Do ADHD Stimulants Prevent Sleepiness After 25 Hours Awake? What a 59-Man Study Measured

Short answer: In a 2026 study of 59 young men who remained awake in a laboratory for 25 hours, men with ADHD who were not taking medication reported substantially greater sleepiness the next morning than either men without ADHD or men with ADHD taking their usual prescribed stimulants. The groups followed different sleepiness trajectories over the 25 hours, but the study also reported a nonsignificant overall group main effect, and the medicated ADHD group did not differ significantly from controls on the modeled final sleepiness rating. This is an observational comparison of existing treatment groups with repeated subjective ratings, not a randomized medication-versus-placebo trial, a measurement of blood stimulant concentration, or proof that anyone can safely operate a vehicle after being awake all night.

The distinction matters because a person may report feeling relatively alert without demonstrating normal vigilance, reaction speed, judgment, or safe driving. The investigators recorded hourly responses to the Karolinska Sleepiness Scale (KSS), a nine-point self-rating. They did not determine whether any individual had a measurable methylphenidate or amphetamine concentration in legally collected human venous whole blood, and they did not validate a drug-based method for reconstructing ingestion time or attributing a particular accident. Their findings provide useful evidence about reported sleepiness under sustained wakefulness, subject to the limitations of clinical grouping, small numbers, and heterogeneous medication exposure.

What is the original primary study?

Iris Haimov, Orrie Dan, Shahar Eisenstein, Kfir Asraf, and Ami Cohen published The effects of stimulant medications on the sleepiness curve of young men with Attention-Deficit Hyperactivity Disorder (ADHD) in BMC Psychiatry, volume 26, article 361, in 2026. The original empirical paper reports the actual recruited groups, 25-hour observation procedure, repeated KSS ratings, pre-experiment sleep measurements, mixed-model comparisons, and statistical tests of very sleepy ratings. This discussion draws on those original methods, results, tables, and figures rather than treating a review’s description as independent experimental evidence.

The central question was not whether taking a prescription stimulant renders sleep deprivation harmless. It was whether men with ADHD receiving their preexisting prescribed medication would report a different time course of sleepiness from unmedicated men with ADHD, and whether their curve would resemble that of men without ADHD. The original research therefore separates a comparison of group trajectories from a causal test of an assigned drug. That distinction sets the limit on clinical and forensic interpretation from the outset.

How many independent people participated?

The biological sample was 59 distinct men, not 59 per treatment and not 59 multiplied by the number of hourly questionnaires. Twenty were controls without ADHD. Thirty-nine had the combined presentation of ADHD; among these, 22 were unmedicated during the experiment and 17 were treated with their existing stimulant prescriptions. The investigators described recruitment of men approximately 18–35 years old, although a narrower upper age is mentioned elsewhere in the discussion. The mean age of the overall cohort was 25.46 years, with standard deviation 4.37 years.

The authors recruited participants through community advertisements and conducted symptom interviews and questionnaires to characterize ADHD and exclude certain other psychiatric conditions. The ADHD and control designations were characteristics of the people being studied, not randomized assignments. Each participant supplied repeated observations while remaining awake; repeated ratings increase information about one person’s curve, but do not create new independent people. A statistical model can account for within-person repetition without changing the actual biological sample size.

Which medications were involved, and were their doses standardized?

Of the 17 already medicated men with ADHD, 13 were taking methylphenidate and four were taking an amphetamine medication. The study states that these participants received their regular doses at approximately 08:00, when observation began, and again at 00:00 midnight. It does not supply one common investigational milligram dose, a standardized drug formulation for every participant, or a drug-specific randomized allocation against placebo. The original reported medication schedule therefore permits a statement about usual-treatment exposure, not a fixed milligram-per-kilogram experiment.

These agents must not be conflated. Methylphenidate is not amphetamine, and neither generic stimulant category should be relabeled as methamphetamine. Lisdexamfetamine, an inactive prodrug that yields dextroamphetamine, is another separate entity and was not identified as a distinct measured exposure here. The paper’s four-person amphetamine subgroup is too small, and too sparsely described, to establish a separate amphetamine-versus-methylphenidate effect. Nor can the paper isolate the effect of the morning dose from the midnight dose, because the medicated men received both and the study did not include a randomized single-dose comparator arm.

What exactly did 25 hours of wakefulness mean in the laboratory?

