before driving. The experiment had only 25 completers, and a nonsignificant result is not a certification of absence of impairment.

Did the individual lane-position tasks follow the global score?

Not consistently. During car-following, the mean peak SDLP change was 2.2 cm (SD 6.0) under placebo, 8.9 cm (SD 7.6) after 25 mg THC alone, 6.3 cm (SD 8.3) after 10 mg THC plus 0.05% BrAC, and 9.3 cm (SD 11.8) after 25 mg THC plus 0.05% BrAC. The original table identified those three active comparisons as significantly different from placebo. These are observed simulator-lane outcomes, not estimates of distance traveled while chemically impaired on a public highway.

The divided-attention SDLP results were less consistent. Although the analysis found overall condition and timing effects, the co-use conditions did not show statistically significant divided-attention SDLP differences from placebo in the reported planned contrasts. The high-dose combination’s divided-attention peak was 8.3 cm (SD 17.4), illustrating substantial interindividual variability. A global composite can be significant while one constituent task is not. Selecting only the significant lane-position result—or only the nonsignificant one—would distort the tested endpoints.

This matters in litigation because a “driving impairment” phrase may conceal the variable actually assessed. The first questions should be which task, which reference condition, which minute, which direction of change, and which prespecified statistical comparison supports the claim.

Did standardized field sobriety testing detect the same changes?

Often it did not. Trained examiners used horizontal gaze nystagmus, walk-and-turn, and one-leg stand elements and scored the study’s total clues. The field examination samples performance under instructions, coordination and oculomotor cues; it does not reproduce steering, divided attention, hazard response or lane control inside a moving vehicle. Field clues, a driving composite and a chemical result therefore answer materially different questions.

In particular, the 10 mg THC-plus-0.05% BrAC arm had a mean peak clue change of 0.8 (SD 2.2) versus 0.2 (SD 1.3) for placebo, with a planned comparison reported as p > .99, not statistically significant. Its mean GDS change was nevertheless 1.6 (SD 1.4), significantly worse than placebo. This pairing is direct evidence of a study-level discordance between measured tests, not a numerical estimate of field-test sensitivity for all cannabis users.

The limitation cuts in both directions. SFSTs did show a statistically significant result under alcohol 0.08% alone: clue changes 2.2 (SD 2.2) versus 0.2 (SD 1.3) under placebo (p = .008). The 25 mg THC-plus-0.05% condition also differed in clues from alcohol 0.05% alone (2.1 versus 0.0, p = .04). It would therefore be incorrect to announce that all field-test comparisons in the study were negative. The conclusion is narrower: certain simulator changes were not reliably mirrored by the study’s field-test clue totals.

What about the DRUID tablet-based assessment?

The investigators separately administered DRUID, a digital cognitive and sensorimotor test designed to capture aspects of drug effects. Its measured score is not another name for GDS or for horizontal gaze nystagmus. The study found significant placebo contrasts under 10 mg THC alone, 25 mg THC alone and alcohol 0.08% alone, but not robust statistically significant DRUID contrasts for the combined edible-and-0.05%-alcohol conditions. The high-dose combination’s reported peak change was 2.6 (SD 20.5), against placebo −2.1 (SD 6.8), showing pronounced variability.

That is a warning against a single surrogate “impairment meter.” A digital test may detect some laboratory drug-related changes and miss others, depending on task content, drug timing and noise. Neither the absence of a DRUID difference nor a negative field-test finding establishes chemical abstinence, safe actual driving or lack of possible psychomotor risk in a particular person.

How were THC and alcohol measured in biological specimens?

For alcohol, the protocol measured exhaled breath alcohol concentration to guide and assess the assigned drinking condition. BrAC is not automatically interchangeable with a forensic venous whole-blood ethanol result. Conversions can depend on legally prescribed assumptions and measurement context; this paper did not demonstrate a case-specific criminal-laboratory ethanol result from sampled whole blood.

For cannabinoids, researchers collected venous blood and assayed whole-blood parent Δ9-THC and the distinct metabolites 11-hydroxy-THC and 11-nor-9-carboxy-THC using liquid chromatography–tandem mass spectrometry. These measurements are not serum or plasma, and 11-COOH-THC is not parent psychoactive THC. An administered edible quantity, an observed concentration and a functionally observed driving score must remain separate categories of evidence. The site’s published guide to biological matrices explains why identifying the sampled compartment and specific analyte is indispensable.

