Does Oral d-Methamphetamine Impair Simulated Driving? Why a 20-Person Trial Found a Slower Emergency Speed but No Significant Overall Deficit

Short answer: In a double-blind crossover experiment involving 20 licensed adult recreational stimulant users, a single oral dose of 0.42 mg/kg d-methamphetamine did not produce a statistically significant difference in the study’s overall simulated-driving error score compared with placebo (reported p = 0.19). Participants drove about 3.1 km/h slower at one simulated freeway emergency under methamphetamine (103.0 versus 106.1 km/h; unadjusted p = 0.02). The apparently greater number of red-light stopping failures did not reach conventional statistical significance (p = 0.11). The important distinction is between a narrowly observed laboratory contrast and any claim that a particular driver was safe, impaired, or responsible for an actual collision.

These findings come from Silber and colleagues’ original 2012 controlled experiment, The effect of d-methamphetamine on simulated driving performance, Human Psychopharmacology: Clinical and Experimental, 27:139–144, DOI 10.1002/hup.1238. This article evaluates the paper’s directly measured drug specimens, its separate behavioral endpoints, the substantial multiple-testing problem, and the boundaries of any forensic inference.

What did the experiment actually test?

Investigators recruited 20 healthy adults, ten women and ten men, aged 21–32 years, with a reported mean age of 25.4 years. Each held a full driver’s license, had previously experimented with illicit stimulants, and reported nonregular stimulant use. The group was neither a representative sample of all motorists nor a cohort of chronic, sleep-deprived, intoxicated, injured, or severely dependent methamphetamine users. These differences matter because prior use, tolerance, fatigue and other substances may alter both drug exposure and performance.

Every person underwent two experimental conditions: an oral dose of 0.42 mg/kg d-methamphetamine and a matched placebo. Treatment order was counterbalanced and both participants and investigators were blinded. A two-week washout separated the visits. A crossover design is valuable because each participant supplies their own placebo comparator, reducing some between-person variation. It does not remove practice effects, unmeasured order effects, fatigue fluctuations, or the restrictions imposed by a short laboratory driving exercise.

The tested drug was d-methamphetamine—the dextrorotatory enantiomer—not racemic d,l-methamphetamine, d-amphetamine, lisdexamfetamine, MDMA, or a mixture with alcohol. Those compounds and preparations should not be interchanged simply because their names contain “amphetamine.” The trial involved a single monitored oral dose, not repeated illicit smoking, injection, or prolonged wakefulness.

How was driving evaluated?

About two and a half hours after dosing, participants completed a set of simulated roadway exercises with a steering wheel, accelerator, brake pedal, signal controls, projected roadway scene, and dashboard. The protocol included day and night freeway and city scenarios. Four approximately five-minute segments produced roughly 20 minutes of simulated driving per study visit.

The investigators recorded 35 error or behavior variables. Each contributed to an overall weighted performance score according to an assigned severity loading factor. Higher total scores reflected more weighted errors within that simulator’s scoring framework; they were not measured probabilities of collision and not a validated conversion to a legal definition of impairment. Many of the component variables involved signaling, following distance, braking, lane position, speed, and emergency responses. Some measures, such as a “collision” variable, were simulated events or scoring counts—not actual roadway crashes.

Two levels of analysis must be kept separate: a prespecified or overarching composite score that aggregates error types, and 35 individual comparisons that search for changes within specific behaviors. It is entirely possible for one component to differ nominally while the global score does not. Whether that component represents a real reproducible effect requires an additional statistical and substantive assessment.

What happened to the overall driving-error score?

The reported mean weighted error score was 95.3 under placebo (standard error 3.8) and 102.4 after d-methamphetamine (standard error 6.9). The source reports F(1,18) = 2.17 and p = 0.19 for the overall treatment effect. That is a nonsignificant group comparison at the conventional 0.05 threshold. The numerical mean was higher, not lower, in the drug condition; describing the result as a demonstrated performance enhancement would therefore also be unsupported.

A nonsignificant result is not proof of no effect at every dose, for every task, in every driver, or over every time interval. Twenty people provide limited statistical precision, particularly when task outcomes are variable. The study does not report a diagnostic sensitivity for methamphetamine impairment, a minimal clinically important change, or a calibrated probability of crash as a function of its score. It can support the narrower conclusion that this protocol did not resolve a statistically significant difference in its chosen composite endpoint.

The investigators also reported no significant interaction or difference by the day versus night scenarios in the relevant analysis (F(1,18) = 0.09, p = 0.76). These exercises were both brief simulations. The result cannot establish that complex real nighttime conditions have the same safety risk as daytime driving.

