Does a 13-Hour Classroom Effect Prove Amphetamine Remains in Blood? A 99-Child Oral-Suspension Trial

What does the original experiment actually answer?

In the original randomized pediatric classroom experiment, children taking dose-optimized, once-daily oral amphetamine extended-release suspension had better blinded classroom behavior ratings and completed more arithmetic problems than children given placebo. The treatment difference was demonstrable at the first scheduled examination, one hour after dosing, and remained detectable at the last scheduled examination, thirteen hours after dosing. That is evidence about observed performance under a specific controlled protocol; it is not a direct measurement of when amphetamine first entered blood, when its concentration peaked, or whether a particular person was impaired.

The central interpretive question is therefore narrower than whether an oral stimulant “lasts thirteen hours.” The source shows a maintained statistical difference across scheduled teacher-rating and mathematics assessments in a selected pediatric ADHD sample. It never takes paired postdose blood specimens for a confirmatory amphetamine assay. A behavioral effect horizon and a chemical detection window are not interchangeable clocks.

Which single primary scientific paper supports this account?

The full primary report is Ann C. Childress, Sharon B. Wigal, Matthew N. Brams, John M. Turnbow, Yulia Pincus, Heidi W. Belden, and Sally A. Berry (2018), Efficacy and Safety of Amphetamine Extended-Release Oral Suspension in Children with Attention-Deficit/Hyperactivity Disorder, Journal of Child and Adolescent Psychopharmacology 28(5):306–313, DOI 10.1089/cap.2017.0095. This account uses the empirical study rather than substituting a product label, later commentary, a review, or another group’s ADHD trial.

The report contains a five-week initial active-medication dose-optimization phase followed by a one-week randomized, double-blind, parallel-group assignment. Its primary endpoint was an observer-rated classroom measure at four hours, not an amphetamine concentration. The authors also disclosed that the manufacturer supported the work and participated in study design and analysis. That sponsorship is relevant to evaluating possible bias, but it does not automatically negate an observed randomized treatment difference.

How many independent children were studied?

The source records 108 children enrolled, 107 in the initial safety population, 100 randomized for the blinded week, and 99 ultimately included in the intent-to-treat efficacy population. Forty-eight efficacy participants received placebo and fifty-one received active suspension. In the randomized safety population, the active denominator was 52 rather than 51 because an additional randomized child did not contribute to the required efficacy observations.

The 99-person efficacy count is a biological participant count. The same child’s SKAMP ratings at predose and eight subsequent periods are repeated measures, not nine independent children. The same applies to the ten-minute mathematics worksheets taken at successive classroom intervals. Treating each repeated rating as an independent person would inflate the effective sample and misstate precision.

Was this crossover testing, and why does the design matter?

No. In the blinded phase, each child was assigned to one treatment arm for the study week. That is a parallel comparison after all eligible children had experienced active-dose optimization. It is not the two-period, within-child active-versus-placebo crossover used in some other laboratory classroom trials. When interpreting this paper, do not transfer assumptions from the separate eighteen-child oral-suspension pilot or from a transdermal dextroamphetamine patch study.

Randomization improves comparability of treatment groups during the double-blind week, but the preceding optimization can select children whose response and tolerability were suitable for continued participation. The resulting sample is not a population-based experiment on amphetamine exposure in all healthy children, adolescents, drivers, or adults. The trial also excluded several substantial psychiatric and medical comorbidities, which limits generalization.

Which amphetamine was administered, and by what route?

The investigated product was amphetamine extended-release oral suspension, not methamphetamine, methylphenidate, lisdexamfetamine, a dextroamphetamine skin patch, or an injected solution. The investigators administered it by mouth once each morning. During the unblinded optimization phase, initial doses were 2.5 or 5 mg/day and increases were individually titrated toward clinical benefit and tolerability, with a 20 mg/day maximum.

For entry into the blinded phase, a child needed to tolerate a stable 10–20 mg/day optimized dose. Across treated participants, the report gives a mean final optimized dose of 15.4 mg/day (SD 4.10), corresponding to a mean 0.47 mg/kg (SD 0.16). These are administered product amounts derived from the treatment regimen. They must not be relabeled measured amphetamine concentrations in venous whole blood, plasma, urine, or saliva.

What did the observers and arithmetic tests measure?

SKAMP-Combined sums thirteen items reflecting observed classroom attention and deportment, using a rating scale on which higher totals mean more impairment. Trained assessors rated children in laboratory-school sessions. Its change-from-predose quantity is a behavioral score difference; the source’s negative active-minus-placebo contrast means relatively better recorded ratings with medication, not a negative drug level.

PERMP is a brief written arithmetic performance task: ten minutes of problems attempted and answered correctly. Unlike SKAMP, it records completed work, but it still measures school-task performance in the artificial classroom. Neither endpoint is a roadside psychomotor validation instrument, a toxicology screen, a blood assay, an eyewitness-deception test, or a validated one-person assessment of legal fitness to drive.