Participants entered a controlled observation period beginning in the morning and continuing across the following night into the next morning. The central outcome was recorded once each hour across approximately 25 consecutive hours awake. The researchers used a repeated-measures design to examine a changing curve rather than relying solely on one endpoint. During the night, the medicated men received their second ordinary prescription dose at midnight, an event the authors marked on the original plotted sleepiness curves.

One practical consequence of this protocol is that clock time, time awake, treatment group, and second-dose timing are related. A rising overnight rating could reflect prolonged wakefulness, circadian timing, symptoms, expectations, medication exposure, or interactions among these influences. Because there was no randomized placebo substitution within the medicated group and no alternative midnight-dose schedule, the study cannot determine how much of any particular hour’s reported alertness was produced by a precise measured drug concentration.

What did the Karolinska Sleepiness Scale actually measure?

The KSS is an immediate subjective rating of how sleepy or alert a person feels. The scale runs from 1 to 9; higher values indicate more perceived sleepiness. Participants rated their condition during the observation period rather than completing a contemporaneous laboratory blood assay. A response of eight or nine expresses that the person feels markedly sleepy. It does not quantify psychomotor reaction time in milliseconds, quantify a stimulant in nanograms per milliliter, or directly score collision avoidance.

Because the question is subjective, differences in how people interpret and use the scale can affect the averages. KSS is valuable for describing within-person perceived sleepiness and group-level time courses; it is not interchangeable with the psychomotor vigilance test (PVT), driving simulator lane keeping, electroencephalographic sleep intrusion, multiple sleep latency testing, or a formal polysomnogram. The researchers did not independently establish that a person reporting a six instead of an eight had restored all cognitive abilities needed for an operational task.

What did the full 25-hour mixed model show?

The investigators reported a statistically significant time-by-group interaction, F(48, 1392.08) = 2.14, p < .001. That term addresses whether the trajectory changed differently over time across the three observed groups. A large time main effect, F(24, 1392.08) = 55.79, p < .001, indicated that reported sleepiness varied substantially across the prolonged observation. But the corresponding overall group main effect was F(2, 58.99) = 1.81, p = .171—a genuine nonsignificant result.

Those results are not contradictory. A nonsignificant group average across the whole recording does not rule out differences at specific times, and a significant interaction does not demonstrate that every pair of groups differed at every hour. The time course matters: clinical groups that are similar for portions of a day can diverge during the overnight period. An analysis that reports only the favorable interaction while omitting the nonsignificant group main effect would misrepresent the balance of the original evidence.

How different were the three groups by the following morning?

At the end of the 25-hour observation, the original modeled KSS estimates were 8.42 ± 0.45 for unmedicated ADHD, 6.21 ± 0.45 for medicated ADHD, and 6.70 ± 0.43 for controls; the uncertainty numbers here are the source-reported standard errors, not standard deviations or confidence intervals. The unmedicated group differed from controls (p = .007) and from the medicated group (p < .001). The difference between medicated participants and controls was not statistically significant (p = .437).

It would be a mistake to conclude that the medicated men were therefore immune to sleep loss. Their modeled mean KSS still exceeded six on a scale where high values signify substantial sleepiness. The test establishes a difference in subjective ratings among the observed groups under this protocol; it does not establish how fast a medicated participant would respond to an unexpected hazard, whether someone would have a microsleep, or whether a driver could safely continue operating a vehicle.

First numerical evidence table: the original KSS time-course and comparisons

The estimates reflect the source’s fitted analysis and should not be casually replaced with arithmetic averages calculated from an image or with counts of invented individual respondents. The manuscript reports fractional denominator degrees of freedom because of its chosen repeated-measures mixed-model approximation; these degrees of freedom are not a count of additional participants. Most importantly, none of these values measures medication molecules in blood. The table preserves both significant and negative results so that evidence of changing group trajectories is not mistaken for uniform proof of medication protection.

Did differences also appear during the hours after midnight?

Yes, in the prespecified reported nocturnal comparison from approximately 01:00 onward, the source reported an overall group contrast of F(2, 58.95) = 4.49, p = .015. Mean KSS estimates across that interval were 7.16 ± 0.38 for unmedicated ADHD, 5.57 ± 0.39 for medicated ADHD, and 6.01 ± 0.38 for controls. The comparison of unmedicated ADHD against controls reached p = .039 and against medicated ADHD p = .005. Medicated men versus controls remained nonsignificant, p = .419.

