Can Plasma Dexamphetamine or Zolpidem Predict Attention? A 60-Man Controlled Trial

Direct answer: No dependable individual prediction was demonstrated. In a randomized, double-blind trial of 60 healthy men, researchers measured dexamphetamine or zolpidem in separated venous plasma and tested attention, reaction time, and working memory before dosing and at three and eight hours afterward. Both drug groups made fewer sustained-attention omissions than placebo at eight hours, and the zolpidem group showed shorter simple reaction times. Yet the investigators found no detectable correlation between plasma drug concentration and the cognitive outcomes studied. A drug concentration in plasma is therefore not interchangeable with a determination that a particular person was alert, impaired, or fit to drive.

The distinction is especially important because the trial compared a stimulant with a sedative–hypnotic. The label assigned to a drug did not straightforwardly predict every laboratory-task result. Nor did a concentration measured at a scheduled sampling time supply an individualized performance threshold. The original experiment is Pfoser-Poschacher and colleagues, Effect of a Single Oral Dose of Dexamphetamine or Zolpidem on Attention and Reaction Time in Healthy Men, Fundamental & Clinical Pharmacology (2026), DOI: 10.1111/fcp.70092.

What question did the researchers actually test?

The central question was whether a single, controlled oral dose of dexamphetamine or zolpidem changed narrowly specified computerized cognitive endpoints and whether the observed changes tracked drug concentration in plasma. Sustained attention, basic visual reaction time, phasic alertness, and a short working-memory task were measured with Psytest. Subjective experience, pulse, blood pressure, and drug tolerability were assessed separately.

That is more limited than asking whether a driver operated a vehicle safely, whether a witness recalled an actual event accurately, or whether a toxicological concentration identifies the cause of an accident. No road test, vehicle simulator outcome, collision investigation, or forensic case reconstruction was part of this trial. Even a statistically significant effect on one laboratory endpoint should not be automatically promoted into a global finding about real-world competence.

Who participated, and what does the sample size mean?

Sixty-four prospective participants were evaluated; 60 healthy men were enrolled and randomized. Their ages ranged from 20 to 43 years, with an average of 26 years and standard deviation of five years. The three mutually exclusive treatment groups contained 20 men each, and the authors reported no study dropouts. The trial was a parallel-group experiment, not a study in which the same man received all three drugs.

Each participant completed repeated measurements at baseline, three hours, and eight hours. Those observations are correlated within the same person. Sixty biological participants did not become 180 independent people because three assessments were made. Similarly, multiple questions within a cognitive battery cannot be treated as independent human replication. The sample also excludes women, older people, and patients with many relevant medical or substance-use histories. Extrapolating this experiment to chronic users, multidrug exposure, or a heterogeneous forensic population would require additional evidence.

What doses were administered, and when were measurements made?

Participants received a single encapsulated oral treatment: 30 mg dexamphetamine, 5 mg zolpidem, or placebo, with 250 mL of water. Randomization and masking were intended to limit expectation and observer bias. Cognitive tests were performed approximately one hour before dosing and again three and eight hours after administration. Scheduled venous samples were collected before dosing and at the two postdose testing times.

The dose labels describe what researchers administered, not what was later recovered from a specific blood specimen. Thirty milligrams swallowed is not a concentration of 30 ng/mL in blood, and the zolpidem dose does not define how much drug reached the brain at any particular moment. Meals were supplied after two hours and during the later postdose period. Timing, absorption, distribution, metabolism, measurement, and individual variability all separate an administered amount from a sampled concentration.

Which specimen was chemically analyzed?

The investigators obtained venous blood into EDTA-containing collection tubes, but they centrifuged the blood and analyzed the separated plasma. The samples were stored frozen before batch analysis. Sample preparation included methanol-based protein precipitation and solid-phase extraction. The analytical system used reversed-phase liquid chromatography coupled with high-resolution mass spectrometry, with external calibrators prepared in reference plasma.

This is direct chemical quantitation of the named parent drugs in plasma. It is not quantitation of an unseparated forensic venous whole-blood sample. Plasma and whole blood are different physical matrices: cellular partitioning can affect the relationship between their concentrations. The authors did not generate a validated plasma-to-whole-blood conversion equation for this study. They also did not report corresponding quantitative urine, oral-fluid, hair, breath, vitreous, milk, or tissue concentrations. Analytical identification and behavioral inference must be treated as distinct operations.

What were the observed plasma concentrations?

Group-average concentrations declined between three and eight hours for both active treatments. Dexamphetamine was present at a higher average numerical concentration than zolpidem at both sampling points, but these are different drugs with different pharmacology. A direct numerical comparison of their ng/mL values is not a potency or impairment ranking. The reported variability also matters: the zolpidem group had substantial between-person dispersion at three hours.

