Can an Amphetamine Liquid Improve Classroom Ratings Before Math Performance? An 18-Child Crossover Trial

Quick answer: In 18 children with ADHD, oral extended-release amphetamine suspension produced a significant observer-rated classroom improvement at 30 minutes, while the contemporaneous mathematics-task comparisons were not statistically significant. Both favored the active preparation at three hours. The trial did not quantify blood amphetamine or test individual driving impairment.

What is the answer from the original experiment?

In a small randomized crossover pilot involving 18 children diagnosed with attention-deficit/hyperactivity disorder (ADHD), an oral extended-release amphetamine suspension produced a statistically significant improvement in observer-rated classroom behavior at 30 minutes compared with placebo. At that same 30-minute assessment, the number of timed mathematics problems attempted and answered correctly did not show statistically significant active-versus-placebo differences. By three hours, both classroom ratings and the two mathematics measures favored the active preparation. The research therefore supports an early difference on one measured endpoint, not a conclusion that all aspects of performance had already improved.

The distinction matters whenever a scientific report is offered as proof of a drug’s real-world functional effect. A rating-scale difference, a timed classroom work product, a plasma drug concentration, and an individual’s ability to operate a vehicle are different forms of evidence. This pilot directly measured the first two. It did not conduct postdose forensic blood testing, an on-road driving evaluation, or an individualized accident reconstruction.

Which primary study is being analyzed?

The evidence comes from Childress, Kando, King, Pardo and Herman’s 2019 original clinical experiment, published in the Journal of Child and Adolescent Psychopharmacology. The researchers specifically asked whether amphetamine extended-release oral suspension (AMPH EROS) could produce a detectable improvement by the first 30-minute laboratory-classroom observation. It was an empirical crossover intervention, not a meta-analysis, a case report, or a retrospective database association.

The complete seven-page source provides the randomization sequence, dosing procedures, definitions of both tests, model-based comparisons, clinical observations, safety summaries and limitations. Its identifier is linked in the source section below. A separate, earlier large classroom investigation described a later first testing point; its participants and results should not be silently pooled with this independent 18-child pilot.

Why does the study have 18 children rather than 36 or 54?

There were 18 independently enrolled children, all included in the efficacy and safety analyses. Fourteen were boys and four were girls; the mean age was 9.0 years. Nine were assigned active treatment followed by placebo, and nine to the reverse sequence. Every child underwent the active and placebo double-blind classroom conditions. That is a two-period crossover using repeated observations on the same 18 people, not two independent treatment populations.

Predose ratings, the 30-minute measurement, and the three-hour measurement are repeated observations nested within those children. The study also included preliminary screening, medication adjustment, a practice classroom day and an open-label interval between blinded classroom visits. Counting each occasion as a new participant would falsely inflate precision and exaggerate the evidentiary scale. The statistical analysis accordingly modeled sequence, period and within-person effects.

How was the oral amphetamine actually administered?

The intervention was an oral liquid extended-release amphetamine suspension, not transdermal dextroamphetamine, methamphetamine, methylphenidate or inhaled amphetamine. During individualized dose optimization, study medication could begin at a nominal 5–15 mg based on treatment history and increase to as much as 20 mg daily according to tolerability and symptom control. The double-blind classroom doses were 15, 17.5 or 20 mg, matched to an identically sized placebo volume.

One child had no previous stimulant treatment. The mean daily dose during the blinded classroom portion was 15.8 mg; across the entire study, the mean was 13.3 mg. A nominal product dose establishes what investigators intended to administer. It is not a chemical determination of the parent drug reaching an individual’s brain or the concentration present in any bodily fluid at a specific time.

Why are the two classroom tests different?

The Swanson, Kotkin, Agler, M-Flynn and Pelham Combined (SKAMP-C) assessment is an observer-rated scale of classroom attention and deportment. It assesses visible behavior during the standardized classroom period. A lower score indicates fewer observed difficulties. Ratings can improve while a particular timed work task remains statistically indistinguishable from placebo, because the instruments do not measure the same variable.