These are findings about a specified late-night interval, not evidence that medication levels were stable or known for its duration. The midnight dose occurred immediately before the interval, but the study did not measure serial drug concentrations or assess which formulation reached which circulating peak. Different preparations, ordinary clinical doses, and persons’ pharmacokinetic variation could produce different exposure curves. The nocturnal comparison can motivate more controlled work while remaining insufficient to establish a dose-response model or universal countermeasure to fatigue.

What about the percentage of men reporting severe sleepiness?

Investigators also evaluated whether ratings exceeded KSS 7 near the end. Their logistic model gave χ²(2) = 10.98, p = .004. Reported high-sleepiness percentages were 88.2% in unmedicated ADHD, 55.0% among controls, and 36.9% in medicated ADHD. The unmedicated-versus-control comparison was significant, p = .038, as was unmedicated-versus-medicated, p = .004. The medicated-versus-control comparison was p = .258, again not significant.

These percentages need careful handling. The original graph and statistical model do not furnish a verified participant-by-participant event table. It is therefore not justified to reverse-engineer exact event counts from the plotted percentages or assert a specific count of affected medicated participants without raw source data. A population proportion is not an individual observation and cannot substitute for a participant-level forensic record.

What did the five nights before the laboratory observation contribute?

The investigators assessed sleep before admission by combining the Pittsburgh Sleep Quality Index (PSQI), a retrospective subjective sleep questionnaire, with approximately five nights of wrist actigraphy. An actigraph uses movement patterns to estimate intervals of sleep and wakefulness; it is not the same as a polysomnogram with EEG-based sleep staging. Participants had to meet a minimum pre-experiment sleep requirement, and the study described no exclusions for falling below that minimum in the observed lead-in period.

Between-group total sleep duration, reported sleep-onset latency, and wake after sleep onset were not significantly different in the original summary tests. By contrast, estimated sleep efficiency was higher in controls, with a significant overall group comparison. That baseline difference deserves attention because sleep efficiency is a plausible contributor to perceived fatigue, even when adjusted group analyses indicate a remaining difference. A null comparison on one sleep metric does not prove that every participant entered the experiment with identical sleep pressure or equivalent sleep health.

Second numerical evidence table: baseline sleep and tested null findings

The source’s different test families reflect different descriptive distributions and statistical choices; a PSQI mean and an actigraph median should not be combined as though they were one continuous outcome. Also, the correlation tests did not show a significant simple linear relationship between sleep efficiency and starting or ending reported sleepiness. Their null results neither establish complete lack of confounding nor prove that sleep quality cannot matter physiologically. They limit the claim the authors can make about this particular measured correlation in this sample.

Can a low baseline sleep-efficiency correlation eliminate confounding?

No. Correlation near zero for a single group-pooled predictor cannot eliminate unmeasured clinical, circadian, sleep-disorder, expectancy, or medication-selection differences. The investigators excluded certain comorbid conditions and attempted to standardize the preceding nights; those procedures improve interpretability but do not create treatment randomization. The medicated and unmedicated ADHD groups may differ in symptom severity, access to care, tolerability, typical routines, adherence, or responses to being observed. With only 17 medicated men—and just four receiving an amphetamine preparation—those differences are not reliably separated by this design.

A higher control-group actigraph efficiency could plausibly be relevant even though the paper did not demonstrate a significant efficiency–KSS association. Conversely, the lack of a statistically significant efficiency correlation does not establish that controls and medicated participants were genuinely equivalent in objectively assessed sleep. The correct statement is narrower: the source observed selected baseline actigraphy differences and did not detect the two simple KSS correlations that it tested.

Does this study prove a stimulant reverses the biological effects of sleep deprivation?

It does not. The medicated group experienced the same 25-hour wakefulness protocol but reported less sleepiness than the unmedicated ADHD group late in the experiment. That finding can support an association between existing treatment and subjective alertness in this sample. It cannot distinguish a direct medication effect from all consequences of the treatment-selection process, nor show restoration of reaction time, attention accuracy, vigilance lapses, memory consolidation, executive judgment, or physiological sleep homeostasis. There were no randomized drug-free retest periods for the same medicated men.

The distinction between a group comparison and a within-person causal contrast is particularly important when interpreting treatments. The experiment did not administer a standardized drug to the unmedicated group, then withdraw it under blinding, then compare that individual’s objective performance under matched sleep deprivation. Such a design could better isolate medication effects, although it would require appropriate safety safeguards and careful measurement. The present paper makes a more limited contribution: a longitudinal description of reported sleepiness across existing diagnostic and treatment groups.

