Does Modafinil Outlast Caffeine for Overnight Aircrew Vigilance? A 32-Person Plasma Study

What did the aircrew trial actually ask?

The immediate scientific question was whether wake-promoting medication could reduce the decline in vigilance after a long waking day and during the early-morning circadian low. Rather than compare people who happened to have taken different products, the investigators assigned the same volunteers to modafinil, caffeine and placebo sessions. Randomization concerned the sequence of these conditions; neither stimulant group consisted of an entirely separate sample of aircrew. This within-person design can limit confounding by relatively stable individual traits such as habitual response speed, but it does not eliminate time-of-night effects, imperfect sleep-history control, test-learning, or missing sessions.

Two different types of question run through the paper. First, does a participant press buttons, follow visual targets or report sleepiness differently after a blinded active capsule? Second, what concentration of the tested compound is observed in a laboratory specimen several hours after administration? The investigators answered those questions with different instruments. They did not prove that a particular plasma concentration caused an individual’s task performance, nor did they measure cockpit judgment or an air-traffic safety outcome.

How many independent people supplied the data?

Thirty-two RNLAF employees participated, including 21 pilots and five women. Their mean age was approximately 35 years, with a reported standard deviation of 10 years; the range was 25–59 years. All subjects met the applicable military or civilian flight-medical fitness criteria used for eligibility. This was a selected occupational cohort, not 32 randomly sampled commercial drivers or patients in a forensic case.

Two participants completed only two of the three test days for operational reasons; both missed the caffeine condition. Their modafinil and placebo results remained eligible under the protocol. Consequently, there were 32 modafinil sessions, 32 placebo sessions and 30 caffeine sessions, yielding 94 treatment sessions rather than 96 complete triples. In addition, investigators excluded two outlying participants from the VigTrack parameter analysis because their extreme scores suggested difficulty understanding that task. The statistical denominators therefore should not be casually described as 32 complete paired observations for every endpoint.

Repeated PVT stimuli, joystick tracking intervals, sleepiness ratings and venous samples do not inflate the number of independent humans. The sample size is 32 people; the repeated observations are nested within people and trial sessions. This distinction is especially important when interpreting small p values derived from many observations in a crossover study.

What were the precise drugs, routes and timing?

Each participant was scheduled for three nonconsecutive laboratory visits, separated by a washout of at least seven days. The randomized sequences balanced the six possible orders of the three conditions. The capsules were designed to appear the same. The active treatments were 200 mg of oral modafinil on one visit and 300 mg of oral caffeine on another, with placebo on the remaining visit. These were single-dose conditions, not repeated dosing across an overnight shift and not a combined modafinil-plus-caffeine condition.

Participants arrived in the late afternoon, completed familiarization, and undertook their first primary baseline tests around 6 p.m. A second pre-dose observation occurred around midnight. Medication was administered just after midnight, defined as study time zero. Subsequent task blocks were conducted at approximately 1, 2, 3, 4, 6 and 8 a.m. Vital signs were assessed at specified times. Participants could continue their usual caffeine consumption until 5 p.m. on a test day, after which additional caffeine was prohibited. That restriction did not guarantee that caffeine was chemically absent at midnight.

The reported median time already awake at medication administration was about 17 hours, with a range of 15.5–20 hours. The experiment involved extended wakefulness, not a verified identical 17-hour period of deprivation imposed on every person. Sleep duration and waking times before the visit differed among subjects, and the authors identified that variation as a limitation.

Which physical specimens were measured, and what did the laboratory report?

The paper states that qualified personnel collected blood at midnight, and approximately three, six and eight hours after medication. Its published analytical results are explicitly identified as plasma concentrations in Table 2. That is a separated biological fluid, not automatically equivalent to an unseparated venous whole-blood concentration. The authors expressed caffeine as micrograms per milliliter (µg/mL) and modafinil as milligrams per liter (mg/L). Numerically, 1 µg/mL equals 1 mg/L, but the analyte identity and matrix remain essential.

The investigators describe analysis by an external qualified diagnostic laboratory. They do not give a complete forensic method-validation dossier in this report: there is no detailed instrument platform, specimen preparation, lower limit of quantification, calibration traceability, uncertainty budget or chain-of-custody documentation sufficient to transform these research results into a casework laboratory report. The data nevertheless represent actual chemical measurements and should not be dismissed as mere assumed pharmacokinetics. For more on why separated plasma and whole blood must be distinguished, see the site’s published biological matrices guide.

