Can 10 mg of Dexamphetamine Offset Alcohol-Impaired Driving? What a 16-Driver Simulator Study Actually Found

What question did the investigators actually test?

The investigators asked whether dexamphetamine would reduce alcohol-related sedation or impairments of simulator driving, and whether the combination would change risk-taking behavior. That is narrower than asking whether stimulant exposure is generally protective, whether an amphetamine-positive blood result proves impairment, or whether two drugs caused a real collision. The study deliberately used four treatments in the same volunteers, making comparisons possible among a stimulant alone, alcohol alone, both substances, and placebo.

The central contrast is particularly important for forensic interpretation. An apparent reduction in one performance deficit when dexamphetamine accompanies alcohol is not enough to establish a reliable compensating effect. The relevant question is whether the paired treatment differs statistically and meaningfully from alcohol alone, and whether it changes the outcomes that matter: staying within a lane, choosing safe gaps, stopping at signals, attending to simultaneous demands, and handling unexpected traffic. These endpoints are not interchangeable with one another.

The experiment is also not an actual roadside validation study. There were virtual roadways and simulated collisions, not police observations of specific motorists or adjudicated accidents. It can provide controlled experimental evidence about group responses to these doses and tasks, but not a certified rule for classifying every driver who has alcohol or amphetamine in a specimen.

Who participated, and what were the four dosing conditions?

Eighteen licensed volunteers enrolled; complete datasets were available for 16 individuals (12 men and four women). One participant stopped after becoming ill on the first trial day; another withdrew for unrelated medical reasons. Those 16 people, not 64 different subjects, were the main repeated-measures cohort. For the lane-position calculation, three more were excluded after off-road driving, meaning that the effective denominator for that outcome was smaller than for the general study.

Volunteers were infrequent recreational users of alcohol and amphetamine-like drugs, had held a driver’s license for at least two years, and practiced the simulator before the four actual test visits. Their average age was 25.7 years and their mean reported annual mileage was approximately 5,600 kilometers. Extrapolating these results to a typical older motorist, a chronic high-dose stimulant user, a person with sleep deprivation, or a patient receiving long-term medication would require evidence not supplied by this experiment.

Across randomized crossover conditions, participants received 10 mg dexamphetamine alone, 0.8 g alcohol per kilogram of body weight alone, both treatments, or placebo. There was a seven-day washout between conditions. Dexamphetamine/placebo was blinded, and the testing period was arranged so that the simulated drive occurred roughly two hours after study-drug ingestion. The paper describes alcohol administration in three portions to produce the planned alcohol exposure. This is a controlled dose, not the same thing as a laboratory measurement of circulating drug.

Which specimens and analytes were measured directly?

Specimen identity is decisive. The investigators drew 10 mL of venous blood before the simulator task to determine dexamphetamine concentrations. They reported mean drug concentrations of 20.8 ng/mL in the dexamphetamine-only condition and 20.7 ng/mL in the dexamphetamine-plus-alcohol condition, with broad individual ranges. The publication calls these blood concentrations, but does not fully establish in the presented Methods the processed fraction, laboratory method, validation characteristics, or whether the reported figure can be transferred without qualification to a particular legal whole-blood reporting convention. It would be unsound to invent a serum-to-whole-blood correction, an assay uncertainty, or an unreported analytical technique.

Alcohol was different: investigators used a Dräger Alcotest 6510 on breath and transformed the readings into a BAC-equivalent display in per mille. Thus the paper’s table titled “Blood alcohol concentrations” does not mean that contemporaneous venous blood was drawn and chemically analyzed for ethanol at every time point. At the beginning of the simulated drive, the displayed mean was 0.85‰ with alcohol alone and 0.91‰ with the combination; at the end of driving both groups averaged 0.64‰. Such estimates depend on the breath instrument and conversion conventions and must not be mislabeled direct blood GC results.

The study also obtained previsit breath screens and a urine drug screen, but it did not report quantitative dexamphetamine concentrations in urine, saliva, hair, or other media as outcome measurements. Nor did it report contemporaneous direct ethanol concentrations in human forensic venous whole blood. The distinction between an administered 0.8 g/kg alcohol dose, a breath-derived BAC-equivalent, and a measured venipuncture dexamphetamine concentration should remain explicit in any forensic explanation.