Were any stimulant concentrations measured in legally collected human venous whole blood?

No quantitative drug concentrations were reported. The published outcomes do not include a collected forensic human venous whole-blood sample assayed for methylphenidate, ritalinic acid, amphetamine stereoisomers, or any stimulant parent or metabolite. Nor did the paper report a quantitative stimulant result in separated plasma, serum, urine, oral fluid, hair, sweat, vitreous, breast milk, breath aerosol, or tissue. It supplies no analytical LC-MS/MS or GC-MS assay validation, limit of detection, limit of quantification, laboratory measurement uncertainty, collection tube specification, concentration in ng/mL, or paired drug concentration–KSS calibration.

Whole blood and separated plasma or serum are physically different biological matrices. A hypothetical plasma concentration would not become a whole-blood result just by relabeling the specimen; matrix-specific partitioning, analytical method, time, and validation would matter. But that conversion question is not even reached here, because no quantitative drug concentration was measured. The two reported ordinary medication times—08:00 and midnight—are scheduled administrations, not measured adherence verified by a chemical analyte and not a source-derived basis for backward extrapolation of a blood level.

Why is subjective alertness not a validated driving-impairment measurement?

A driver can subjectively feel capable of continuing while still making slower or inconsistent responses to unexpected events; the opposite may also occur. KSS describes perceived sleepiness, not lane position, braking latency, hazard recognition, collision avoidance, response inhibition, or field sobriety results. The Haimov study measured no on-road operation and no simulated roadway, aircraft, or industrial machinery task. Thus even a statistically supported group difference in KSS cannot be converted to a probability that a particular driver was or was not impaired at a specified time.

One must also avoid treating a prescription as proof of safety or danger. The four amphetamine-treated men and 13 methylphenidate-treated men were not individually tested for safe operation, specific driving errors, toxic doses, or serum and blood drug levels. The study did not compare therapeutic plasma exposures with legally relevant venous whole blood concentrations, did not establish an impairment concentration cutoff, and did not estimate a validated probability of a crash conditional on a KSS score. Any such claim would be an extrapolation beyond its actual measurements.

What uncertainty should accompany the reported p values?

The participant sample was restricted to young men. Sex, age, medication class, dose, release formulation, baseline sleep needs, clinical severity, and circadian differences may influence how results generalize. Mixed models permit repeated-hour observations, but an impressive denominator degree of freedom from hundreds of ratings does not transform 59 biological subjects into hundreds of independent randomized participants. Selected late-night pairwise comparisons also sit within a broader set of measurements, meaning readers should inspect the authors’ analysis decisions rather than describe every isolated p value as a separate confirmatory experiment.

Reporting limitations matter as well. The article did not disclose participant-level raw response records, and the reader cannot independently reconstruct an individual’s hourly alertness from its aggregate curves. These limits are reasons to preserve source-reported values with their actual meaning, refrain from manufacturing exact unreported headcounts, and avoid attaching false mathematical precision to a new person’s medical or legal condition.

What forensic questions remain explicitly unmeasured?

The published research provides no validated relation between a medication’s nominal dose and a chemically measured whole-blood level, no individual drug elimination slope, no validated last-ingestion time reconstruction, and no method for telling whether a person’s alertness was affected by one specific blood concentration. It also provides no individual diagnostic sensitivity or specificity of KSS for stimulant intoxication, withdrawal, circadian sleepiness, or residual fatigue. No road or simulator crash, field sobriety examination, legal case, event witness statement, airline performance outcome, or injury causation endpoint was observed.

These omissions are not criticisms of a study that principally asked a sleepiness question; they define how it can be cited accurately. It is legitimate to say that the groups’ reported sleepiness diverged overnight under the tested design and that some accompanying group contrasts were nonsignificant. It would be illegitimate to say the data identify the responsible drug concentration in a hypothetical defendant, prove a specific ingestion clock, establish that a named person could safely drive all night, or supply a person-specific opinion about the cause of an accident.

What research would strengthen the causal interpretation?

A stronger treatment-effect study could recruit sufficiently large, diverse groups and experimentally compare precisely documented drug regimens, formulations, and dose timing against well-designed controls where ethical. It would predefine outcomes and distinguish subjective KSS scores from objective PVT lapse and reaction-time endpoints, cognitive accuracy, electroencephalographic sleepiness markers, and operational simulation tasks. An appropriate crossover or randomized design could reduce the confounding between who chooses or receives medication and how that person otherwise sleeps or responds to sustained wakefulness.