Why can plasma concentrations not be converted directly into a whole-blood legal cutoff?

Whole blood contains cellular components plus plasma. A plasma assay measures the analyte concentration in the separated fluid; a whole-blood assay measures the analyte in an unseparated specimen according to a validated method. Distribution between these components varies by compound, sample handling and analytical procedure. A figure such as 5.20 mg/L of modafinil in plasma must not be presented as a measured whole-blood modafinil value. Nor does it identify an equivalent whole-blood threshold for impairment.

The concentration samples also preceded or followed specific computerized tasks but were not taken continuously. The sparse four-time-point scheme cannot establish an exact individual maximum concentration, precise absorption curve, exposure area under the curve, or a unique point-by-point pharmacodynamic relationship with each task response. The investigators themselves discussed the possibility that the true modafinil peak occurred between the +3 and +6 hour draws. A proper specimen interpretation must preserve the sample time, material, units, analyte, method and laboratory’s actual validated reporting purpose.

This is why the published specimen collection guidance emphasizes collection-site and timing records. One cannot repair missing case-specific specimen details by borrowing a population median from a study of healthy aircrew.

How was vigilance measured instead of actual flying?

The investigators used three principal domains. The Psychomotor Vigilance Task (PVT) required rapid responses to intermittent screen stimuli; the primary outcomes were reciprocal mean reaction time and the number of lapses, defined in this trial as responses of 500 milliseconds or longer. The VigTrack combined joystick compensation for a moving target with a concurrent vigilance response to occasional shape changes, and produced reaction time, omission percentage and tracking-error measures. The Stanford Sleepiness Scale (SSS) was a self-report rating of how sleepy participants felt.

These endpoints measure different things. A PVT lapse is not a real runway incursion; joystick error is not automatically a measurable deviation from a flight path; a sleepiness rating is not a direct chemical test. A person may benefit more on a simple reaction task than on a complex divided-attention task. That possibility is visible here: the investigators reported a treatment difference for several response measures while finding no significant overall treatment main effect for VigTrack mean tracking error.

The tests took place in a controlled research setting. No actual cockpit flight, instrument approach, emergency maneuver or accident reconstruction was performed. Investigators cautioned that real operational demands and arousal differ from these standardized tasks.

What did the original statistical comparisons actually show?

Because the source used repeated-measures analyses, some degrees of freedom were adjusted for sphericity and are not simple integer counts of independent subjects. Several reported follow-up comparisons depend on time of assessment. The main-effect p value says whether treatment differed on average across relevant measurement occasions, whereas a treatment-by-time interaction says whether treatment differences evolved. Neither one directly yields a personal likelihood of safe task completion. The table keeps the source’s reported tests separate rather than inventing a pooled effect size.

Why did the psychomotor vigilance results favor modafinil later in the morning?

The clearest late difference concerned PVT reciprocal reaction time. Both modafinil and caffeine were associated with less deterioration than placebo from two hours after administration through the final scheduled observations. Yet the investigators found that modafinil also outperformed caffeine at the +6 and +8 hour PVT speed measurements. The number of PVT lapses showed both active treatments better than placebo across the same later observations, with a modafinil advantage over caffeine specifically at +8 hours.

That sequence is consistent with the broad expectation that the two compounds differ in pharmacokinetic duration, but the particular trial did not determine the exact duration of individual efficacy beyond the final sample. Its time course ended at approximately 8 a.m. The paper cites earlier published half-life estimates; these are background literature values, not half-lives independently calculated from the sparse plasma concentrations in this cohort.

Equally, a favorable PVT result at eight hours does not mean the person would make accurate threat decisions, avoid fatigue-related operational errors, or remain awake during every minute of an aircraft mission. PVT is designed to detect lapses and response slowing; it does not validate a case-specific pilot-fitness or criminal-driving threshold.

What happened in the more demanding tracking task?

VigTrack required participants to compensate continuously for a moving visual target while monitoring shape changes. Compared with placebo, both active drugs reduced the deterioration in its mean reaction-time measure at the +4, +6 and +8 hour assessments. Omission percentage showed a significant treatment effect, with a reported post hoc advantage for modafinil over placebo. At selected late times, both active conditions had fewer omissions than placebo.