Active-drug concentrations are means ± standard deviations, not medians, percentile limits, legal cutoffs, or standard errors. Group assignment sizes are shown; the paper does not provide an individualized concentration-performance rule. A single reported average cannot resolve when any unrelated person consumed a medication or how that person performed at an unobserved time.

Did either drug change sustained attention?

The sustained-attention task required recognizing repetitions or matches among presented visual stimuli. Errors and missed targets were tallied. Relative to placebo, the baseline-adjusted difference in omissions at three hours was not statistically significant for either active treatment. At eight hours, the estimated differences favored both: dexamphetamine was associated with 3.90 fewer omissions and zolpidem with 3.08 fewer omissions in the model comparison.

This is an endpoint-specific finding, not proof that zolpidem generally improves safe real-world behavior. A participant might miss fewer targets in this tightly controlled experiment while differing in other capabilities not evaluated here. The source’s eight-hour zolpidem comparison was also close to the conventional 0.05 threshold. The investigators examined multiple drugs, time points, and outcomes; they described their p-values without a global multiplicity correction. Interpretation should follow the size, precision, study design, and scope of each comparison instead of reducing the trial to a collection of significant labels.

What do the placebo-adjusted comparisons actually show?

The original analysis adjusted for baseline values and compared changes in each active-treatment group with the placebo group. Negative values for omissions or errors indicate fewer missed responses or errors; negative values for median reaction time indicate faster responses, not necessarily worse performance. The selected contrasts below retain the source’s measurement units, estimated confidence intervals, and reported p-values.

These are model-based group contrasts, not individual changes or diagnostic sensitivities. The authors conducted additional comparisons not reproduced in this selected table; the underlying original article remains the authority for the full endpoint set. Importantly, the lack of a statistically significant result does not prove two treatments were equivalent, and the presence of a statistically significant difference does not establish legal impairment.

Why did simple reaction time and phasic alertness tell different stories?

Simple visual reaction time measures speed in answering a straightforward stimulus. Phasic alertness represents a separate component of readiness after a warning signal. These domains should not be combined simply because both fall under a broad heading of attention. The researchers found shorter simple reaction times in the zolpidem group relative to placebo at three and eight hours, but not a significant dexamphetamine-placebo reaction-time difference at either time.

The quantitative post hoc comparisons for phasic alertness were not significant for either drug at three or eight hours. This is important because sections of the source’s narrative discuss zolpidem-associated impairment in phasic alertness even while the numerical table and main Results state that no group changed significantly. The measured table should take precedence over an unsupported emphatic description. The public conclusion should therefore identify the inconsistency cautiously, without manufacturing a significant effect that the reported comparison does not demonstrate.

What happened in the working-memory test?

The working-memory component required retaining numbers in sequence and determining whether an item matched one shown earlier. For this tested version of working memory, the number of omissions and errors did not show a statistically significant treatment effect at the postdose measurements. A participant’s general capacity for memory, concentration under fatigue, complex professional decisions, or handling competing traffic hazards cannot be deduced from these limited null results.

Consider what would be required for a larger statement. A genuine broad cognitive assessment would need multiple tasks that measure different domains, a sufficiently representative population, clinically or legally meaningful performance criteria, and statistical controls adequate for multiple comparisons. This nine-page original report did not provide an external validation that a few Psytest endpoints classified individual unsafe drivers. The null working-memory results are real scientific data, but they are neither a certification of normal functioning nor an absence-of-risk guarantee.

Was higher plasma concentration associated with better performance?

The authors reported no detectable correlation between plasma concentration of either active drug and the cognitive-performance measures considered. This is the most consequential negative finding for many proposed forensic interpretations. The trial contained actual chemical measurements as well as actual laboratory performance outcomes, so it did not merely assume exposure from self-report. Nonetheless, the available data did not yield a demonstrated concentration-response relationship for the tested endpoints.

Failure to detect a correlation is not evidence that concentration could never matter. With 20 men assigned to each active arm, two scheduled postdose concentrations, particular doses, and a finite number of tests, the study has limited range and statistical power for characterizing individual nonlinear responses. It did not validate a diagnostic threshold, estimate classification accuracy for intoxication, or define a probability that a specific person with a specified plasma number was safe. The correct inference is that this experiment did not demonstrate a useful individual concentration-to-Psytest association.

Why is the separation between plasma and whole blood legally important?

A forensic toxicology report must name the physical specimen actually tested. A hospital or research laboratory may analyze serum or plasma after removing cells, while an impaired-driving statute, laboratory reporting policy, or expert-retrograde calculation might refer to unseparated venous whole blood. Those concentrations can differ because the blood-cell and fluid fractions do not contain identical analyte distributions. A result reported as plasma ng/mL cannot simply be relabeled whole-blood ng/mL.

