Does a Dextroamphetamine Patch Work Within One Hour? A 106-Child Crossover Trial

Short answer: Not on the timed mathematics measures in this trial. In 106 children and adolescents with attention-deficit/hyperactivity disorder (ADHD) who entered a randomized crossover assessment, a nine-hour transdermal dextroamphetamine patch did not significantly outperform placebo on arithmetic problems attempted or correct at one hour after application. The adjusted between-treatment differences were 10.3 problems attempted (p = .231) and 10.9 correct (p = .204). By two hours, both mathematics comparisons were statistically significant, and the reported benefits continued to 12 hours after application, three hours after planned patch removal. Observer-rated classroom behavior also favored active treatment beginning at two hours. Those statements describe group-level test results, not a direct assay of a child’s blood dextroamphetamine, a personal drug-effect threshold, or fitness to drive.

This distinction comes from the complete nine-page primary clinical trial by Cutler and colleagues, published in the Journal of Child and Adolescent Psychopharmacology in 2022. The original study is available at DOI 10.1089/cap.2021.0107. The original paper separates an optimized nominal transdermal dose, timed observer ratings, objectively scored classroom arithmetic, adverse-event surveillance, and subsequent pharmacokinetic/pharmacodynamic modeling. These are not interchangeable forms of evidence.

What precise question did the trial answer?

The experiment asked whether a once-daily dextroamphetamine transdermal system, applied to the hip for nine hours, improved ADHD-related behavior and academic-task productivity relative to placebo during a controlled laboratory-school day. Investigators also sought to identify the first assessed postapplication interval with convincing group separation and whether efficacy remained measurable after the patch was removed. The treatment was dextroamphetamine delivered through skin, not oral immediate-release amphetamine, racemic amphetamine sulfate, methamphetamine, methylphenidate, or lisdexamfetamine.

A formulation’s onset is operational: it depends on the applied patch, the wear protocol, the measured behavioral instrument, the timepoints sampled, and the population selected. A positive result at two hours does not reveal an exact minute when every participant experienced a benefit. Conversely, an inconclusive one-hour comparison is not proof that all individuals had zero pharmacological effect at that time. The defensible answer is narrower: the prespecified classroom outcomes did not demonstrate statistically significant one-hour arithmetic improvement, while later comparisons did.

How many independent children took part?

The source reports 110 participants in its initial safety population. Of these, 107 reached a dose judged optimal during the open-label phase, and 106 entered the double-blind crossover phase. The reported mean age in the safety population was 10.5 years; 80 were 6–12 years old and 30 were 13–17 years old. The laboratory-school efficacy analysis used 106 participants. The label N = 106 does not mean 106 children on active drug plus another independent 106 on placebo. The crossover design exposes the same recruited individuals to treatment and placebo in separate periods; measurements at multiple hours remain nested within each child.

During optimization, 7 participants received the 5 mg/9-hour patch strength, 35 the 10 mg, 42 the 15 mg, and 23 the 20 mg strength, totaling 107 optimized patients. Because one individual withdrew after optimization, those 107 counts cannot simply be treated as the double-blind efficacy distribution. A child could contribute multiple observation points to the laboratory classroom schedule, but each score was not an independent biological participant. This distinction matters whenever apparent data density is used to argue statistical power or transportability to another setting.

What was actually randomized and blinded?

Every participant began with a five-week open-label dose-optimization phase, starting with a 5 mg patch applied to alternating hips for nine hours daily. Investigators could titrate across labeled 5, 10, 15, and 20 mg/9-hour strengths according to clinical response and tolerability. Those are intended amounts delivered during a specified nine-hour wear period; they are not blood concentrations, chemically measured delivered mass for each child, or a uniform milligram dose shared by all randomized participants.

After optimization, patients entered two blinded weeks in a crossover sequence: an optimized active patch during one period and a matching placebo during the other. Each treatment’s laboratory-school observations took place within that controlled protocol. The investigators’ statistical treatment of sequence is important. They observed a carryover or sequence signal in the primary analysis (p = .009) and undertook a permutation sensitivity analysis that still supported a treatment effect (p = .008). Crossover design improves within-subject comparability but does not make potential period or sequence effects disappear.

Where was the patch placed and when was it removed?

The transdermal system was placed on the hip and worn for nine hours. On the laboratory-school day the investigators followed performance from the time of application through 12 hours after application. That schedule included observations while the patch remained attached, and then additional observations after removal. On the study’s figures, patch removal occurs at nine hours. Findings at 10 or 12 hours therefore cannot be described as testing a patch still on the skin at those times.