In contrast, the overall mean tracking-error treatment main effect was not statistically significant (p=.392). There was a significant interaction of treatment with assessment time (p=.014). At +6 and +8 hours, modafinil showed less impairment of mean tracking error than placebo, while the paper did not identify a significant caffeine-versus-placebo tracking-error difference. This is a qualified, time-dependent result—not blanket evidence that both medications improved every component of divided attention.

The source excluded two participants from the VigTrack analyses after reviewing unusually extreme task results. Such exclusions may be defensible in an experiment, but they should appear in any expert assessment of denominators, quality controls and applicability. A forensic writer should not infer that the experiment produced perfect complete-case data for all 32 people and all three treatments.

Did subjective sleepiness track the objective results?

The SSS findings partly aligned with PVT and VigTrack, but not at every time and treatment contrast. Group differences in sleepiness ratings appeared at approximately +4, +6 and +8 hours. The investigators reported less sleepiness with modafinil than placebo at each of those assessments. Caffeine was associated with less sleepiness than placebo at +4 and +6 hours, but not as an established significant benefit at +8 hours. Modafinil also produced a lower sleepiness rating than caffeine at +8 hours.

Those are subjective scale observations. They help describe participants’ experience and support an effect on perceived fatigue within this study. They do not independently validate accurate instrument handling, make an individual witness more truthful, or establish how much caffeine or modafinil was present in a person’s whole blood. In a different setting, a person might feel capable of performing despite objective lapse risk. The present experiment did not quantify that possibility as a crash probability.

What does measurable caffeine before dosing reveal?

At midnight before study medication, the source’s caffeine plasma column reported a median of 1.30 µg/mL, with an interquartile range of 0.43–3.45 µg/mL. This is a useful reminder that instructions to stop consuming caffeine are not the same as proof of a zero analyte concentration. The protocol allowed usual caffeine products until 5 p.m. and then prohibited further use. Its baseline concentration reflects carryover from earlier exposures and individual elimination variability.

The existence of measurable baseline caffeine also cautions against describing the placebo visit as proof of a drug-free person in all respects. Placebo in the trial means that the participant received the blinded comparator capsule, not that every stimulant compound had been completely absent from the body. The modafinil plasma baseline median was zero; that source result should not be generalized into an all-drug negative toxicology screen.

For any litigation matter, a numerical concentration without the identity of the compound, the time of collection, the physical matrix and the laboratory’s method is incomplete evidence. The site’s published confirmatory testing guidance explains the distinction between laboratory confirmation and the separate interpretive question.

Can measured plasma levels explain exactly who performed well?

No individual performance classifier was established. The plasma table offers treatment-level medians and interquartile ranges at four collection times, while the study’s statistical performance results compare repeated task outcomes across experimental sessions. It does not publish a validated personal concentration-response equation that can accurately classify one aircrew member as impaired, safe, fatigued, or sufficiently vigilant. It also does not quantify a modafinil-versus-caffeine equivalence point in plasma.

Nominal oral dose, measured plasma exposure, pharmacological effect and complex occupational performance are related but not interchangeable. An eight-hour plasma result is not equivalent to eight hours of proven safe functioning. Differences in habitual caffeine use, wakefulness duration, sleep debt, tolerance, metabolic disposition, circadian timing, task interest and other factors may all modify performance. The experiment shows treatment-associated group differences under a particular protocol, not how to reverse-engineer the exact cause of a specific person’s conduct.

How important are the missing observations and the small sample?

The study had useful methodological strengths: randomized treatment order, identical-looking capsules, two baseline opportunities, seven-day washout, repeated objective tasks, subjective ratings and actual plasma drug quantitation. It also had limits that directly affect interpretation. Thirty-two volunteers is a relatively small sample for predicting uncommon effects or identifying distinct subgroups. Only five were women, so sex-specific effects could not be established with useful precision. Two participants had no caffeine visit. Two further VigTrack outliers were removed for that specific analysis.

The participants’ waking times and preceding sleep durations were not made identical. The study was conducted with healthy, screened, fit-to-fly employees rather than older people with comorbid illnesses or members of the general driving public. Medication guesses at the end were correct in 54 of 94 instances, rather than every condition being consistently identified; this offers some information about blinding but does not establish perfect expectancy control. The original article reported no adverse events during these brief sessions and no drug-related vital-sign changes. A small overnight experiment cannot establish long-term safety or exclude rare complications.

Why is a controlled test not an actual aviation-safety validation?

