Can Modafinil Preserve Vigilance During a 29-Hour Wake Shift Without Making People Feel Less Sleepy? An 18-Person Trial

Answer: In a small randomized, double-blind laboratory study, modafinil helped preserve selected objective attention and cognitive-throughput measures during repeated 28.57-hour scheduled waking periods and changing circadian phases. Yet participants did not report a corresponding improvement on the study’s main subjective sleepiness scales. A separate objective recording showed less inadvertent sleep during assigned waking time: approximately 1.1% of scored epochs with modafinil versus 2.2% with placebo. The study therefore supports a narrow conclusion about task-dependent laboratory performance under experimentally disrupted sleep timing, not a claim that someone receiving modafinil was awake, alert, fit to drive, or unimpaired in an individual incident.

This analysis addresses Grady, Aeschbach, Wright and Czeisler, Effect of Modafinil on Impairments in Neurobehavioral Performance and Learning Associated with Extended Wakefulness and Circadian Misalignment, Neuropsychopharmacology 35:1910–1920 (2010), primary research DOI 10.1038/npp.2010.63. The original full Methods, Results, illustrations, neurobehavioral ANOVA table, and Discussion—not an abstract alone—form the evidentiary basis.

What was the precise question?

Hours spent awake and the phase of the body’s internal clock are separate sources of risk to performance. Being awake for a long interval raises homeostatic sleep pressure. Working during a circadian low can degrade performance even when a person has not been continuously awake for an unusually long time. Real shift work combines both pressures, which makes it difficult to attribute a slow reaction time or lapse to just one cause. This laboratory protocol intentionally allowed the circadian oscillator to run out of alignment with imposed sleep and wake times.

The relevant question was not whether modafinil made people generally more intelligent or enabled safe activity indefinitely. It was whether a prescribed, repeated wakefulness-promoting intervention changed specific performance measurements while investigators independently characterized time awake and circadian phase. An additional question was whether any objective advantage was accompanied by subjects feeling less sleepy, measurable changes in sleep episodes, or poorer recovery sleep.

Who participated—and what does “N = 18” actually count?

Nineteen healthy adults began participating. One male assigned to the active-drug condition withdrew on the second active-treatment day after symptoms that the study physician judged possibly drug-related. The authors analyzed 18 completers: nine in the modafinil arm and nine in the placebo arm. Each analyzed arm included five men and four women. The analyzed participants averaged 23.9 years of age (SD 4.0; range 18–30). Those 18 independent individuals—not the hundreds or thousands of later test occasions—are the relevant biological sample size for comparing randomized groups.

The investigators screened physical and psychiatric suitability and used clinical chemistry, electrocardiograms, and drug screening near entry. Screening established eligibility; it was not serial quantitative testing of modafinil exposure. The group was young, monitored, healthy, and housed in a dedicated laboratory. These selection and environmental conditions matter: professional shift workers, patients with obstructive sleep apnea, commercial drivers, people with multiple medications, and individuals with uncontrolled illnesses were not the tested populations. Randomization protects comparison within this small laboratory, but it cannot supply data for those absent groups.

What was the unusual forced-desynchrony schedule?

Over a prolonged study that included baseline and recovery days, participants followed an imposed 42.85-hour “day” during forced desynchrony, with 28.57 hours scheduled awake followed by 14.28 hours allotted to bed/rest. Fourteen such artificially lengthened cycles extended through calendar days 4–29. Clocks and ordinary time cues were removed, and research personnel supervised adherence. This schedule is deliberately different from natural civilian shift work or a conventional 24-hour calendar day.

The internal core-temperature rhythm remained closer to a normal circadian period: estimated means were 24.11 hours in placebo and 24.23 hours in modafinil, with between-group p = .5. Because the internal rhythm did not entrain to the artificial 42.85-hour schedule, episodes of waking progressively occurred across differing biological phases. The authors divided observations into six internal circadian phase bins and six waking-duration bins. The result permitted a statistical distinction between time since scheduled waking and internal circadian timing. It did not identify a universal unsafe clock hour applicable to every individual.

How much modafinil was administered?

The active-treatment arm received a total of 400 mg of oral modafinil per 42.85-hour scheduled cycle, split into 100 mg at scheduled waking, 100 mg after approximately 9.58 hours awake, and 200 mg after approximately 19.16 hours awake. Placebo participants received matched tablets on the same schedule. The authors described a randomized, double-blind, parallel-group comparison, not a three-way crossover; no participant switched between active drug and placebo during these randomized periods.