After the 25 mg THC edible without alcohol, mean whole-blood parent THC maximum concentration was 3.21 ng/mL, with an observed range of 1.34–6.70 ng/mL, even as the group showed substantial GDS worsening. A mean peak concentration cannot be paired with every individual’s peak driving score as though they necessarily occurred simultaneously. Nor does it validate a fixed whole-blood THC cutpoint for diagnosing impaired driving in every person, particularly across routes and histories of prior cannabis use.

Was there a pharmacokinetic interaction between edibles and alcohol?

The researchers did not find a statistically significant effect of adding alcohol on the examined whole-blood THC and metabolite concentration maxima or exposure measures, nor a significant effect of adding cannabis on the measured BrAC outcomes. This is one of the paper’s important negative findings. The stronger simulator decrement under co-use therefore cannot be explained simply by asserting that alcohol raised everyone’s assayed THC concentration. The study’s functional differences can arise through combined pharmacodynamic actions without a detectable mean change in blood pharmacokinetics.

That null result remains bounded by the particular dose sequence, sample size and sampling schedule. It does not prove that pharmacokinetic interactions are impossible for different products, higher doses, drinking patterns or populations. Likewise, a whole-blood THC curve cannot reconstruct an individual’s exact last ingestion time without an independently validated model and reliable case-level dosing history. See the site’s published pharmacokinetics interpretation guide for the distinction between observed concentration-time data and individualized back-calculation.

Which planned contrasts were significant, and which were not?

Results should be described by endpoint and tested contrast, not by a blanket “impaired” or “unimpaired” label. The second evidence table selects especially probative primary results and distinguishes statistically significant differences from nonsignificant findings. A nonsignificant test is a constraint on what was demonstrated; it is not positive evidence of clinical equivalence. For clarity, “GDS” means the standardized simulator driving composite, while “SFST” means the clue count from performance tests. Different rows do not share a common physical unit.

The authors used repeated-measures statistical methods and multiplicity adjustment for planned contrasts. These controls reduce some false-positive risks but do not confer unlimited power on a 25-completer study. Other possible contrasts and endpoints were not all tested as formal equivalence hypotheses. Selecting an isolated small p value while concealing corresponding negative results is not a faithful scientific appraisal.

What did participants report about intoxication and confidence to drive?

Reported drug effects and willingness to drive are important but subjective. The mean peak decrease in confidence to drive was −14.2 (SD 19.1) under placebo, −60.0 (SD 30.8) after 25 mg THC alone, −44.8 (SD 31.5) under 0.05% BrAC alone, and −63.0 (SD 24.3) after 25 mg THC plus 0.05% BrAC. The comparison between the latter combination and alcohol 0.05% alone was significant (p = .04). This shows participants perceived a difference; it does not demonstrate an individual can accurately self-assess every functional deficit.

Conversely, the 10 mg THC-alone condition did not significantly reduce confidence relative to placebo under the cited comparison. Feeling capable of driving is not identical to performing well in the simulator, and a weak self-reported “high” does not negate objective task changes. A forensic evaluator should identify whether the record concerns perceived intoxication, volunteered willingness to drive, objective performance, or chemical detection rather than presenting those measures as one scale.

When did the driving-score effects occur and resolve?

Group-level GDS changes were visible approximately 1.5 hours after swallowing the brownie and generally greatest around 3.5 hours, which was roughly 2.5 hours after completion of alcohol ingestion. Under the study’s repeated testing, group means approached baseline by about 7.5 hours after the edible. Those are features of these regimens and observation windows, not a scientifically validated “safe-to-drive-after-X-hours” instruction.

Individual absorption, product formulation, meal composition, previous use, concomitant medication, repeated dosing and personal functional sensitivity may shift the time course. A retrograde reconstruction from one cannabinoid concentration cannot automatically tell whether the driving-score peak happened at a specified moment. The timing of the blood draw, breath examination, actual ingestion and alleged operation of a vehicle must be documented independently.

What limitations matter most in a DUI or collision case?

First, 25 participants completed a carefully structured laboratory crossover, fewer than the 32 initially targeted in the planning assumptions. They were screened healthy young adults with specified prior cannabis and alcohol experiences. The experiment was not an epidemiological survey of drivers and did not include actual roadway driving, officer stops, crashes, injuries or retrospective liability adjudication. It cannot provide an individual crash-causation probability or prove whether a particular driver could avoid a pedestrian or respond to an emergency.