Source-measured exposure and study denominator: evidence table 1

The approximately matched dexamphetamine concentrations make it more plausible that the combination and stimulant-only conditions produced comparable measured exposure to dexamphetamine at this sampling point. They do not imply equivalent brain responses or provide a validated person-specific impairment threshold. Likewise, similar estimated BAC values across the two ethanol arms reduce one exposure-comparison concern without converting the breath measurements into direct blood results.

What happened to lane keeping under alcohol and dexamphetamine?

The most readily interpretable continuous driving measure was the standard deviation of lateral position (SDLP), or how much a vehicle drifts sideways while following a straight roadway. Lower values generally reflect more stable lateral positioning in the particular simulator task, although different road scenarios can have different baselines. Investigators assessed separate monotonous drives with posted speed limits of 80 and 100 km/h. Three volunteers were excluded from this specific analysis because they drove off the road; those exclusions matter when discussing strength of inference.

At 80 km/h, the reported mean SDLP was 20.3 cm under alcohol versus 15.4 cm under placebo, a 4.9-cm difference. At 100 km/h, the corresponding means were 20.1 and 15.2 cm. The authors reported alcohol-versus-placebo statistical differences at both speeds (F(1,12)=7.12, p<.02; F(1,12)=6.49, p<.03). Those findings concern the alcohol condition in this simulator: they are not universal threshold values for legal whole-blood alcohol or a direct estimate of collision probability.

With both substances present, mean SDLP was 17.2 cm at 80 km/h and 17.7 cm at 100 km/h, compared with 17.8 and 18.8 cm for dexamphetamine alone. Although the combination looked numerically better than alcohol alone, the paper did not demonstrate a statistically reliable separation of combination from alcohol in this endpoint. The authors’ equivalence exercise left a confidence interval spanning both no-effect and alcohol-level impairment references. It would be an error to describe these descriptive means as proof that 10 mg dexamphetamine safely counteracted alcohol.

Did drivers choose riskier crossing gaps?

The simulated junction tasks gave participants approaching vehicles and required them to judge whether they could cross safely. The investigators measured accepted time gaps and distances, which reflect a specific operational choice rather than a global personality characteristic. There were two junction configurations, including traffic approaching from left to right and a Y junction with right-of-way constraints.

Alcohol and alcohol-plus-dexamphetamine produced shorter accepted gaps than dexamphetamine alone or placebo. For left-to-right crossing, the group-level treatment effect on accepted time gaps was F(3,42)=5.13, p=.004 and on distance F(3,42)=3.87, p=.02. The Y-junction effects were likewise reported significant for time (p=.02) and distance (p=.009). These are meaningful simulator risk-choice indicators: participants selected tighter opportunities to cross the roadway when alcohol was present.

The critical null comparison is that gap distance did not significantly differ between alcohol alone and the combination. Thus the stimulant was not shown to reverse the shorter acceptance distances associated with drinking. The raw source figures depict averages with standard-error whiskers, not an individual dangerousness classifier. The task also involved no real motor vehicles or exposed road users, so these judgments cannot be converted into a crash-specific causal opinion without independent evidence.

What did red-light violations and simulated collisions show?

In traffic-light dilemma scenarios, drivers sometimes encountered a sudden signal requiring a stop. The original report states that drivers failed to stop in 64% of scenarios with both drugs, 59% with alcohol alone, and 30% with dexamphetamine alone or placebo; the overall chi-square comparison was 10.8496 with three degrees of freedom, p=.02. Those percentages describe simulated traffic-light occasions, not the percentage of subjects legally determined to have run a light on public roads. The 64%-versus-59% difference by itself was not reported as evidence of a significant interaction.