“400 mg” is an administered amount, not a laboratory finding of “400 ng/mL,” a measured whole-blood concentration, or an instruction for an individual worker. Nor is the 400 mg per artificially long study cycle automatically interchangeable with a calendar-day prescription. Drug tablets were administered under the experimental schedule, and effects were assessed through neurobehavioral and physiological instruments. The study did not quantify individual serial modafinil in venous whole blood, separated plasma, serum, oral fluid, hair, urine, or exhaled breath and did not establish a forensic concentration–impairment cutoff.

Which outcomes were directly measured?

Every two hours during scheduled wakefulness, a roughly 30-minute test battery examined mathematical throughput, digit-symbol substitution, probed recall memory, reaction-time vigilance, and a fine-motor trackball task. The ten-minute Psychomotor Vigilance Task (PVT) generated median reaction time, the slowest and fastest reaction-time deciles, and lapses defined by responses longer than 500 milliseconds. These were direct behavioral task measurements—not direct chemical concentrations, blood tests, or actual road-driving outcomes.

The Karolinska Sleepiness Scale (KSS) and visual-analog scales measured subjective sleepiness, alertness, and an attentive–dreamy dimension. Polysomnography recorded sleep-stage activity during assigned rest periods; during scheduled waking, electroencephalogram and eye-movement channels helped identify inadvertent 30-second sleep epochs and slow eye movements. Core body temperature was recorded by rectal thermistor. These modalities measure different constructs. A person can complete a reaction-time task relatively well while reporting sleepiness, or subjectively feeling awake while producing objective lapses. Neither measure alone certifies safe real-world operation.

What was the strongest objective performance evidence?

The study’s principal evidence is a pattern of repeated-measures effects, not one pooled “modafinil improved everything” score. In the original numerical table, the Addition Task showed a drug-group main effect, F(1,16) = 6.9, p = .02. Drug-by-waking-duration interactions appeared for PVT slowest 10% of responses, F(5,80) = 4.5, p < .01; PVT median reaction time, F(5,80) = 4.5, p < .01; PVT lapses, F(5,80) = 3.4, p < .01; DSST, F(5,80) = 2.6, p = .03; and trackball losses, F(5,80) = 4.2, p < .01. The advantage tended to emerge as time awake increased.

There were also drug-by-circadian-phase interactions on addition accuracy, F(5,80) = 3.3, p < .01, and PVT median reaction time, F(5,80) = 4.5, p < .01. The effect was especially evident near the post-minimum portion of the body-temperature rhythm in this imposed protocol. Because these are interactions, the proper interpretation is that the difference between treatment groups changed across wakefulness duration or circadian phase. It is not valid to turn an interaction p-value into a constant magnitude of improvement for every hour, subject, and task.

Which effects were genuinely nonsignificant?

Several findings actively resist an overbroad enhancement narrative. The main drug effect on PVT median reaction time was F(1,16) = 1.2, p = .30, whereas the interaction with hours awake was significant. PVT lapse count showed a nonsignificant drug main effect, F(1,16) = 2.0, p = .17, again despite an interaction with waking duration. DSST had a drug main effect of F(1,16) = 0.1, p = .93. Probed Recall Memory had F(1,16) = 0.3, p = .61 for the main drug comparison, with drug-by-wake p = .50 and drug-by-phase p = .88. The fine-motor trackball main effect was F(1,16) = 0.1, p = .80.

The authors additionally noted a near-threshold drug-by-phase pattern for the PVT fastest 10% of responses (reported around p = .057 in the prose), not a conventionally significant result. These distinctions are important in testimony and scientific writing: nonsignificant means the specified comparison did not meet the selected evidentiary threshold in this sample, not that no possible effect exists. The contrast between main effects and interactions underscores why a single “impaired/not impaired” label cannot be reconstructed from aggregate task statistics.

Evidence table 1: drug effects on measured performance

This reorganization separates main drug contrasts from interaction effects; values refer to different model terms and cannot be pooled into one effect size. The original authors transformed certain PVT outcomes before statistical testing because raw reaction-time distributions were not normal. Figures display some raw descriptive measures, while the inferential tests may use their transformed counterparts. When evaluating opposing expert statements, the transform and relevant denominator degrees of freedom should be identified rather than inferred from a figure alone.

Did the medication prevent inadvertent sleep?

The strongest direct physiological contrast was objective inadvertent sleep during assigned waking: approximately 2.2% of scored epochs in placebo participants versus 1.1% under modafinil, with a drug main effect F(1,16) = 5.7, p = .03. Drug-by-waking-duration also reached significance (p < .01), meaning the treatment difference changed across the prolonged waking period. The exact 30-second scoring design matters. A figure derived from an epoch percentage cannot be turned into the probability that a particular driver will fall asleep at the wheel, because the laboratory denominators, monitoring, task demands, and consequences differ dramatically.