If forensic concentration questions were among its objectives, the follow-up would separately collect timed physical specimens under validated conditions and report actual parent and metabolite measurements in the expressly named matrices, with analytical uncertainty and relevant time points. It would not treat brain imaging, motor activity, a prescription list, or a nominal dose as a proxy for measured venous whole blood. Finally, before making individual risk claims, researchers would need independent validation on real people and tasks rather than inferring person-level accuracy from group means.

Frequently asked questions

1. Did the researchers study 59 patients in every treatment group?

No. There were 59 independently recruited men in total: 20 controls without ADHD, 22 with unmedicated combined-type ADHD, and 17 with medicated ADHD. Each person’s repeated hourly KSS responses remain repeated measures within that person. The study’s 25-hour time series does not turn 59 biological participants into 1,475 unrelated human subjects or multiply the already small four-person amphetamine subgroup.

2. Were the medicated men assigned the same methylphenidate dose?

No. Thirteen medicated participants used methylphenidate and four used an amphetamine medication, each on his ordinary regimen. The paper identified administrations at 08:00 and midnight but did not report a single standardized investigational dose in milligrams for all 17 men. It therefore cannot provide a drug-specific or dose-specific concentration-response model, nor cleanly separate the two medication classes’ effects.

3. Did medicated ADHD men have statistically less sleepiness than controls?

Not on the relevant reported contrasts. At the final assessment the medicated-versus-control comparison gave p = .437, and during the approximately 01:00-onward interval the comparison gave p = .419. Both are true nonsignificant tests. The absence of a significant contrast does not establish statistical equivalence or proof of identical alertness in every participant, especially with small groups and subjective reporting.

4. Was the full-period group effect significant?

No. The full mixed model’s group main effect was F(2, 58.99) = 1.81, p = .171. The time main effect and time-by-group interaction were significant, which means the trajectories differed as wakefulness progressed. Those hypotheses are distinct. A defensible account reports the nonsignificant overall group term rather than only the later contrasts that favor a medication-associated difference.

5. Does KSS above seven diagnose intoxication or unsafe driving?

No. A score above seven expresses substantial self-rated sleepiness on a nine-point scale. It is not a chromatography result, a drug-impaired-driving cutoff, or a validated legal classification of a particular driver’s performance. The study did not measure driving or accidents and cannot establish who was impaired at a specific case-related time from this threshold alone.

6. Was amphetamine or methylphenidate measured in blood or plasma?

No quantitative parent-drug or metabolite assays were reported in venous whole blood, separated plasma or serum, or any other forensic matrix. Prescription times and a KSS number are neither blood concentrations nor analytically validated proxies for them. Without an actually collected and calibrated physical specimen, no source-grounded ng/mL value, detection threshold, clearance estimate, or individual ingestion-time reconstruction is available.

7. What is the soundest forensic conclusion from this study?

Under one controlled overnight observation schedule, 22 unmedicated men with ADHD reported higher late and final subjective sleepiness than 17 men with ADHD on usual prescriptions and 20 controls, while important other comparisons were nonsignificant. The result illustrates how clinical grouping and time awake relate to self-reported fatigue, but supplies no individual stimulant blood measurement, dose-response proof, driving-safety validation, or evidence of what caused a particular incident.

Bottom line: a changing subjective sleepiness curve, not a forensic safety test

Haimov and colleagues documented a useful but circumscribed contrast in a 59-man observational laboratory study: the unmedicated ADHD group’s subjective sleepiness rose more sharply with overnight wakefulness, while existing stimulant-treated participants resembled the controls on selected later self-ratings. Positive group-by-time and overnight comparisons must be read alongside the full-period group-main-effect null, medicated-versus-control nulls, and baseline actigraphy findings. The evidence does not show that sleep deprivation has become harmless, that the treatment groups were randomized, or that the drugs had a measured plasma or whole-blood exposure.

For an accurate forensic account, preserve the original group denominator, the distinction between KSS and objective vigilance, the distinction between medication administration and chemical assaying, and the lack of any actual driver or event measurement. The underlying methods, tables, figures, and inferential tests are available in the full original 2026 primary empirical study. Nothing in this population-level comparison establishes the venous whole-blood drug concentration, precise ingestion time, safe operation, culpability, or incident causation of an individual.