When were observations collected?

Assessments occurred immediately before the classroom-day morning dose, then at one, two, four, six, eight, ten, twelve, and thirteen hours after dosing. Accordingly, “onset by one hour” means a statistically significant separation was already present at the earliest sampled postdose time. The investigators did not run a continuous minute-by-minute study able to distinguish onset at, for example, twenty, forty, or fifty minutes.

Similarly, “duration through thirteen hours” means the effect remained detectably different from placebo at the final measured point. It does not demonstrate a precisely located offset at fourteen hours or later, nor does it imply that drug molecules disappear from biological specimens at the time a teacher-rating difference becomes nonsignificant. Actual clinical effect, detection, and impairment each require their own time-resolved evidence.

Why was four hours the prespecified primary endpoint?

The four-hour test was the primary comparison of adjusted change from each child’s predose SKAMP-Combined score. The investigators prespecified a mixed-effects repeated-measures model that accounted for study center, treatment, observation time, and their interaction, using a participant random intercept. The analysis did not simply compare two raw postdose score means without regard to their baselines.

At four hours, the observed mean score was 8.2 in the active group and 21.1 in the placebo group; the observed mean predose scores had been 17.3 and 15.5, respectively. The model-adjusted active-versus-placebo change was −14.8 rating points, with a standard error of 1.61 and confidence interval −17.9 to −11.6. The primary p value was less than .0001. That result supports a group difference in rated classroom impairment during the trial—not a quantified blood-amphetamine cutoff.

What are the actual behavioral effect estimates?

The original report provides an adjusted change-from-predose comparison at the primary four-hour endpoint and two other clock times in the continuous prespecified sequence. It also supplies an average of the multiple classroom observations for the arithmetic task. These results should be kept in their original units: teacher ratings and the number of written problems, not milligrams or nanograms of amphetamine in a specimen.

The one-hour and thirteen-hour estimates are not directly measured blood concentrations. The PERMP summaries are averages across repeated classroom time points, not additional independent children. The source marked differences at each sampled time point, but did not provide a high-resolution minute-by-minute onset curve.

Does a significant one-hour score specify absorption or peak concentration?

It does not. One hour is the first observation at which the study established a difference. Drug absorption and distribution may begin earlier, and the measured classroom outcome combines pharmacological, behavioral, situational, and measurement processes. The research design did not sample a blood concentration at thirty or sixty minutes to link a numeric parent-amphetamine amount directly to the contemporaneous score.

The paper discusses a separate prescribing-information description of a plasma concentration curve peaking approximately four to five hours after dosing. That background description was not generated from new drug assays conducted on these ninety-nine participants during this study. It cannot be cited as if the investigators obtained each child’s paired plasma profile, still less as if they measured the same person’s legal venous whole-blood concentration.

What happened at the final thirteen-hour classroom assessment?

The last source-reported SKAMP adjusted difference was −9.2 score points, standard error 1.61, p less than .0001. All scheduled examined postdose SKAMP-Combined differences favored the active arm. The trial’s onset-and-duration procedure assessed the ordered observation points, so the affirmative final result supports the statement that an observable group effect persisted through the final scheduled examination.

A trial with no fourteenth-hour observation cannot establish the first later time at which the groups converged. Nor did the protocol ask whether a given child with a particular dose, meal, sleep pattern, adherence history, or disease severity experienced the same effect across the full period. The estimated mean differences are not a clock calibrated for a specific other individual or an unrelated forensic event.

Did mathematics performance independently agree with the rating results?

The report found that PERMP problems attempted and problems correct were higher on medication across the repeatedly sampled classroom session. The pooled, model-based treatment differences across observation times were 35.7 problems attempted (SE 4.59) and 33.0 problems correct (SE 4.38); the source reported p less than .0001 for both. The two related endpoints corroborate a standardized academic-task benefit under laboratory conditions.

They are nevertheless not the same as the SKAMP rating and cannot establish independent pharmacokinetic replication. A faster arithmetic response may depend on practice, motivation, classroom structure, and task difficulty. More work completed also does not mean the medication abolished all inattentive symptoms, made every child equally accurate, or improved driving performance outside the tested setting.

Were any efficacy comparisons demonstrably nonsignificant?

The published efficacy analysis marked significant comparisons at the sampled postdose times for the prespecified behavioral endpoints. It did not publish an actual p-greater-than-.05 efficacy result in the same series that could properly be called a tested treatment null. Reporting a nonexistent nonsignificant one-hour or thirteen-hour contrast would be fabrication, and inferring equivalence for a different task from an unmeasured endpoint would be equally unsound.