The Permanent Product Measure of Performance (PERMP) is a timed ten-minute mathematics worksheet procedure. Investigators counted how many problems each child attempted and how many were correct. Problem difficulty was selected using a prior ability assessment. This task has different measurement error, baseline constraints, motivation demands and statistical power from an observer rating. A significant SKAMP-C contrast cannot be transposed into a significant PERMP contrast.

What happened to SKAMP-C at 30 minutes?

The observed mean SKAMP-C score at 30 minutes was 19.4 under active suspension and 26.3 under placebo. The unadjusted active-minus-placebo score contrast was −6.9, with a reported model comparison p=.0195 for this postdose score comparison. More importantly for the prespecified primary endpoint, the mean change relative to each condition’s own predose value was −6.1 under the active drug and +2.5 under placebo.

The primary model-adjusted active-minus-placebo difference in that predose-to-30-minute change was −8.6 points (standard error 3.01), with a 95% confidence interval from −14.91 to −2.17 and p=.0118. Those numbers are a between-treatment outcome comparison. They are not an amphetamine concentration, a sensitivity estimate for a toxicology test, or proof that every one of the 18 children had the same response.

Do baseline differences alter the interpretation?

Before the relevant classroom doses, the source reported SKAMP-C means of 25.4 for the active-day observations and 23.8 for the placebo-day observations. Simply comparing 19.4 versus 26.3 after dosing ignores those starting values and the pairing of each child with himself or herself. The trial therefore emphasized change from each respective predose score and a model that included treatment, period and sequence.

This is why the table below distinguishes the observed score comparison (p=.0195) from the prespecified change-score primary contrast (p=.0118). Both favor the active treatment, but they represent different estimands. Collapsing them into a single “amphetamine works in 30 minutes” claim would omit the endpoint definition and make independent review of the statistics more difficult.

What did the mathematics task show at the same 30-minute time?

At 30 minutes, the active-minus-placebo least-squares difference in mathematics problems attempted was +17.3 (standard error 11.01), with p=.1361. The corresponding comparison for problems correctly answered was +18.8 (standard error 11.19), with p=.1128. Both estimated differences point numerically toward more work under active medication, but neither meets the study’s conventional .05 significance threshold.

This is a genuine measured nonsignificant result, not an unperformed analysis. It does not prove that the true effect was exactly zero, and it certainly does not justify describing the two outcomes as established improvements. With only 18 children, the estimates were imprecise enough that the researchers could not exclude plausible differences. The observed positive point estimates and statistical uncertainty should be communicated together.

What changed by the three-hour assessment?

At three hours, the reported mean SKAMP-C change was −14.3 points with active medication and +2.9 with placebo. The model-based treatment difference was −17.2 (standard error 3.65), p=.0002. These values indicate a larger behavioral-rating contrast at the later measured observation point, within this specific laboratory experiment.

The mathematics outcomes also reached statistical significance at three hours: the active-minus-placebo difference was +60.3 problems attempted (standard error 12.93) and +61.6 correct (standard error 13.16), each with p=.0003. Those are group estimates from a short standardized classroom task. They do not quantify an individual child’s test-score improvement at school, and they cannot demonstrate effects at unobserved times between or after the two sampled postdose intervals.

Which numbers establish the important significant-versus-null distinction?

Table 1 separates the prespecified behavioral outcome, the observed behavioral score and the distinct mathematics endpoints. Reported differences are active minus placebo. Negative SKAMP-C values indicate fewer observed difficulties; positive PERMP values indicate more completed worksheet items. The source’s original standard errors and p-values are retained rather than reconstructed from individual data, which were not supplied in the published article.