This study did not analyze paired plasma and whole-blood aliquots to establish a drug-specific correction factor. It also did not evaluate an oral-fluid collection kit or a urine screening threshold. For the broader matrix question, see biological matrices in forensic toxicology. The appropriate specimen, validated analytical method, collection timestamp, calibration evidence, and preservation record should be obtained before interpreting a concentration in litigation. A source paper measuring plasma alone cannot settle all of those case-specific questions.

Does LC–MS measurement prove drug identity, dose, timing, or impairment?

Liquid chromatography–mass spectrometry can be highly informative for compound identity and quantitation when properly validated. In this trial, it quantified the assigned drugs in separated plasma. That strength does not make the technique a detector of cognitive impairment. Concentration measurements are chemical observations; performance tests are behavioral observations. Neither alone identifies the mechanism behind an event that researchers did not witness.

In a forensic case, the analytical record should distinguish screening from confirmation and explain target analyte, specimen, calibration, internal and external quality controls, detection range, uncertainty, collection time, and sample integrity. The general principles are developed in this site’s confirmatory drug testing guide. The laboratory record—not a convenient cross-study number—governs analytical interpretation. A controlled 30 mg dose does not provide a universally valid backward calculation for a stranger’s unknown dose or exact last use.

What did the participants report feeling?

The subjective Drug Effect Questionnaire provided a different picture from the narrow objective cognitive results. At three hours, the median perceived effect was 32% in the dexamphetamine group, compared with 10.5% for zolpidem and 4% for placebo. The median drug-liking rating was 47% after dexamphetamine at the same time. At eight hours, the dexamphetamine group’s median perceived effect was still 12.5%, whereas the other groups’ medians were zero.

These subjective ratings should not be substituted for measured task accuracy, measured drug concentration, or a legal state of intoxication. People may report liking a stimulant when the studied reaction-time contrast is nonsignificant. Conversely, a reduced error count on a laboratory task does not establish that the person feels normal or can drive safely. The study was designed to observe such differences, not to declare a validated relationship among subjective effect, pharmacokinetics, and real-world safety.

What cardiovascular and tolerability findings matter?

At three hours, mean arterial pressure and pulse rate were elevated in the dexamphetamine arm. The table reported mean arterial pressure increasing from 94 mmHg at baseline to 104 mmHg and mean pulse from 68 to 75 beats per minute. Four men receiving dexamphetamine reported tachycardia and/or palpitations, and some reported other transient events. No adverse effects were reported in the zolpidem arm within this brief monitored experiment, which should not be mistaken for evidence that zolpidem is risk-free.

These hemodynamic and tolerability measures are important clinical context but do not constitute quantitative blood-drug testing or documented driver behavior. Acute transient changes need not map to the task endpoint that reaches statistical significance at eight hours. Moreover, investigator-rated tolerance and participant-perceived tolerance were not identical assessments. A cautious reading separates observed vital signs, reported symptoms, laboratory performance, and later theoretical claims about risk.

Why do apparently favorable zolpidem results require particular restraint?

Zolpidem is normally classified as a sedative–hypnotic, yet this 5 mg daytime laboratory trial showed some favorable changes in missed targets and simple responses at selected time points. The experiment does not negate the possibility of sedation, complex behavior, residual effects, or individual vulnerability in other settings. Its narrow dosing regimen, all-male healthy sample, limited observation window, repeated testing, and small group sizes prevent such a sweeping generalization.

The zolpidem plasma averages were also below concentrations the authors discussed as typical in some therapeutic-monitoring contexts. Without higher-dose, repeated-dose, nighttime-use, or actual driving measurements, it would be unsound to argue that a drug’s familiar clinical category always dictates a specific Psytest score—or that one surprising Psytest result proves a medication categorically safe. It would be equally unsound to portray a nonsignificant phasic-alertness comparison as a significant impairment. The data demand endpoint-by-endpoint interpretation.

How should statistical uncertainty change the forensic interpretation?

The study was powered around a projected difference in reaction time, not around proving an individual concentration threshold or a dose-specific driving outcome. Baseline-adjusted analyses compared changes from baseline, with Dunnett post hoc contrasts. Numerous endpoints and times were examined and the investigators indicated that p-values were presented descriptively without a comprehensive multiplicity adjustment. A borderline p = 0.0495 for one omissions contrast is therefore a signal to interpret carefully, not an absolute law of human behavior.