The important null-type observations in this paper concern particular recorded events, not failed efficacy tests: no postbaseline pulse above the defined 110-beats/minute cutoff during the randomized week, no serious or severe treatment-emergent adverse events recorded, and no reported suicidal ideation or behavior. These are observed zeros within stated monitoring conditions, not a statistical proof that those harms cannot occur. Missing laboratory-drug measurements are unmeasured outcomes, not “negative” specimens.

What adverse events and vital-sign changes were recorded?

During the initial active dose-optimization stage, 64 of 107 exposed children reported one or more treatment-emergent adverse events. Decreased appetite was reported in 28 (26.2%); insomnia in 14 (13.1%); affect lability in 10 (9.3%). The blinded stage showed lower observed rates, but its participants had already undergone an optimization process and the observation interval was shorter.

In the randomized safety group, the report described 9 of 52 active participants (17.3%) versus 6 of 48 placebo recipients (12.5%) with one or more treatment-emergent events. A ten-millimeter-mercury or greater diastolic pressure increase occurred in six participants in each blinded arm. Those descriptive observations are not a reported adjusted adverse-event risk ratio or a test proving equal risk. The absence of a severe event in this modest sample cannot establish clinical safety for every future patient.

What do the safety observations actually count?

Safety reports use different denominators because the open-label exposure group, randomized safety group, and efficacy population are not identical. The numerical contrasts below are newly organized from the original report and are descriptive unless an inferential test was explicitly published. A small study with zero recorded severe events has not demonstrated that the risk is zero in a larger population.

Table 4 of the original source labels the blinded active-adverse-event column as n=51, whereas its prose and randomized-safety disposition support n=52 for the cited 9/52 = 17.3% total. The discrepancy should not be erased by inventing a denominator. For detailed individual assessments, obtain the original subject-level protocol and records.

Why are “no events observed” and “no effect” different?

An observed zero means the study did not record an event using its definition, population, sampling schedule, and reporting window. It does not establish that the true probability in the population is literally zero. Infrequent cardiovascular or psychiatric harms can be missed in a small trial; some effects may require longer exposure or a different clinical setting to detect.

A nonsignificant statistical comparison would require investigators to specify a hypothesis, estimate the difference and its uncertainty, and report that the evidence did not cross a prespecified significance criterion. This paper did not supply that for every descriptive safety count. Therefore the appropriate conclusion is limited: the monitored severe-event and pulse-cutoff categories had no recorded cases in the observed sample, while certain other adverse symptoms and blood-pressure changes did occur.

Did the investigators chemically measure the administered amphetamine?

The report provides no contemporaneous quantified parent amphetamine or amphetamine-metabolite results from postdose human venous whole blood. It also does not present child-specific analyzed separated plasma or serum drug concentrations, validated urine or oral-fluid confirmatory concentrations, hair results, sweat-patch concentrations, vitreous measurements, or drug concentrations in exhaled aerosol. The article contains baseline health-screening procedures, symptom observations, dose records, and vital-sign data, but not a method validation for an amphetamine toxicology assay.

It consequently reports no study-derived calibration curve, chromatographic mass-to-charge ion criteria, matrix-specific limit of detection or quantitation, extraction recovery, or whole-blood-to-plasma distribution ratio for the drug. Comparing reported milligrams swallowed with nanograms per milliliter obtained in a different case would require external validated pharmacokinetic information and relevant collection facts; this particular study supplies neither individual postdose drug concentrations nor that bridge.

Why do whole blood and plasma require separate treatment?

Venous whole blood includes blood cells and the liquid fraction; separated plasma is a distinct physical matrix generated after collection, and serum is different again. Drug partitioning between components can vary with compound properties, specimen processing, and method. Neither a general pharmacokinetic narrative from a product label nor a child’s classroom score is a scientifically valid conversion formula from plasma concentration to whole-blood concentration.

When a real toxicology report is at issue, the analyst must identify exactly which material was sampled, when, whether it was venous or another collection type, how it was stored, which parent drug or metabolite was targeted, and which quantitative method was validated. The separate biological-matrices guide discusses why specimens must remain tied to their measured compartment. None of those requirements can be bypassed by substituting this trial’s behavioral score.

Can the result define a last-use time or individual legal impairment?

No. The study is evidence for an average treatment-group difference under planned administration and classroom observation. It was not a retrospective time-since-ingestion validation study. A person’s recorded exposure, formulation, individual metabolism, dosing interval, drug-drug interactions, biological matrix, and sampling lag must be evaluated separately. The source does not provide the necessary paired chemically quantified time series for a last-use reconstruction.

The trial recruited children diagnosed with ADHD, not adult motorists or eyewitnesses, and administered a therapeutic product according to a controlled protocol. No participant completed a validated real-world driving test, hazardous-operation exercise, criminal-event assessment, or witnessed accident reconstruction. Evidence of better average classroom ratings and arithmetic performance cannot determine whether a particular motorist at a different time was intoxicated, competent, negligent, deceptive, or a cause of a collision.