Second, the treatments were precisely selected oral THC servings and alcohol protocols, not random commercial edibles or unknown drinking sequences. The 0.08%-alone condition served as a comparator, but no edible-plus-0.08% arm was studied. The later-testing time course should not be generalized to every formulation or multiple-edible use. Even careful simulator practice cannot guarantee that laboratory driving scores transfer identically to public-road situations.

Third, evidence types must not be merged. BrAC is a breath measurement; THC values reported here came from venous whole blood; THC-COOH is a metabolite; field-test clues assess standardized performance components; DRUID is a digital proxy; GDS is a simulator score; and subjective confidence is a report by the volunteer. The published specimen collection guide outlines why method, matrix and sampling times matter when a report is used in a legal setting.

Finally, the safety observations were limited: transient dizziness or headache occurred in four participants in the 25 mg THC-plus-0.05% condition, three in the 25 mg-alone condition, and one in the 0.08%-alcohol-alone condition; no serious adverse event occurred during the study. That safety description is not a claim that the tested combinations are medically safe or appropriate for driving.

What is an appropriate expert use of these findings?

A defensible expert may explain that oral THC and lower-level breath alcohol jointly changed simulated driving performance under the tested conditions, with a particularly notable GDS difference relative to alcohol 0.08% alone. The expert may also point out that certain SFST and DRUID contrasts failed to detect statistically significant changes even where simulator impairment was observed. These findings help frame what a particular test can—and cannot—detect. They do not establish a universal field-test false-negative rate, an individualized blood-THC impairment cutoff, or the cause of a particular accident.

Before applying the research to an incident, review the actual dosing history if reliable, blood and breath collection times, whether an assay measured whole blood or a separated fraction, cannabinoid identification and reporting limits, measurement uncertainty, all field-test and observed driving evidence, participant health and concomitant exposures, and the governing law at the relevant date. A persuasive scientific analysis identifies disagreement among independent kinds of evidence rather than choosing whichever result favors one side while concealing limitations.

The bottom line is neither “positive THC proves impairment” nor “negative field sobriety proves safe driving.” In this carefully controlled experiment, some oral-THC-plus-alcohol conditions altered simulated driving more clearly than certain screening and performance proxies reflected. That is a bounded, experimentally observed result with genuine forensic value precisely because its limits are preserved.

Seven frequently asked forensic questions

1. Did this study show that a 0.05% breath-alcohol result alone is equivalent to 0.08% with an edible?

No. The study showed numerical similarity for particular simulator outcomes in selected combined conditions, and a significant worse-GDS contrast for 25 mg THC plus 0.05% versus 0.08% alcohol alone. This was not a formal clinical equivalence study or a universally applicable legal threshold comparison.

2. Were field sobriety tests always negative after edibles and alcohol?

No. Several planned comparisons were nonsignificant, including the 10 mg THC-plus-0.05% versus placebo clue contrast, but the 25 mg THC-plus-0.05% condition had more clues than 0.05% alcohol alone (p = .04). One cannot turn selective nonsignificance into blanket test failure.

3. Did the researchers measure actual whole-blood THC or just plasma THC?

They directly assayed venous whole-blood THC and distinct metabolites. The alcohol outcome used exhaled-breath concentration. A blood cannabinoid assay and a BrAC device measure different substances in different specimens.

4. Does a whole-blood THC concentration below 5 ng/mL rule out impaired simulator driving?

No. After 25 mg THC alone, the reported mean maximum whole-blood THC was 3.21 ng/mL while the group showed a significant simulator decrement. The trial did not validate any person-level 5 ng/mL classifier or directly decide a jurisdiction’s concentration statute.

5. Did adding alcohol increase measured whole-blood THC exposure?

The analyzed group-level pharmacokinetic comparisons did not find a significant co-use increase in the tested cannabinoid exposure measures. That does not mean the functional effects could not combine, or that all alternative alcohol–THC sequences have been excluded.

6. Is a DRUID score the same as a driving simulator result?

No. DRUID samples particular digital cognitive and motor tasks; GDS combines multiple simulated driving metrics. The two can produce different patterns of statistical significance within the same experimental conditions.

7. Can the trial determine whether an identified motorist caused a crash?

No. The participants were laboratory volunteers, and no real crashes or case reconstructions were observed. A specific incident requires contemporaneous driving facts, valid toxicology, alternative explanations, and relevant legal standards in addition to what controlled experiments establish about group-level effects.