Confidence intervals indicate the precision available for the measured group differences. Some include zero and therefore do not support the corresponding conventional significance claim. Others exclude zero by only a small margin. Negative estimated differences are simply directions on the particular endpoint’s scale. The researcher must ask what was measured, whether improvement is clinically meaningful, how robust it is across tasks and populations, and whether findings hold under independent replication. The paper did not quantify individual predictive sensitivity, specificity, or an impairment decision boundary.

What can an expert responsibly use this study to explain?

A qualified expert can use it as a controlled demonstration that chemically measured plasma drug concentrations and measured cognitive effects are analytically distinct forms of evidence. The paper can also illustrate that a nominally sedating medication and a stimulant may both alter selected test outcomes in ways that do not correspond neatly to lay expectations. It is useful in evaluating overgeneralized testimony that a measured number alone establishes an inevitable cognitive state.

The expert should not use this study to certify safe driving after zolpidem, declare stimulant intoxication irrelevant, translate plasma directly into whole blood, or infer a specific defendant’s dose, ingestion hour, crash contribution, or witness reliability. Such conclusions would demand case records, verified specimen matrices, uncertainty and assay documentation, independent behavioral evidence, and an appropriate causal analysis. Additional context on timed therapeutic measurements appears in therapeutic drug monitoring; a therapeutic reference range is not automatically a legal impairment line.

Seven frequently asked forensic-science questions

1. Did the study directly measure dexamphetamine and zolpidem?

Yes. Both active agents were chemically measured in EDTA plasma separated from venous blood, using chromatography coupled with mass spectrometry. The measured group averages are given above at three and eight hours. The study did not provide a corresponding numeric legal whole-blood panel or validated paired plasma–whole-blood conversion. Reporting the physical matrix is necessary before comparing its numbers with any forensic standard or another study.

2. Did a higher plasma level identify the men who made the most errors?

No demonstrated association of that kind was reported. The authors found no detectable correlation between their timed drug levels and the cognitive endpoints. This is a failure to detect an association in the studied conditions, not proof that every possible concentration is behaviorally irrelevant. A forensic prediction would require a separate validated model with individual-level performance, calibration, uncertainty, and population applicability.

3. Did zolpidem improve reaction time at every measured moment?

No. Relative to placebo, statistically significant reaction-time differences were reported at three and eight hours, but the baseline assessment preceded dosing and was not a drug effect. Phasic alertness was a different endpoint, with no significant drug-placebo contrast in the original numerical comparison table. These results do not warrant a general claim of faster, safer driving or improved complex decision-making after zolpidem.

4. Why might a drug influence attention but not working memory?

The Psytest battery defined those constructs with different tasks and scoring rules. Detecting or matching repeated stimuli is not identical to holding a short number sequence in mind. Drug effects can vary with task demands and time after administration. The actual trial found selected sustained-attention effects but no significant working-memory omissions or errors. The difference is a measured domain distinction, not a proven mechanism for every participant.

5. Can the trial establish how recently another person swallowed a pill?

No. Doses and sampling times were precisely controlled for these volunteers, but the study did not validate an unknown-dose inversion procedure or a person-specific ingestion-time estimator. A concentration obtained in an unrelated case might involve repeated administration, different physiology, mixed drugs, different sampling conditions, or an entirely different matrix. The paper’s three- and eight-hour group averages are not universal pharmacokinetic clocks.

6. Was driving performance tested?

No driving course, simulated vehicle-control endpoint, crash, or collision was assessed as an outcome in this primary trial. Psytest measures and subjective scales are laboratory observations. Although traffic safety provided background motivation, that does not convert a cognitive test into actual road performance. A case-specific impairment opinion would have to distinguish evidence of ingestion, evidence of effect, observed driving behavior, and evidence establishing causal contribution to an event.

7. What is the most defensible bottom line for litigation?

In 60 healthy men randomized among oral dexamphetamine, zolpidem, and placebo, some attention and reaction-time comparisons differed at scheduled times, whereas working-memory and several alertness comparisons did not. Plasma concentrations were directly assayed, yet no concentration–performance correlation was demonstrated. Those findings support cautious interpretation of specific laboratory data; they do not supply an individualized legal impairment threshold, a whole-blood equivalent, or proof of incident causation.

Original primary source

Pfoser-Poschacher V, Wolf M, Bayerle-Eder M, Wimmer T, Boryshchuk D, Gerner C, Wolzt M. Effect of a Single Oral Dose of Dexamphetamine or Zolpidem on Attention and Reaction Time in Healthy Men. Fundamental & Clinical Pharmacology. 2026;40:e70092. https://doi.org/10.1111/fcp.70092. Original trial registration: EudraCT 2021-005381-17. This article discusses study evidence for general scientific education; interpretation of a particular case requires that case’s own validated measurements and records.