Does Blood THC Predict Lane Weaving the Morning After Cannabis? A 130-Driver Simulator Study

Short answer: In a 2026 observational study of 130 licensed adults, including 65 frequent cannabis smokers tested approximately 12–15 hours after their last smoking session and 65 controls, laboratory-measured blood THC and oral-fluid THC did not show a statistically significant association with the simulator driving measures after correction for multiple testing. Nor did the two groups differ significantly on the study’s adjusted lane-position measures after that correction. These findings restrict what the study can support about the morning after; they do not demonstrate that a particular driver is unimpaired, that cannabis never affects driving, or that a THC concentration can resolve collision causation.

The primary study is Zakala and colleagues, Driving by frequent cannabis users ‘the morning after’ last use of smoked cannabis: an observational driving simulator study (2026), DOI: 10.1186/s42238-026-00416-w. Its distinctive contribution is the simultaneous examination of directly collected blood and oral-fluid cannabinoids alongside measured simulator lane control and related tasks at a defined interval after self-selected smoking. A same-cohort cognitive report addresses different tests; the evidence below concerns driving-simulator outcomes.

What question did the investigators test?

Many forensic disputes collapse three different questions into one: whether a specimen contains THC; whether someone performs differently on a specified driving task; and whether drug effects caused a particular driving event. A chemical result is not itself a lane-position measurement, and a simulator score does not reconstruct a traffic collision. This investigation asked whether frequent smokers, tested the morning after use, displayed different simulator performance from controls and whether drug concentrations in blood or oral fluid related to that performance.

Its design matters. Participants were not randomized to receive cannabis or placebo. Rather, one group already used smoked cannabis at least four days each week and the other reported no cannabis use during the prior month. The smokers purchased a legal pre-roll, chose their own product, smoked it at home, and were assessed the following morning. Such a comparison can describe group differences under observed circumstances; it cannot isolate cannabis exposure from every characteristic associated with frequent use.

Who was studied, and when?

The analysis included 65 frequent cannabis smokers and 65 controls who completed procedures: 130 distinct individuals, not 130 drug administrations and not 130 repeated assessments per condition. Each group included 33 men and 32 women. The investigators began with 201 consenting participants; screening exclusions, withdrawals, and losses occurred before the final groups were formed. Eligible participants were licensed drivers aged 19–45 years. Smokers were asked to use their selected pre-roll roughly 12–15 hours before the test session; their reported mean interval was approximately 13 hours.

The groups were reasonably comparable on age, sex, reported driving experience, driving exposure, and sleep, but race and years of education differed. Those imbalances matter because a between-person difference in driving tasks may have more than one plausible explanation. The study did not supply a sober, within-person pre-smoking simulator baseline for every cannabis participant, and there was no randomized exposure allocation or counterbalanced placebo driving day.

Participants practiced the simulator at screening. On the test day they completed standard and distracting-task drives, braking and following-distance tasks, specimen collection, and subjective ratings. A breath-alcohol eligibility screen was used to exclude acute alcohol consumption, not to derive a forensic venous blood-alcohol concentration. Measurements were obtained in a research environment, not in an instrumented reconstruction of any defendant’s vehicle or actual collision.

How was driving measured rather than inferred?

The study used a Virage VS500M driving simulator. A key endpoint was standard deviation of lateral position (SDLP): how much the simulated vehicle varied laterally while traveling along the roadway. The investigators also ran a separate divided-attention condition in which participants counted backward by threes from 900 while driving. They examined speed, speed variability, following distance, and brake-response latency. Each of these is an observed task output—not a measured THC concentration or a standalone legal finding of impairment.

A larger SDLP generally indicates more lateral variability in that standardized task, but the number has meaning only in the context of the route, simulator, instructions, and comparison group. It cannot be converted into an individual probability of a real-world collision from this study. A lower SDLP can coexist with other deficits, compensatory driving, different task priorities, or unmeasured risk. The research’s multiple endpoints are therefore not interchangeable measures of one person’s overall driving ability.

Which physical specimens actually contained measured cannabinoids?