The PVT and VigTrack are useful controlled measures of selected vigilance and divided-attention capabilities. The Royal Netherlands Air Force setting makes the scientific question operationally meaningful, but employees were completing standardized laboratory tests, not flying aircraft at night. The experiment did not expose subjects to real air-traffic coordination, changing weather, emergency communications, instrument failures, control handoffs, or a flight crew’s shared decision making. It did not observe a crash, a near miss or a legally adjudicated impairment event.

The source itself notes that simpler test conditions may exaggerate how readily an intervention’s benefit translates into complex real-world operations. The relevant conclusion is that selected measures of vigilance benefited under trial conditions; the separate assertion that a particular medication eliminated an individual’s aviation risk was not studied. Any case-specific opinion would require independent facts about the event, actual drug administration, sleep history, time course, medical conditions, observed performance and any specimen tested.

How should this evidence be handled in a forensic report?

An evidence-grounded analysis can state which controlled conditions differed, name the exact behavioral endpoint, preserve the difference between the whole-person N and repeated measurements, and identify whether a specimen was plasma or whole blood. It should distinguish a treatment-by-time interaction from a treatment main effect and report genuine nonsignificant findings alongside positive results. It should also keep the reported plasma medians in their original units and avoid inventing individual pharmacokinetic reconstructions.

The report should not equate modafinil with amphetamine, describe caffeine and modafinil as a combined administered treatment, call a plasma value a whole-blood concentration, or infer that a high vigilance score validates accident-free flying. The study provides no individual safe-driving threshold, no incident-causation probability and no instrument-validated forensic concentration cutoff. The absence of those endpoints reflects the scope of the protocol rather than a nonsignificant statistical finding.

Frequently asked scientific and forensic questions

1. Did all 32 participants complete every modafinil, caffeine and placebo condition?

No. Two subjects missed caffeine sessions because of operational scheduling, leaving 32 modafinil and placebo sessions but 30 caffeine sessions. Two further participants were excluded specifically from VigTrack analyses as task outliers. These are different missing-data issues and should not be combined into a fictional complete-case denominator.

2. Was the modafinil concentration measured in whole blood?

No. The published results table expressly reports plasma modafinil in mg/L and plasma caffeine in µg/mL. Plasma findings cannot be copied into a whole-blood case report without the appropriate validated specimen conversion evidence.

3. Did 200 mg of modafinil outperform 300 mg of caffeine at every hour?

No. Both active drugs differed from placebo on selected outcomes at multiple observations. The modafinil-versus-caffeine advantage was clearest in the later PVT measurements at six and eight hours and in sleepiness ratings at eight hours. The source does not establish uniform superiority on every task at every assessment.

4. Was VigTrack tracking error significantly improved overall by either treatment?

The treatment main effect for mean tracking error was nonsignificant (p=.392). A treatment-by-time interaction was significant (p=.014), and late modafinil-versus-placebo contrasts favored modafinil. These statements can both be true because they test different aspects of the data.

5. Did the study measure a blood modafinil level capable of proving one person’s impairment?

No. It measured plasma levels as research pharmacokinetic observations, but it did not validate a concentration-based rule classifying a person’s legal impairment, attentional capacity, fitness to fly or accident causation. No individual decision threshold was supplied.

6. Did a caffeine placebo condition mean no caffeine was present?

No. Placebo refers to the blinded test capsule. The source reported measurable caffeine at the pre-dose observation, consistent with the allowance for usual caffeine consumption until late afternoon. A placebo visit is not a universal negative chemical screen.

7. Can the results establish that a particular fatigued pilot should have been allowed to fly?

No. The study assessed standardized reaction tasks and sleepiness scales among medically screened air-force employees during controlled overnight visits. It did not test actual flying, operational mishaps or personalized safety thresholds. Individual fitness and event causation require separate factual and medical evidence.

Bottom line: a late vigilance advantage, not an individual safety certificate

The authors provide useful controlled evidence that a single oral modafinil dose and a single oral caffeine dose reduced selected overnight vigilance decrements compared with placebo, with a stronger late modafinil advantage on PVT outcomes and sleepiness. They also directly measured concentrations of each drug in separated plasma, which anchors exposure more firmly than relying on nominal doses alone. The scientifically important limits are equally clear: variable and missing visits, incomplete plasma time-course sampling, a genuinely nonsignificant VigTrack tracking-error main effect, a healthy occupational cohort, and no actual flight or case-specific impairment test. The original paper remains available through its primary DOI record. Accurate forensic interpretation requires stating both what this trial demonstrated and what it did not investigate.