The investigators counted 18 virtual collision events across the tested conditions: six with dexamphetamine plus alcohol, eight with alcohol alone, one with dexamphetamine alone, and three with placebo. They reported significantly more events with alcohol-containing conditions than placebo (F(1,13)=6.06, p<.03). These are event counts, not 18 separate injured people and not an estimate of the population risk of fatal crashes. Within-person dependence and sparse counts limit the precision of any relative-risk reconstruction; it would be inappropriate to calculate a courtroom odds ratio from these totals without the original event-level data.

No reliable difference between dexamphetamine alone and placebo was identified for collisions. Two instances of improper overtaking via the emergency lane occurred in the alcohol-only condition. The reasonable summary is that alcohol-containing arms showed unsafe simulated choices and collisions, not that a stimulant uniquely created those events or proved guilt in any specific incident.

Key driving and cognitive results: evidence table 2

Table rows are selected source endpoints, not an invitation to pool heterogeneous measures into a synthetic “impairment score.” The simulations contained repeated tests and multiple correlated outcomes. The paper does not supply the raw subject-level dataset or a multiplicity-adjusted overall traffic-safety index from which to reverse-engineer an individual’s probability of impairment.

Did the stimulant help participants feel awake?

Yes, in a limited subjective sense. On the nine-point Karolinska Sleepiness Scale, the overall treatment effect was F(3,45)=6.82, p=.0007. Participants given dexamphetamine, alone or with alcohol, reported less sleepiness than those given placebo or alcohol alone. Profile of Mood States ratings also reflected greater perceived energy and less fatigue under dexamphetamine in relevant comparisons.

But a person may feel alert while still accepting a dangerous crossing gap or failing to stop at a simulated traffic light. Reduced subjective drowsiness is not a validated proxy for restored divided attention, lane stability, judgment, or legally adequate driving. The study’s own main conclusion distinguished stimulation or subjective wakefulness from reliable reversal of alcohol-related traffic-risk behavior.

For testimony, a statement such as “the person was not sleepy, so alcohol could not have impaired them” would exceed the evidence. Equally, a positive stimulant result does not dictate how this one person’s choice-making must have changed. The correct evidentiary bridge requires task-specific data or case-specific observations, interpreted within the validated limits of that evidence.

What happened on divided attention and vigilance testing?

After simulated driving, participants completed computer tasks in two later sessions as breath-derived alcohol estimates were falling. On the divided-attention task, alcohol alone produced the most pronounced difficulties: greater tracking error, longer reaction times, more missed targets, fewer hits, more losses of tracking control, and false alarms in the aggregate comparisons. The first session occurred while alcohol-arm mean estimates were roughly 0.55–0.64‰, and the second while estimates were around 0.23–0.37‰. Timing therefore matters: later cognitive test values should not be mapped mechanically onto the earlier drive.

The initial DAT tracking error was 19.0±1.2 mm under alcohol alone, 17.3±0.9 mm under combined exposure, 14.5±1.4 mm under dexamphetamine alone, and 16.0±1.2 mm under placebo. The study reported an overall tracking-error treatment effect p=.004. A clear between-arm result was that false alarms with the combination exceeded those with dexamphetamine alone, but this does not by itself prove the combination was worse than alcohol alone.

On the concurrent vigilance-and-tracking task, alcohol alone likewise produced larger tracking error and slower reactions. The omission-percentage model was p=.07, so it should not be recast as significant. The separate critical tracking, psychomotor vigilance, and visual-field selective-attention tests showed no significant between-treatment differences. These null findings must appear alongside the positive results; selective reporting of only impressive task differences distorts the original experiment.

Did measured drug levels correlate with performance?

Investigators searched for associations between tasks and their reported alcohol estimates or venipuncture dexamphetamine levels. They did not demonstrate relevant statistically significant relationships for the examined divided-attention or VigTrack measures. That observation is not an assay validation exercise and should not be turned into “blood drug concentrations never relate to impairment.” With only 16 complete participants and narrow controlled dosing, the analysis may have had limited power to detect concentration-response relationships.

Nor does a group mean of approximately 20.8 ng/mL identify the amount of impairment expected at 20.8 ng/mL in a different adult. The time since administration, analytical specimen, measurement uncertainty, tolerance, concomitant substances, task demand, and population all matter. A prosecutor or expert who cites the mean as an automatic cut point would be importing a threshold the study did not establish.