But removal also does not guarantee immediate disappearance of dextroamphetamine already entering or circulating in the body. The observed persistence is consistent with delayed absorption and elimination, while the study’s behavioral endpoints do not establish an individualized pharmacokinetic disappearance curve. Some of the paper’s later plotted drug-concentration relationships arose from pharmacokinetic/pharmacodynamic modeling and simulations, not contemporaneous trial-specific forensic blood draws at each classroom hour.

Which endpoints measured behavior and which measured schoolwork?

SKAMP is the Swanson, Kotkin, Agler, M-Flynn, and Pelham classroom rating scale. Trained observers use it to rate ADHD-related impairment during the laboratory-school setting; lower SKAMP scores indicate improvement in the assessed behavior. The trial’s primary efficacy comparison used SKAMP total scores over the observed day, with onset and duration explored through time-specific contrasts. SKAMP measures a defined observational behavior construct, not blood or plasma concentration, and not an individual’s general competence in every real-world activity.

The Permanent Product Measure of Performance (PERMP) evaluates arithmetic-task output, recording the number of problems attempted and the number answered correctly during a timed classroom worksheet. Counting correctly answered problems reduces reliance on subjective impressions, but it remains an artificial, structured school task. Different test constructs can show different onset patterns even after the same patch application. Neither SKAMP nor PERMP has been validated by this paper as a stand-alone test for intoxication, driving impairment, incident reconstruction, or credibility.

What happened at one hour after application?

The original numeric PERMP table is unusually informative because it includes genuine negative findings rather than only the headline positive results. At one hour, the adjusted active-minus-placebo difference for attempted mathematics problems was 10.3, with a 95% confidence interval from −6.6 to 27.2 and p = .231. The interval includes zero. For correctly answered problems, the adjusted difference was 10.9, 95% CI −5.9 to 27.7, p = .204. That interval also includes zero.

These are not significant treatment contrasts at the conventional .05 threshold, despite positive point estimates. A defensible explanation avoids two opposite errors. It must not promote the positive estimates to proven one-hour benefit, but it also must not translate a null hypothesis test into a demonstrated absence of every possible benefit. The paper describes its observed classroom efficacy onset at two hours. There was no trial measurement at every minute between zero and two hours capable of pinpointing a universal onset instant.

What changed at two hours and beyond?

At two hours, PERMP attempted problems favored active patch by an adjusted 36.4 problems (95% CI 19.5–53.2, p < .001). Correctly answered problems favored active patch by 36.0 (95% CI 19.2–52.8, p < .001). The time-specific differences remained significant in the source table at 3, 4.5, 6, 7, 9, 10, and 12 hours. At 12 hours, after patch removal, the adjusted difference was 37.3 problems attempted (95% CI 20.4–54.2) and 38.9 correct (95% CI 22.0–55.7), each p < .001.

The SKAMP observations reached reported significance at two hours and remained favorable through 12 hours. The investigators reported an overall least-squares treatment difference of approximately −5.87 on SKAMP total score, with the source’s confidence bounds presented in the negative direction and p < .001. Those are statistical descriptions of a selected trial population after dose optimization. They do not demonstrate that every child improved, that the magnitude generalizes to an unoptimized first-dose patient, or that a child’s blood contained a particular quantity of dextroamphetamine at any observation point.

Source-grounded table: a genuine null followed by later positive results

The arithmetic figures are adjusted differences from the original study’s PERMP results, not independent recalculations from a digital graph. A 12-hour treatment difference is not equivalent to wearing the patch for 12 hours: the planned wear time was nine hours. No row should be interpreted as a forensic amphetamine concentration or as a validated threshold for performance outside the classroom.

Why is the nine-hour removal point scientifically important?

The source reports a significant difference on the selected outcomes three hours after scheduled patch removal. This finding addresses clinical duration, but it does not prove that removal instantly terminated molecular absorption or that the same duration would follow a prematurely removed patch. Skin delivery creates a temporal relationship among patch application, cutaneous drug availability, systemic distribution, and behavior; removal changes the input conditions, not necessarily the immediate state of all compartments.

In the paper, investigators used pharmacokinetic/pharmacodynamic simulations to explore differing patch-wear times, including early removal. The resulting plotted dextroamphetamine concentration curves were model outputs. A modeled median expressed in ng/mL is not a forensic laboratory’s measured whole-blood result from a specifically identified patient. It is particularly important not to take those model curves, assign them to one of the enrolled children, and then assert an observed plasma, serum, or blood concentration that the clinical results do not report.

What is known about the population’s optimized patch strengths?

The five-week optimization period matters for interpreting performance. The full safety group entered the trial with substantial ADHD symptoms, and dosing was adjusted before the randomized classroom experiment. The 107 optimized patients had differing nominal patch strengths; in the study’s distribution, the majority used 10 or 15 mg per nine hours. Consequently, the estimated average treatment effect is a pooled result across individualized optimized dosing. It is not a controlled head-to-head dose-response test proving that 20 mg is better than 10 mg, or that an arbitrary individual at 5 mg will experience the average clinical trajectory.