Which individual driving measurements changed—and which did not?

The standout nominal comparison was speed at a simulated freeway emergency: 106.1 km/h under placebo and 103.0 km/h after d-methamphetamine, Wilcoxon T = 44, p = 0.02 in the original numerical table. Both means nevertheless exceeded the 100 km/h speed limit referenced by the investigators. A slightly lower mean speed at this event does not by itself identify improved hazard recognition, better braking, an intentional safety response, or greater legal fitness to drive.

Red-light nonstopping scores were higher in the active-drug condition, 4.1 versus 1.0, but the paired test returned p = 0.11. The authors described roughly fourfold more such violations; that phrasing describes the sample means rather than a proven population-level relative risk. The score variation was considerable: standard deviations were 7.6 and 3.1, respectively. A high ratio from small variable means is particularly sensitive to a few observations.

Other tests were nonsignificant: the source’s simulated collision measure was 4.5 versus 4.0 (p = 0.81), excessive-speed-limit score 6.0 versus 7.2 (p = 0.19), inappropriate following-distance score 48.3 versus 44.5 (p = 0.51), and freeway emergency stopping distance 99.7 versus 97.3 in the simulator’s distance scale (p = 0.94). An unnecessary-stopping comparison landed exactly at the paper’s reported borderline p = 0.05, with more scoring under placebo; it should not be transformed into a broad clinical benefit.

Original evidence table: global endpoint and selected behaviors

Why does testing 35 individual variables matter?

The authors performed a Wilcoxon signed-rank test for each of 35 individual driving variables after evaluating the overall score. More testing creates more opportunities for a small p value to occur by chance. With 35 nominal tests evaluated at 0.05, the expected number of false-positive results under a complete set of true null hypotheses would be 1.75, although the actual error structure depends on correlations among outcomes. This simple calculation does not prove that the freeway result is false; it shows why it should not receive the same confidence as a single prespecified primary endpoint.

A straightforward Bonferroni familywise threshold for 35 tests would be approximately 0.00143, much smaller than the observed freeway-speed p = 0.02. Bonferroni may be conservative when driving measures correlate, and different prespecification or multiplicity strategies could be justified. Crucially, the paper does not present the freeway comparison as surviving a validated adjustment for the family of 35 comparisons. A careful forensic characterization is therefore “nominally significant and exploratory,” not “the drug reliably improves emergency driving.”

The contrast between the overall composite and its components is also substantive. Scores for 35 different behaviors can dilute a real task-specific difference, while isolated components can be noisy. Both possibilities justify additional replication with clearly defined primary outcomes, wider exposure conditions, and preplanned controls for multiplicity—not a definitive positive or negative individual impairment classification.

Did the investigators chemically measure methamphetamine?

Yes. The team obtained venous blood by antecubital venipuncture and oral-fluid specimens using an absorbent collection swab. The methods describe broad class screening in blood by ELISA/EMIT, followed by specific amphetamine measurements in blood and oral fluid by gas chromatography–mass spectrometry. The paper reports mean d-methamphetamine concentrations at multiple post-dose intervals. The blood concentrations are reported as “blood,” without establishing in sufficient analytical detail that the values represent a specified unseparated forensic whole-blood fraction rather than a processed laboratory fraction; it would be unsafe to silently equate them with a jurisdiction’s validated legal whole-blood result.

At approximately 120 minutes after administration, mean methamphetamine in the reported blood matrix was 72 ng/mL, whereas oral fluid averaged 285 ng/mL. At 170 minutes these values were 67 and 223 ng/mL; at 240 minutes, 59 and 190 ng/mL. These are descriptive group means, and the saliva dispersions were substantially larger. A paired individual oral-fluid reading is not a fixed multiple of blood concentration. The swab was collected into an assay buffer, so device-specific extraction, dilution, calibration and reporting conventions also deserve scrutiny before cross-matrix comparison.

Original evidence table: measured concentrations, not impairment thresholds

Can a measured concentration explain the simulator result?

The experiment did not establish a concentration–effect curve mapping each person’s methamphetamine level to their driving-error score. Dosing was weight adjusted, and the group subsequently showed different concentrations and responses. A group mean of 67 or 72 ng/mL does not mean every participant drove at that concentration, nor that different people with the same concentration would produce the same outcome.

Sampling time is relevant because the simulator was run around the post-dose interval, whereas drug concentrations were measured at specific times on either side. A test obtained after an event is not automatically equivalent to one at the event. The trial also does not supply an individual validated back-extrapolation equation, person-specific absorption profile, or validated per-se cutoff. Pharmacokinetics and pharmacodynamics are distinct: detectable parent methamphetamine supports exposure, but a driving task and individual behavior are separate measurements.