In the alcohol arms, converted breath estimates were declining over the experimental session; this does not show that a later negative or lower specimen measurement identifies a person’s earlier exact performance. Retrograde extrapolation of an individual’s toxicological result requires independent case-specific pharmacokinetic assumptions and data absent from the crossover paper.

What did the ECG monitoring show?

Continuous electrocardiographic monitoring identified two episodes of paroxysmal supraventricular tachycardia during the combined dexamphetamine-plus-alcohol condition. The paper states the episodes were asymptomatic and resolved spontaneously. Other between-treatment comparisons of heart rate, blood pressure, and reported adverse effects did not show significant differences, and no severe adverse effects were reported in this small study.

The two ECG episodes deserve faithful reporting because omitting them could falsely suggest an entirely unremarkable safety observation. They do not establish that a given subsequent arrhythmia, medical injury, or fatality was caused by this dose combination. Event counts in a small crossover experiment are signals for interpretation, not definitive comparative clinical risk estimates. The authors did not design this trial as a sufficiently powered cardiac-outcomes study.

Why is the sample size smaller for some measures?

A common reporting error is to quote the enrollment count of 18 as though every table analyzed 18 complete independent subjects. Two subjects withdrew, so 16 provided general full datasets. For SDLP, the original researchers additionally excluded three for leaving the roadway; the F-test denominator degrees of freedom align with the reduced data, but individual condition-level raw records were not supplied. The exclusion of people who drove off the road could affect both average and variability and may preferentially omit extreme performances.

Additionally, each participant experienced four conditions. A collision or traffic-signal response in multiple visits is a repeated event, not a new driver. Even when the sample size remains 16, group summaries across outcomes depend on different event opportunities or missing observations. A table of means is never equivalent to a validated personal accuracy or calibration table.

The study is useful precisely because it controlled treatment conditions, yet the convenience sample was small. It did not sample motorists randomly from the driving population and cannot quantify how common a given reaction would be among the many subgroups encountered in criminal, civil, clinical, or occupational cases.

What design limitations constrain a forensic conclusion?

First, the dose combination was a relatively low 10 mg dexamphetamine with substantial alcohol; higher or repeated amphetamine regimens were not tested. Second, a 50-minute simulator with specific scripted junctions is informative about those tasks, not automatically interchangeable with road traffic emergencies, weather, distractions, injuries, or fatigued driving. Third, the participants were young, experienced at least two years of driving, and infrequent recreational users; generalization to chronic medication users, heavy drinkers, or particular clinical groups is uncertain.

Fourth, alcohol exposure was expressed with instrument-derived estimates, and analytic fraction/validation details for venipuncture dexamphetamine were not reported at the level necessary to reproduce a courtroom whole-blood assay. Fifth, outcome-specific missing observations and repeated within-person events affect denominators. Sixth, many simulator and cognitive outcomes were tested without a single externally validated aggregate legal safety threshold. Multiple positive findings therefore should be interpreted alongside genuinely null comparisons and source-reported confidence intervals, not cherry-picked from a long menu.

Finally, the relevant comparator was the alcohol arm itself. The combined condition cannot be called protective simply because some numerical means were smaller than alcohol alone. The report expressly states that the clinical significance of the combination’s SDLP pattern was ambiguous, rather than proving either complete rescue or uniform injury.

How should toxicologists explain dexamphetamine versus methamphetamine?

The administered stimulant was dexamphetamine (d-amphetamine), not methamphetamine, methylphenidate, or lisdexamfetamine. These drugs are chemically and pharmacologically distinct, have different dose-equivalence questions, and may be assayed or reported differently. A test for parent amphetamine cannot be described as a controlled methamphetamine dose merely because both fall under a broad stimulant category.

The paper reports an administered 10 mg dose of dexamphetamine and a mean venipuncture concentration near 21 ng/mL around the simulator period. It does not characterize a driver’s metabolic conversion of other stimulants, validate an immunoassay confirmation threshold, or show whether a real-life reported amphetamine result reflected prescribed medication, illicit use, contamination, or timing. Those questions need the actual specimen, confirmatory laboratory procedure, collection record, and drug history.