Patients who could not progress through the optimization and trial-entry requirements are not automatically represented by the final crossover efficacy sample. The selected nature of the sample can magnify apparent generalizability if neglected. Dosage in milligrams per nine-hour patch wear should also be kept separate from the total compound physically contained in the patch, the mass actually absorbed by the body, and concentrations subsequently quantified in any biological specimen. The clinical study did not provide a child-by-child forensic absorption audit.

Did safety findings establish that the patch has no physiological risks?

No. Investigators tracked treatment-emergent adverse events, skin responses, pulse, and blood pressure. During the double-blind period, 44 of 105 participants exposed to active patch (41.9%) and 43 of 105 exposed to placebo (41.0%) reported at least one treatment-emergent adverse event in the source safety table. These are descriptive frequencies, not a formal statistical equivalence finding. Decreased appetite was reported in 13 active-patch periods (12.3%) versus 2 placebo periods (1.9%). The study did not report deaths or serious treatment-emergent adverse events during the observed trial period.

Application-site symptoms occurred during dose optimization, including pain and itching. The trial reported occasional local skin reactions and described most adverse events as mild or moderate. Absence of a serious event in a particular sample is not proof that serious adverse effects can never occur, and a symptom appearing after drug exposure does not on its own establish pharmacological causation. Nor do group blood-pressure or pulse measurements substitute for quantitative testing of dextroamphetamine in a specific child’s venous specimen.

Source-grounded table: denominators, route, and safety observations

The denominators in the final safety comparisons reflect available exposed observations in the source table. The number of timepoints, questionnaire items, and drug-patch strengths cannot be added to obtain more independently recruited children. The table also highlights a critical distinction: the presence of physiological monitoring does not mean the paper performed analytical toxicology on human venous whole blood.

Did the researchers actually measure dextroamphetamine in whole blood?

The efficacy experiment reported laboratory-classroom behavior, mathematics output, adverse events and vital signs. It did not report participant-level contemporaneous validated quantitative dextroamphetamine concentrations in legally collected venous whole blood. It did not present corresponding individually measured drug levels in centrifuged plasma or serum, urine, oral fluid, sweat, hair, vitreous humor, tissue, or true exhaled aerosol specimens as trial efficacy endpoints. There was no paired chemistry/behavior dataset with a published specimen timestamp and validated reporting threshold suitable for a forensic individual-concentration opinion.

Physiological safety measurement is not chemical drug assay. The paper’s simulated pharmacokinetic plots require separate interpretation, because their model-supplied curves do not turn behavioral observations into direct specimen measurements. For the foundational difference between legally collected whole blood and separated biological fractions, consult the independently published same-site guide to biological matrices in forensic toxicology. The same mg/9-hour patch label cannot be translated to a ng/mL court specimen without actual measured data, validated matrix-specific methods, sampling timing and scientific uncertainty.

Could a plasma concentration graph be converted into a forensic whole-blood result?

Not from this paper alone. Whole blood contains cells and plasma, while serum and plasma represent differently prepared fluid compartments. Analytical recovery, protein binding, cell partitioning, collection tubes, preservatives, elapsed storage, extraction and calibration may matter. Even where pharmacokinetic information exists for a drug in one compartment, converting model-estimated plasma values to measured legal whole blood requires validated comparative evidence and does not create an actual chain-of-custody specimen.

The study’s simulated time course can explain why the researchers explored earlier patch removal and duration. It cannot establish the amount in a named child at a particular minute, drug use before a criminal incident, or whether a driver was impaired. A quantitative case interpretation requires the actual specimen, its validated analytical method and result, collection time, matrix identity, reporting uncertainty, clinical and behavioral context, and applicable law. See the site’s published guide to confirmatory drug testing for the distinction between identification, quantitation and interpretation.

Does a statistically significant SKAMP score prove safe driving or functional competence?

It does not. SKAMP assesses observer-rated classroom impairment under a structured research protocol. PERMP measures timed arithmetic performance. Both may matter to pediatric ADHD treatment research, but neither is a validated road test, field sobriety examination, real-world vehicle-control assay, occupational fitness determination or case-specific functional-capacity assessment. Better average school scores cannot be used to conclude that a particular person obeyed traffic law, perceived a hazard, acted without negligence, or was uninfluenced by a stimulant.

Likewise, statistical superiority over placebo is a property of a comparison of outcome distributions under the design actually used. It does not establish a unique drug effect in every individual. Differences in diagnosis, age, dose optimization, concomitant medication, sleep, health, task familiarity and other factors can affect transportability. A specific forensic question requires evidence of the relevant conduct, not simply citation to a group-level clinical treatment effect.