For general specimen handling, the live biological matrices guide explains why blood and oral fluid represent different compartments. The confirmatory testing guide addresses identification and laboratory specificity. These broader resources cannot repair missing analyte, timing, or legal evidence in an individual case.

How do oral-fluid results differ from blood results?

Oral fluid can contain methamphetamine at concentrations that differ markedly from blood because secretion, pH, collection and sample preparation affect its measured level. In this experiment, the reported oral-fluid means were greater than the blood means at all three post-dose times, but the saliva standard deviations were wide. A high saliva-to-blood ratio of group averages should not be treated as a validated ratio for an individual driver or as proof of current dangerous behavior.

Questions about a specific specimen include whether the device collected the expected volume, whether a collection buffer diluted the sample, whether an assay was validated for that medium, and whether the laboratory reported raw collection-fluid or corrected concentration. The paper’s methods describe a swab and buffer system but do not provide the case-specific reporting metadata required to convert its numbers to another collection system. Cross-specimen interpretations should preserve the actual matrix and the instrument’s analyte specificity.

What does a negative overall score mean in a forensic case?

The phrase “not statistically significant” addresses the evidence for a group difference in this limited design, not a certification of driving safety. Nor is it proof that methamphetamine cannot impair driving: driving performance changes with dose, route, prior exposure, sleep debt, other drugs, setting and task difficulty. Equally, the presence of methamphetamine in a driver’s blood does not permit importing a positive result from some other experiment as proof that this particular person failed to drive safely.

In a real case, the forensic scientist should distinguish a laboratory’s chemically confirmed substance finding from a driver’s documented speed, traffic control violations, observations, collisions, medical presentation and alternative explanations. Timing, co-ingestion, route and assay quality are essential. If an expert cites this trial for broad claims, the expert should also disclose the nondifference in the overall composite and the null results for collision and red-light comparisons. Selectively highlighting one p = 0.02 result without the family of tests risks overstating evidentiary strength.

Can the experiment be generalized to racemic methamphetamine or other stimulants?

No direct interchangeability is established. The paper specifically tested d-methamphetamine at 0.42 mg/kg. Racemic d,l-methamphetamine includes a different enantiomer composition and was investigated separately in other work. d-amphetamine and lisdexamfetamine are also different drugs or prodrugs, although related pharmacologically. Comparisons require their own study designs, exposures, analytes and endpoints. An article about a different stimulant cannot be substituted as if it tested the same intervention.

The same caution applies to administration method. A single supervised oral dose does not represent smoked or injected high-dose methamphetamine, a stimulant taken during an extended period without sleep, or a combination with alcohol, opioids, sedatives or other psychoactive substances. Likewise, the results say nothing direct about crash risk in a person with cardiovascular symptoms or another medical condition excluded by the original study.

Why the participants and simulator matter

Twenty relatively young, medically screened, nonregular recreational stimulant users completed brief simulated tasks. That selection increases internal control for a pharmacological experiment but sharply limits population generalization. Licensed driving status confirms prior qualification, not uniform driving experience, real highway competence, or equivalence to a commercial driver after an overnight shift. The study measured no actual roadway collision, police observation, roadside divided-attention battery, or objective sleep restriction.

Participants performed four short tasks in a two-dimensional simulated roadway. Real driving can impose prolonged monotony, sustained attention, complex intersections, unpredictable road users, adverse weather, and consequences that cannot be recreated by a brief simulation. A modest change in freeway emergency speed inside that environment should not be used to infer the probability of a pedestrian crash or a reduction in stopping distance on a particular road.

Moreover, the composite used severity loading factors assigned to individual error categories. Such scoring can be useful for consistent within-study comparisons, but it is not the same as a validated clinical diagnosis or a cross-jurisdiction legal impairment standard. A weighted sum of simulated mistakes is neither a direct drug assay nor a causal finding about an accident.

What scientific limitations should counsel examine?

At least six limitations affect case use. First, the biological sample of 20 people is small. Second, the observed window centered around 2.5 hours after one oral dose, leaving earlier absorption and later effects less studied. Third, only one active dose was randomized; there is no robust dose–response assessment. Fourth, the 35 individual comparisons create a multiplicity problem. Fifth, broad composite scores and laboratory scenes may miss behaviors relevant to actual traffic. Sixth, the paper does not validate an individualized drug concentration or simulator score as a legal impairment classifier.