The researchers collected blood in lavender-top tubes and oral fluid at the time of the drive, then arranged laboratory analysis for parent THC, its psychoactive metabolite 11-hydroxy-THC (11-OH-THC), inactive metabolite THC-COOH, and CBD. The report distinguishes blood from oral fluid and lists concentrations in ng/mL. It does not establish a universal breath-to-blood conversion, nor should the blood numbers be silently re-labeled as venous whole-blood gas-chromatography results in a separate forensic case without checking that case’s actual laboratory method and specimen.

In the smoker group, mean reported blood THC was 2.8 ng/mL (standard deviation 4), while the median was 1.2 ng/mL (interquartile range 0.5–2.7). The distribution was heterogeneous, with measured values spanning approximately 0.1–18 ng/mL in the study table. Mean oral-fluid THC was 31 ng/mL (standard deviation 75), with a broad 0.1–399 range. These very different distributions arise in distinct matrices; 31 ng/mL oral-fluid THC is not equivalent to 31 ng/mL blood THC or a THC dose reaching the brain.

For results below the study’s 0.2-ng/mL quantification limit, the investigators assigned 0.1 ng/mL for some quantitative analyses. Such substitution matters when interpreting a low measured average or apparent control value. The analytic floor is not evidence of a chemical-free specimen. Inactive THC-COOH in blood is different from pharmacologically active parent THC and is not, by itself, proof of contemporaneous driving impairment.

The study’s experimental design and chemical measurements

The figures are original research-group summaries, not predicted values for a hypothetical driver. A participant’s THC result, product label, and route timing would still require identification, collection history, laboratory validation, and uncertainty analysis in an actual forensic matter.

Did the cannabis group weave more in the simulator?

Not in the study’s adjusted group means. In the ordinary lane-control condition, the cannabis group’s mean SDLP was 0.280 m (standard error 0.008), compared with 0.305 m (0.008) among controls. In the distracted condition, the respective adjusted means were 0.256 m (0.006) and 0.272 m (0.006). The cannabis group therefore had numerically less lateral variability in both comparisons. Before multiple-comparison adjustment, the group contrasts had p values of .031 and .046. After the specified correction, both adjusted p values were .277—not statistically significant at the conventional .05 threshold.

This is not evidence that cannabis improved anyone’s driving. Nonrandom allocation, between-person variation, possible compensatory behavior, the limited simulator environments, and the multiple-comparison correction all constrain the inference. Conversely, failure to reject a group null does not prove individual equality or that no participant had a residual deficit. It means these investigators did not demonstrate a statistically robust group effect on these specific simulator outputs under these conditions.

Source-based comparison of selected simulator and self-report outcomes

Adjusted group means and multiplicity-adjusted p values are reproduced from the original article’s Table 2; SE means standard error, not between-participant standard deviation. An adjusted p value of .277 is not interchangeable with an unadjusted p value of .031 or .046.

What did the other driving tasks show?

Mean speed, speed variability, maximum speed, following distance, and brake latency did not exhibit statistically significant cannabis-versus-control differences after adjustment for multiple comparisons. Brake latency was numerically shorter in the cannabis group (0.783 versus 0.819 seconds), but its unadjusted p value of .072 and adjusted p of .287 do not substantiate a group difference. These measures also probe distinct demands: holding speed, reacting to a stop sign, and maintaining distance from a lead car are not synonyms for lateral lane control.

The study also compared smokers who had chosen infused pre-rolls with those who had not. Seventeen cannabis participants used products identified as infused. The subgroup analysis did not identify significant simulator differences attributable to that distinction. Because individuals selected commercial products rather than receiving randomized assigned formulations, this subgroup result does not establish dose-response equivalence or the safety of high-potency products.

Collision counts within the research simulator were not adequate to estimate real crash-risk sensitivity, specificity, or an individual likelihood of collision. A simulator collision and a road crash have different exposure denominators and circumstances. None of the standard, dual-task, braking, or following-distance outcomes justifies a numerical prediction about a particular roadway incident without additional independent evidence.