What limitations should govern a forensic discussion?

First, the participants were children and adolescents with clinically diagnosed ADHD who had undergone individualized dose optimization. The study was not a random sample of drivers, criminal defendants or otherwise healthy adults receiving one experimental dose. Second, the double-blind crossover provided repeated within-person contrasts and exhibited a sequence/carryover signal that the investigators addressed using additional analysis. Third, one-hour and later timepoint findings come from different instrument-based comparisons; a single phrase such as “works immediately” conceals the measurement details.

Fourth, the investigator-funded clinical design evaluated efficacy and safety, not forensic chemical attribution. Its pharmacokinetic plots include modeled results, while its clinical outcome tables concern behavior and arithmetic output. Fifth, the study cannot isolate a fixed-dose causal gradient because the randomized contrast was active optimized patch versus placebo. Finally, the source’s 12-hour benefit after nine hours of wear should not be turned into a reliable individual blood concentration, a precise last-use inference or a finding about a specific person’s behavior on a particular day.

What is the defensible answer in a clinical or legal evidence review?

A careful answer identifies the treatment and the endpoint together. In a dose-optimized pediatric crossover experiment, nine-hour transdermal dextroamphetamine demonstrated statistically significant group-level classroom symptom and arithmetic differences beginning at the two-hour observation and continuing to 12 hours. At one hour, the original PERMP attempted and correct comparisons were not statistically significant. That apparent delayed onset is a feature of this protocol and these measurement times, not a universal time-to-effect clock. The finding persists after removal in the scheduled group measurements, but drug-concentration modeling should not be passed off as direct forensic bioanalysis.

For an expert reviewing a report or legal record, the relevant sequence is to determine what was administered, by which route, for how long, how often the behavioral endpoint was assessed, whether the actual test result reached statistical significance, and whether an independently validated drug assay existed in the specimen relevant to the question. A case-specific opinion about exposure, last use, impairment or causation would need additional evidence that this article does not contain.

Frequently asked forensic-science questions

Did the transdermal patch significantly improve arithmetic at one hour?

No statistically significant between-treatment difference was demonstrated in this study’s one-hour PERMP comparisons. Attempted problems differed by an adjusted +10.3 (p = .231) and correct by +10.9 (p = .204). Neither result proves zero possible benefit, but both prevent a claim of established one-hour mathematics improvement.

When was the first observed significant benefit?

The paper reports onset at two hours after patch application. At that point the active-placebo adjusted arithmetic differences were +36.4 attempted and +36.0 correct, each p < .001, with an accompanying significant SKAMP total-score contrast. It did not continuously test every moment between application and the two-hour assessment.

Why were effects still observed after removal?

The patch was removed at nine hours, but group effects persisted through the last assessed 12-hour timepoint. Removal stops intentional continued patch wear; it does not immediately establish a zero-body-drug state. The study also explored alternative wear periods using pharmacokinetic/pharmacodynamic simulations, which must not be represented as direct blood tests in each enrolled child.

Were 106 children randomized to active drug and another 106 to placebo?

No. There were 106 participants entering the double-blind crossover, in which individuals received active patch and placebo at different periods. Safety counts use their own denominators. Repeated SKAMP ratings, PERMP worksheets and treatment periods are not independent additional recruited children.

Do the 5–20 mg patch strengths specify blood levels?

No. A nominal 5, 10, 15 or 20 mg over nine hours is a product dosing designation, not a laboratory-reported whole-blood concentration. The trial contains no paired patient-specific validated whole-blood dextroamphetamine assay values from which such a conversion could be established.

Does the paper prove a dextroamphetamine concentration or individual impairment cutoff?

No. SKAMP and PERMP are behavioral and schoolwork measures, and the pharmacokinetic plots include simulations. The work did not validate a legal venous whole-blood concentration threshold, a last-dose estimator, an individual driving impairment classifier or an incident-causation model.

What should be reviewed before applying this study to a forensic case?

Identify the exact formulation, dose, wear and removal times, clinical selection and titration, and the statistical endpoint at the relevant assessment time. Separately verify any actual biological specimen’s matrix, collection chain, analytical method, validated reporting range and uncertainty. Then assess the individual’s real-world conduct independently rather than importing a classroom-treatment average as a forensic conclusion.

Primary research

Cutler AJ, Suzuki K, Starling B, Balakrishnan K, Komaroff M, Castelli M, Meeves S, Childress AC. (2022). Efficacy and Safety of Dextroamphetamine Transdermal System for the Treatment of Attention-Deficit/Hyperactivity Disorder in Children and Adolescents: Results from a Pivotal Phase 2 Study. Journal of Child and Adolescent Psychopharmacology 32(2):89–97. Original complete research report: DOI 10.1089/cap.2021.0107.