The paper also contains inconsistencies in how the timing of collection is summarized between method paragraphs. These do not justify inventing a new time series. The directly reported 120-, 170- and 240-minute numerical results can be described as printed, while a case analysis requiring exact sampling chronology should request the protocol, specimen identifiers, chain of custody, laboratory analytical records and participant-level timestamps. The results table does not substitute for those underlying records.

Statistical adjustment is another necessary distinction. One nominal p = 0.02 result from dozens of correlated tests could reflect a task-specific drug effect, statistical noise, or both. Without an independent replication and prespecified error-control strategy, it should be presented as a question for further study rather than a proven mechanism. Conversely, nonsignificant red-light and collision scores do not establish that the true drug effect is exactly zero.

Practical interpretation of a drug-positive driver

A defensible forensic report should begin with exactly what was chemically analyzed and when: the reported blood fraction, blood collection time, oral-fluid device, chemical confirmation, calibration, and uncertainty. The next step is the contemporaneous driving evidence—not a generalized label based on an analyte concentration. Evidence of speed, signals, following distance, braking, crash reconstruction, fatigue or medical events may support or contradict a theory, but each should be treated according to its own reliability.

Controlled data like this trial can test proposed mechanisms. They cannot identify which person in the study produced a particular event from the group mean, and they cannot reconstruct an unrelated defendant’s absorption, last dose or mental state. The absence of a significant trial-wide decrement cannot be used as affirmative proof of an individual driver’s fitness; the presence of a compound cannot be treated as a biological verdict of impaired operation.

For discussions of what serial concentrations can—and cannot—do, the site’s published therapeutic drug monitoring guide offers general background on dose timing and specimen context. It is background, not a substitute for validating a methamphetamine assay or establishing a driving threshold.

Frequently asked questions

Did 0.42 mg/kg oral d-methamphetamine significantly impair overall simulated driving?

Not in the primary reported overall weighted error comparison: mean scores were 102.4 after drug and 95.3 with placebo, with p = 0.19. This nonsignificant difference does not establish universal safety, equivalence, or the absence of a specific person’s impairment.

Did the study show that methamphetamine improves emergency driving?

No. Freeway emergency speed was 103.0 versus 106.1 km/h under drug versus placebo, with a nominal unadjusted p = 0.02, but there were 35 component tests, both mean speeds exceeded 100 km/h, and the study did not validate this single component as safer driving or reduced collision risk.

Were red-light violations significantly higher after methamphetamine?

The red-light nonstopping score averaged 4.1 after drug versus 1.0 with placebo, but the paired comparison yielded p = 0.11. It is a numerically higher sample score, not an established statistically significant increase in a population violation rate.

Were methamphetamine concentrations directly measured in blood or only modeled?

The investigators collected actual venous blood and oral fluid and used specific GC–MS analysis. Group means in the reported blood matrix were 72, 67 and 59 ng/mL at the three reported intervals. The article does not justify equating the incompletely characterized matrix with every forensic whole-blood reporting system or converting group averages into individual effects.

Can the oral-fluid concentrations be converted into a driver’s blood concentration?

Not reliably from these group means. Oral-fluid averages were 285, 223 and 190 ng/mL with substantial dispersion. Salivary collection buffer, pH, timing and specimen-specific analytical calibration prevent assuming a universal individual saliva-to-blood relationship.

Why is the 0.02 p-value not enough to prove a reproducible driving effect?

Investigators examined 35 separate component variables, making chance findings more likely. The paper did not demonstrate that the freeway-speed result survives a suitably prespecified familywise correction; the composite effect was nonsignificant. The finding is exploratory and should be replicated.

Can this experiment establish whether a particular methamphetamine-positive driver caused a crash?

No. The experiment measured a small controlled group on a simulator after supervised oral dosing. A specific case requires valid chemical testing, actual collected specimen identity, timing, contemporaneous driving evidence, reconstruction and competing causes. A drug-positive sample alone neither proves nor disproves impaired driving or crash causation.

Bottom line: keep the chemistry, the simulator and the legal question separate

Silber and colleagues provide a narrowly useful experimental finding: after one monitored oral dose of d-methamphetamine, an overall driving error score was not significantly different from placebo, while an emergency freeway speed comparison reached nominal significance among numerous individual tests. Other observed differences—including a larger red-light stopping failure mean—remained nonsignificant. Drug concentrations were measured in reported blood and collected oral fluid, not converted into a validated personal impairment standard. Forensic interpretation should preserve every one of those boundaries.

The complete original source is Silber BY, Croft RJ, Downey LA, Papafotiou K, Camfield DA and Stough C, Human Psychopharmacology (2012), doi:10.1002/hup.1238. A case-specific conclusion requires the actual specimen, method, timing, behavior and surrounding evidence rather than an isolated group-level p-value.