Did the blood THC concentration predict a driver’s SDLP?

After the authors accounted for multiple comparisons, their analyses found no statistically significant correlation between blood THC or blood THC-COOH and the tested simulator driving measures. Likewise, the reported oral-fluid THC associations did not remain significant following correction. That finding addresses these particular observations, concentrations, timing, and population; it does not validate a universal concentration cutoff, prove a chemical assay irrelevant, or show that a driver with an elevated result must be unaffected.

Some exploratory unadjusted correlations involved blood 11-OH-THC. For example, the original paper reports SDLP association r = 0.294 with nominal p = .018, but the adjusted p was .071, above .05. A reported oral-fluid THC versus brake-latency correlation was r = .286, nominal p = .021, adjusted p = .136. Presenting only the nominal p values would materially change what the complete source supports. These are neither confirmed biomarkers of impairment nor proof that the true population association is exactly zero.

The mean blood THC concentration of 2.8 ng/mL also cannot establish how many persons had a valid legally relevant result at any particular driving time. The original distribution includes values below and above that mean, and the study used imputation below quantification. Jurisdictions may use distinct statutory schemes, specimen definitions, and laboratory requirements. Whether a result triggers a statute and whether the drug caused an observed driving deficit are separate questions; this experiment cannot resolve either for an individual defendant.

Why did people still report feeling high?

Although the simulator measures did not differ significantly after correction, the cannabis group continued to report stronger subjective cannabis-related sensations. The adjusted mean for “I feel high” was approximately 12.75 versus 0.64 in controls, with a multiplicity-adjusted p value of .001. The ratings “This feels like cannabis” and “I like this drug effect” also remained significantly different. That coexistence is informative: a subjective report and a task score are not necessarily aligned, and neither can simply be substituted for the other.

Someone may describe residual drug effects without committing a measurable simulator error. Someone else may show concerning real-world behavior while minimizing or inaccurately estimating their own intoxication. The study provides a group contrast of self-reports at approximately 13 hours. It does not test truthfulness in a particular interview, validate credibility scoring, or show that feeling high satisfies or defeats a jurisdiction’s impairment standard.

Why does correction for multiple comparisons matter?

A study that examines many related driving outcomes and numerous cannabinoid correlations has many opportunities to obtain a nominal p below .05 by chance. The researchers adjusted for that multiplicity. It is therefore misleading to elevate the two unadjusted SDLP p values or exploratory correlations while omitting the corresponding adjusted results. The original paper’s adjustment moves both SDLP comparisons to p = .277, while the higher subjective “I feel high” rating remains significant at adjusted p = .001.

Statistical nonsignificance, however, is not a formal demonstration of equivalence. An equivalence or noninferiority conclusion requires prespecified margins, adequate design, and suitably precise confidence intervals. This observational comparison should instead be described as not detecting a robust group difference in these outcomes after multiplicity correction. Neither counsel nor expert witnesses should paraphrase that as proof that cannabis has no residual effects for all users, routes of administration, doses, or tasks.

What prevents the study from proving a specific driver’s impairment?

Exposure was not randomized. The frequent-use group smoked self-chosen cannabis at home before testing, whereas controls were not assigned to do so. The study cannot disentangle recent exposure from every feature of a history of frequent use. The groups differed in education and racial composition, and observational covariate adjustment cannot guarantee that those differences—or unmeasured driving habits—have been neutralized.

The time window was narrow. Measurements approximately 12–15 hours after use provide no direct data about the first hours after smoking, the next afternoon, occasional smokers, concentrates taken another way, or a different person’s rate of change. The actual quantity absorbed cannot be obtained by treating a product label or estimated mass consumed as a validated systemic dose.

The endpoint was simulation. Lane control in one driving program does not reproduce every decision in actual traffic: divided attention from passengers, intersections, lights, pedestrians, unexpected emergencies, fatigue, emotional arousal, and road geometry may differ. A forensic opinion on a crash needs the person’s behavior, timeline, vehicle and roadway evidence, toxicology, and plausible alternative mechanisms.