For a concise explanation of these requirements, the site’s published guides to biological matrices and confirmatory drug testing address the difference between a measured analyte and the conclusions an investigator wants to draw. They provide interpretive context, not additional experiments that replicate this crossover trial.

What can this study support in a DUI or collision analysis?

A defensible limited proposition is that, in a small controlled simulator crossover, alcohol at 0.8 g/kg adversely affected several driving-relevant outcomes and addition of 10 mg dexamphetamine did not establish reliable global reversal. More specifically, alcohol-alone group means showed greater SDLP, shorter junction gaps, and more virtual hazardous actions than placebo on reported comparisons. The combined condition preserved several of those risk-choice patterns even where a subjective alertness measure improved.

What the trial cannot establish is that an identified driver at a reported amphetamine concentration was intoxicated, legally impaired, inattentive, driving aggressively, or the factual cause of a collision. It cannot prove the exact BAC from a later specimen, the time of use from an isolated amphetamine positive, or how a particular witness saw or remembered a traffic light. One must not combine its human experimental data with another person’s drug results without additional validated interpretive steps.

The core forensic questions should remain separate: what compound was identified, in what matrix, by which method; what actually was measured or observed at the event; what uncertainty attaches to timing and assay; and what causal alternative explanations remain. A validated laboratory result, an experimental group association, and a case-specific legal conclusion are three different evidentiary levels.

Seven frequently asked questions

1. Does dexamphetamine cancel alcohol’s effects on driving?

Not according to this study. The 10 mg dexamphetamine dose did not establish a statistically reliable reversal of alcohol-associated simulator impairments or risky junction decisions. Improved subjective wakefulness was not a substitute for safe task performance.

2. Was alcohol measured in venous blood?

No contemporaneous venous chemical ethanol assay was reported. Alcohol was measured with a Dräger breath analyzer and transformed into BAC-equivalent values shown in per mille. Calling those data direct whole-blood gas chromatography would be incorrect.

3. Was dexamphetamine measured directly?

Yes. The team drew venous blood before simulated driving and reported group mean dexamphetamine concentrations around 20.7–20.8 ng/mL, but did not fully characterize the processed specimen fraction or analytical validation details in the published methods.

4. Did all 18 enrolled people contribute to each result?

No. Sixteen provided complete general datasets, and three further participants were excluded from the SDLP analysis for leaving the virtual road. Tasks and events also had different opportunities or missing data.

5. Does a 64% failed-stop rate mean 64% of people are impaired?

No. It was the proportion of certain simulated signal-challenge occasions under the combined condition. It does not describe 64% of an external population or establish a clinically validated personal intoxication classification.

6. Were there important nonsignificant findings?

Yes. The alcohol-only versus combination lane-control differences were not significant, gap distance did not significantly separate those two arms, vigilance omissions reached p=.07 rather than significance, and several standalone cognitive tasks did not differ significantly among treatments.

7. Can the study explain an actual DUI arrest or fatal collision?

It can inform hypotheses about how a particular dose combination affected certain experimental tasks. It cannot, without independent case evidence, decide an individual’s true incident performance, drug source, legal whole-blood concentration, witness truthfulness, or causation.

Research bottom line

The practical lesson is not that stimulant concentrations identify a safe or unsafe driver. The lesson is that reducing sleepiness and restoring driving competence are different experimental questions. The original Simons and colleagues crossover reported alcohol-associated lane variability, riskier accepted gaps, traffic-light violations, and simulator collision events despite the presence of dexamphetamine in a separate combined arm. Other comparisons—including some cognition outcomes and the crucial combination-versus-alcohol contrasts—were nonsignificant or uncertain. The source’s exact dose, 16-person analyzed cohort, breath-estimated alcohol values, venous dexamphetamine measurement, and simulator-specific tasks define the boundaries of the evidence.

For exact details, consult the full original study and its tables. The published therapeutic drug monitoring guide offers additional background on why sampling time and matrix matter when interpreting a concentration; it is not a substitute for laboratory records or person-specific analysis.