In contrast, the frequency of slow eye movements recorded in wakefulness showed no significant main drug effect, F(1,16) = 0.1, p = .81. Neither biomarker alone settles an individual state. A reduced proportion of scored sleep epochs may be consistent with better sustained waking in the study, but it does not prove elimination of microscopic lapses, microsleep in every situation, or the absence of other drug effects. Importantly, the trial did not make physicians or drivers take a blind real-world operation test during the hours portrayed in the graphs.

Why did subjective sleepiness fail to track the objective changes?

On the KSS, the main drug comparison was F(1,16) = 0.2, p = .64. The visual-analog “Alert–Sleepy” rating also produced F(1,16) = 0.2, p = .64. Those outcomes are not statistically significant and should not be rewritten as evidence that every participant felt restored alertness. The authors discussed a more complex attentive–dreamy rating: the source table rounds its main drug term to p = .05, while the prose describes p = .051, and a three-way drug-by-wake-by-phase interaction met p < .01. That scale’s direction and day-specific details do not support a simplistic “always feels more alert” summary.

This objective–subjective separation is one of the trial’s most useful forensic concepts. The relevant question for actual capacity is not merely whether a person took a wakefulness-promoting medication or said they felt normal. Laboratory reaction-time behavior, subjective feelings, physiological sleep recordings, opportunity for rest, and a person’s actual assigned task each contribute different information. The trial did not establish which measure best predicts a collision, medical error, aircraft mishap, or a specific worker’s reliable judgment under legal scrutiny.

What about recovery sleep and other physiological findings?

During scheduled sleep, the study found no statistically significant between-group treatment effects on total sleep time, sleep efficiency, scored sleep stages, sleep latency, or waking after sleep onset. This is a statement about the variables, dose timing, small sample, and laboratory period actually tested. It does not prove that modafinil can never disrupt sleep in other schedules, populations, doses, or combinations. It likewise does not license a conclusion that someone’s nightly sleep was adequate because they used a stimulant.

Body-temperature analyses suggested treatment-related variation by behavioral state—higher average core temperature during scheduled waking and lower during scheduled sleep relative to placebo—while circadian temperature phase parameters were not simply shifted by treatment. The authors considered whether such changes might contribute to measured performance, but did not demonstrate a causal physiological pathway. Core temperature is not drug concentration, psychomotor capacity, or a validated stand-alone individual forensic intoxication marker.

Evidence table 2: objective sleep and subjective experience

Were there adverse events or withdrawals?

One participant assigned modafinil withdrew on the second active-treatment day after nausea, abdominal symptoms, and other complaints; the attending physician considered a possible relation to medication. The original description also records a participant whose pre-existing intermittent tinnitus worsened during and after the study; investigators judged that event unrelated to the medication. Two non-serious events were reported to the institutional review board, and the authors reported no serious adverse events in this small controlled experiment. Importantly, the completed-group analysis of 18 excludes the early withdrawal, so an assertion that all randomized drug recipients tolerated treatment would be inaccurate.

A small number of observed adverse events does not quantify rare risks or establish safety for patients with conditions excluded from the study. It is also not an independent randomized comparison of accident prevention. For source fidelity, the distinction between the 19 initial participants and 18 completers matters when an expert presents a denominator or claims a percentage of people able to tolerate repeated modafinil.

Does prolonged administration imply tolerance or a learning benefit?

The investigators reported a significant drug-by-scheduled-wake-episode interaction for addition-task performance across the repeated forced-desynchrony cycles (p < .00005). The modafinil group maintained higher throughput relative to their own baseline after multiple cycles, while placebo participants did not regain baseline addition output during the protocol. This outcome is plausibly relevant to learning and repeated task practice under imposed sleep disruption, but it was not a direct test of classroom learning, occupational training, or long-term clinical executive function.

Practice effects, experiment scheduling, and the physiological condition all shape this finding. The authors noted that other neurobehavioral endpoints did not show the same across-cycle improvement. It would therefore be incorrect to tell a jury that modafinil universally prevents cognitive fatigue, develops a person’s competence, or eliminates the consequences of circadian misalignment. Equally, this study does not directly demonstrate dose escalation, pharmacokinetic tolerance, or a person-specific drug concentration trajectory across the 25-day experiment.

What did the study never measure about forensic toxicology?

Despite its pharmacologic treatment, the reported endpoints were primarily performance, self-ratings, sleep physiology, and temperature. There were no measured postdose modafinil concentrations in directly sampled forensic whole blood, separated plasma, serum, urine, oral fluid, hair, or vitreous humor reported as the basis for the behavioral findings. Baseline drug screening was for participant eligibility. Accordingly, the dataset cannot translate an administered tablet amount into the numeric drug concentration in another person’s specimen. It offers no matrix conversion factor, postmortem redistribution finding, specimen stability validation, legal per-se cutoff, or toxicological estimate of a driver’s ingestion time.