What biases and applicability limits matter?

The five-week open-label optimization helped identify doses considered both effective and tolerable before randomization, potentially improving the apparent blinded-phase response compared with an unselected population. The trial also depended on one-week placebo exposure, an analog-school environment, trained raters, and a sponsor-involved design. Children with several major psychiatric diagnoses, significant uncontrolled illnesses, or prior amphetamine nonresponse were generally excluded.

Such choices can be reasonable for testing short-term classroom efficacy, but they restrict generalization. The participants were ages six through twelve, with 68 boys and 31 girls in the efficacy sample. That composition does not establish a separate treatment effect for every sex or diagnostic subtype. No prespecified inference showed differential effects by ethnicity, sex, or ADHD phenotype, and the authors explicitly stated the trial was not powered for several such subgroup comparisons.

How does this primary study differ from neighboring stimulant papers?

A separate 2019 eighteen-child pilot of oral amphetamine suspension used a blinded crossover and measured an earlier thirty-minute classroom endpoint, where observer-rated change was significant but arithmetic change was not statistically significant. The present 2018 study is a different set of ninety-nine efficacy participants in a parallel-group trial and its first postdose check was at one hour. Comparing the two experiments may be scientifically informative, but they are not pooled observations from one sample.

Similarly, a transdermal dextroamphetamine trial concerns drug delivery through the skin, a different optimization schedule, different absorption conditions and a separate cohort. Racemic amphetamine sulfate and methylphenidate studies also differ by drug formulation or agent. A search result that merely matches the word “amphetamine” should not be treated as proof the endpoints, original research sample, administered route, or statistical questions are interchangeable.

What can a forensic reviewer responsibly carry forward?

A responsible statement would be that one manufacturer-supported randomized parallel-arm trial of an optimized once-daily oral extended-release amphetamine suspension found improved blinded observer-rated classroom behavior and written arithmetic performance in ninety-nine analyzed pediatric ADHD participants at scheduled observations from one through thirteen hours. The four-hour SKAMP change difference was −14.8 points (95% CI −17.9 to −11.6), and the authors reported statistical differences through the last measured point.

Equally important is what that statement does not mean: the investigation was not a laboratory confirmation of a drug concentration in whole blood or plasma, was not a validated dose-to-effect conversion for a different person, and did not study an actual road event. For analytical questions, specific confirmatory drug-testing procedures and complete contemporaneous case records would be necessary. The scientifically defensible boundary is a population classroom effect, not a blood-level or legal-causation conclusion.

Frequently asked questions

Did the source prove oral amphetamine starts working at exactly sixty minutes?

No. The first postdose measurement was at one hour, so the study demonstrated separation by that observation, not the first minute when any participant experienced an effect.

Did the source prove the drug stops working at exactly thirteen hours?

No. Thirteen hours was the last tested classroom point; the study did not establish the subsequent disappearance of effects or how long the compound remains measurable in any body fluid.

Were there ninety-nine children in both active and placebo groups?

No. The ninety-nine children were the combined intent-to-treat sample: fifty-one active participants and forty-eight placebo participants. Repeated rating times were not extra independent participants.

Does the reported plasma peak describe blood actually collected from these children?

No. The text mentions prescribing-information pharmacokinetic background, not newly assayed paired plasma samples from this efficacy experiment; it contains no directly measured venous whole-blood amphetamine values.

Can fewer adverse events in the blinded week prove that side effects disappear with time?

No. The phases differ in exposure duration and selected participants. Descriptive event frequencies alone cannot establish a causal attenuation curve, and this report did not perform a specific between-phase randomized safety test.

Did the trial produce a statistically nonsignificant classroom efficacy endpoint?

The investigators reported significant treatment differences at scheduled efficacy assessments. They did not publish a contrary nonsignificant efficacy comparison for the same sequence, so no such null p value should be invented. Certain severe-event and vital-sign categories had zero observed cases, a different type of evidence.

May an expert infer the cause of a driver’s behavior from these scores?

No. The experiment involved therapeutic pediatric classroom observations without measured legal whole-blood drug levels, validated individual driving data, or an investigated incident. Individual causation needs independent case and scientific evidence.

Primary research

Childress AC, Wigal SB, Brams MN, Turnbow JM, Pincus Y, Belden HW, Berry SA. Efficacy and Safety of Amphetamine Extended-Release Oral Suspension in Children with Attention-Deficit/Hyperactivity Disorder. Journal of Child and Adolescent Psychopharmacology. 2018;28(5):306–313. Primary original article: 10.1089/cap.2017.0095.

This review is independent scientific interpretation of the original empirical measurements and their limitations. It does not reproduce the original figures or imply that an individual legal case has been evaluated.