There was no validated individual classifier. The investigators did not establish sensitivity, specificity, predictive values, an impairment threshold, or a known error rate that would turn a single blood-THC or oral-fluid reading into an individualized judgment. They did not perform a per-defendant forensic reconstruction. An absence of robust mean group differences cannot furnish those missing operating characteristics.

How should a lawyer or toxicologist use this paper?

First establish the exact biological sample, its collection time relative to the event, the identified parent drug and metabolites, laboratory method, quantification threshold, storage history, and uncertainty. This is why biological matrices, specimen collection, and confirmatory drug testing require separate attention. Neither oral fluid nor blood concentration should be converted to another matrix by assumption.

Second identify what is actually alleged: recent use, chemical presence, behavioral impairment, statutory violation, or causation of an incident. For a purported driving impairment, seek objective observations, driving sequence, contemporaneous field or clinical findings, road conditions, fatigue and other potential causes. Pharmacokinetic interpretation also requires distinguishing measured concentrations from modeled concentration histories. Do not treat an average between-person simulator result as an individualized dose-response function.

Finally, disclose material weaknesses rather than hiding them. This paper concerns frequent smokers at a specified morning-after interval, contains demographic imbalances, reports strongly heterogeneous blood concentrations and self-ratings, and adjusts for many tests. It is relevant when an expert makes a categorical claim that a delayed THC-positive specimen necessarily proves lane weaving. It is equally relevant when someone incorrectly claims the paper has cleared every THC-positive driver of impairment. Neither categorical claim is scientifically supported by this source.

Frequently asked questions

1. Did this study prove that driving the morning after cannabis is safe?

No. It found no statistically significant group differences on its tested simulator-driving outcomes after multiple-comparison correction. It did not validate safety for an individual driver, occasional users, other routes or doses, or actual road collisions.

2. How many drivers were directly studied?

One hundred thirty participants completed the experimental procedures: 65 frequent smokers and 65 nonrecent-use controls. The study was not a randomized crossover in which all 130 people drove both with and without cannabis.

3. Was blood THC still measurable the next morning?

Yes, with substantial heterogeneity. The smoker group had mean blood THC 2.8 ng/mL (SD 4) and median 1.2 ng/mL (IQR 0.5–2.7); some lower observations involved a study rule for values below the quantification threshold. These group summaries are not an individual driver’s value.

4. Did cannabis users have greater lane weaving?

No statistically significant greater SDLP was established. The cannabis group’s adjusted mean SDLP was numerically lower in standard and dual-task simulator conditions; both between-group p values were .277 after correction. That does not demonstrate improvement or equivalence.

5. Did oral-fluid THC predict brake latency?

The exploratory correlation (r = .286, unadjusted p = .021) did not survive the authors’ multiplicity correction (adjusted p = .136). Oral-fluid concentration cannot be substituted for a contemporaneous blood concentration or a validated person-specific impairment measure.

6. Can a positive blood THC result alone establish who caused a collision?

No. The study measured specimens and simulator performance, not causation in a particular incident. Attribution requires separate case-specific evidence, including timing, driver conduct, roadway and vehicle circumstances, other drugs, and competing causes.

7. Does this paper contradict a companion study about next-morning memory?

Not merely because the two studies emphasize different results. The driving paper tests lane position, speed, following distance, and braking. A separate same-cohort investigation addresses verbal recall and other cognitive endpoints. Distinct tasks can yield different findings; neither should be silently used as the other’s outcome evidence.

Primary evidence and bottom line

For an exact account of inclusion criteria, specimen analysis, outcomes, multiple-comparison adjustments, and Tables 1–6, consult the complete original Zakala et al. (2026) study, DOI 10.1186/s42238-026-00416-w. Its evidentiary contribution is narrow and useful: in frequent smokers tested approximately 12–15 hours after self-selected cannabis use, measured blood and oral-fluid concentrations did not yield a validated individualized predictor of these simulator measures, and adjusted group differences on those measures were not statistically significant. Chemical detection, research-task performance, legal impairment, and collision causation remain distinct propositions.