Nor did it establish event-specific capacity. The PVT is a laboratory reaction-time procedure, not a calibrated equivalent of vehicle steering, braking amid hazards, navigation, aviation systems management, medical judgment, or a legal definition of impairment. An observed group mean or treatment interaction cannot identify the causal explanation for a named person’s conduct without independent exposure verification, clinical circumstances, actual sleep history, relevant observed behavior, and event-specific evidence. In a case involving prescribed modafinil, medication history is a relevant factual variable—not proof of either safety or impairment.

How should a forensic expert describe the uncertainty?

State the design first: 18 analyzed healthy adults, nine per parallel treatment group, many repeated observations, and an intentionally extreme 42.85-hour artificial day. Next identify which statistic is a main effect versus a drug-by-wake or drug-by-circadian-phase interaction. Present source-recorded null findings as explicitly as the positives; do not replace a nonsignificant p-value with the sentence “there was no effect” or invent an effect estimate. Emphasize that the numerical ANOVA table describes nested repeated observations, not hundreds of independent subjects.

Explain the validity limits of extrapolation. The laboratory environment controlled time cues, scheduling, and recording, whereas real incidents commonly involve variable chronic sleep restriction, shift rotation, stress, co-medications, individual disease, distraction, and incomplete retrospective evidence. Multiple endpoints and small treatment groups also limit precision. The article’s author affiliations and documented manufacturer-related relationships belong in a transparent assessment of study context, but those conflicts are not substitutes for examining the actual methods and numerical results.

Seven evidence-based questions

1. Did modafinil prevent all errors or lapses during very long wake periods?

No. The randomized group differences and interactions favored active treatment for several tasks under selected waking durations, and inadvertent sleep occupied a smaller fraction of observed epochs. However, nonzero lapses remained, several main drug effects were nonsignificant, and the study never tested an error-free worker or driver.

2. Why can PVT reaction time improve while subjective sleepiness is unchanged?

They measure different outcomes. PVT is a timed behavioral response; KSS and the alert–sleepy VAS record how participants rate their own state. The study observed significant PVT interaction terms alongside nonsignificant drug-group main effects for the principal subjective scales. One is not a validated substitute for the other.

3. Was the administered modafinil amount a blood concentration?

No. The protocol delivered 100 mg, 100 mg and 200 mg orally within an imposed long waking cycle. Milligrams administered are not nanograms per milliliter in venous whole blood, plasma, or another body fluid. A drug concentration would require an actual documented sample and analytical assay.

4. Did this trial prove modafinil works for every shift worker or sleep-apnea patient?

No. The sample comprised selected healthy young participants monitored in an artificial time-isolated laboratory. Patients with particular diseases and people facing ordinary operational shifts were not investigated in the same way. Even within the trial, tasks and subjective outcomes were not uniformly improved.

5. Were wakefulness duration and circadian phase interchangeable?

No. The purpose of forced desynchrony was to disentangle time awake from internal biological phase. The internal temperature cycle continued near 24 hours while scheduled cycles lasted 42.85 hours. Interactions with these two factors describe different scientific relationships.

6. Did modafinil leave recovery sleep unchanged forever?

No. Investigators reported no significant treatment-group differences in measured scheduled-sleep variables during this particular protocol. That small finding cannot rule out different effects under other doses, medication combinations, schedules, or clinical conditions.

7. Can this study determine whether a specific incident was caused by fatigue or medication?

No. It can inform a general scientific assessment of how sleep pressure, circadian timing, and a wakefulness-promoting agent interact with some tasks. Attribution in an actual collision or alleged occupational error requires additional individual evidence. The trial did not validate a person-specific impairment rule, a legal drug cutoff, or a causal reconstruction method.

Bottom line for forensic interpretation

The strongest supported conclusion is a carefully delimited one: in the Grady laboratory protocol, repeated modafinil changed some objective task trajectories under prolonged scheduled waking and circadian misalignment while leaving important memory and subjective-sleepiness comparisons statistically nonsignificant. The active group also had fewer scored inadvertent sleep epochs during scheduled wakefulness, without a detected decrement in the measured assigned-recovery-sleep variables. Those findings warrant consideration as group evidence about selected tasks and sleep physiology. They do not replace actual biological drug testing, independent operational observations, or an event-specific causal investigation.

Original primary source: Grady S, Aeschbach D, Wright KP Jr, Czeisler CA. Neuropsychopharmacology (2010);35:1910–1920. https://doi.org/10.1038/npp.2010.63. The original article and its complete numerical table should be consulted for any litigation quotation or detailed assessment of model assumptions.