Table note: Lower SKAMP-C reflects fewer observed classroom difficulties. PERMP is a worksheet-count endpoint. The source presents separate statistical comparisons at each time, and the two 30-minute PERMP results are not significant. Repeated visits do not increase the independent sample beyond 18 children.

What can the treatment effect size tell us?

The original authors calculated standardized effects using the fitted treatment difference relative to the model’s residual variation. The 30-minute change comparison yielded an effect-size estimate around −0.95; the three-hour SKAMP-C treatment contrast was larger, approximately −1.57. These within-study measures can be informative about the standardized task but should not be mistaken for the probability that any individual will respond.

The study was explicitly powered around an expected large contrast and recruited a small, selected sample. A seemingly large standardized effect can coexist with substantial uncertainty about broader populations. The effect is also dependent on a rating scale and laboratory context; it is not an effect size for driving safety, response to a sudden roadway hazard, criminal intent or an impairment threshold.

Does the word “onset” identify a biological onset time?

The earliest postdose research assessments were 30 minutes and three hours. Finding a significant active-placebo SKAMP-C difference at the first of those points supports a detectable group separation by that assessment. It does not locate the minute at which the drug began to act, the first minute of detectable amphetamine in a blood specimen, or the exact peak of exposure. No testing was performed at every intervening minute.

The formulation’s ion-exchange release system was discussed as a possible mechanistic explanation for prolonged delivery. The trial did not chemically confirm time-resolved parent amphetamine concentrations. An apparent onset on one behavioral endpoint therefore cannot be converted into a numeric legal sampling window or an exact time-since-ingestion statement.

Did the investigators quantify amphetamine in whole blood or plasma?

No postdose chemical amphetamine assay in legally collected venous whole blood appears in this trial. The publication likewise does not report quantitative plasma or serum amphetamine, relevant metabolites, urine or oral-fluid levels, hair concentrations, or a validated limit of detection for a forensic drug assay. The investigators monitored clinical pulse and blood pressure, but those physiological observations are not a drug-concentration measurement.

Whole blood includes cells and plasma; centrifuged plasma and serum are separate matrices with different analytical interpretation. Even if a later paper reported a plasma value after a related formulation, that number could not be silently attached to these 18 participants as though measured in their legal blood specimens. No conversion factor or concentration-to-behavior regression was established here.

What exactly was measured for safety?

The investigators recorded adverse events and vital signs during open-label dosing and the blinded sessions. During the open-label phase, 13 of 18 children reported at least one treatment-emergent event: six had events classified as mild and seven as moderate. Four participants had upper respiratory tract infection, three fatigue and two decreased appetite. There were no reported cases of insomnia, no serious events, and no adverse event that caused early withdrawal in this selected pilot sample.

The authors also recorded potentially clinically significant observations during the blinded portion, including five children whose measured pulse exceeded 110 beats per minute at a relevant time. Those screening-count criteria are not the same thing as proving the active dose caused tachycardia in those five children. Source Table 3 groups counts by assignment sequence rather than a contemporaneously measured individual amphetamine blood concentration.

How do the clinical observations constrain forensic conclusions?

Table 2 presents a small selection of actual denominators and safety observations, all tied to the 18 participants rather than an invented larger sample. Clinical vital-sign counts and adverse events deserve consideration in medication safety, but a small absence-of-event observation does not establish that a formulation is risk-free. The paper excluded children with relevant serious medical conditions and known prior lack of response to amphetamine.

For a legal or medicolegal question, clinically observed attention and pulse are contextual facts, not a validated surrogate assay. A purported inference that “a dose of 15 mg means a certain blood level,” or that an early SKAMP-C response rules out impairment in some later event, would require entirely different evidence.

Table note: The safety observations were descriptive, not designed as an independent controlled toxicity-risk model. Crossover treatment sequences are not separate groups of new participants; pulse counts do not identify drug concentrations.

Why is this study not a clinical blood–behavior correlation?

For a pharmacokinetic–pharmacodynamic concentration-response analysis, one would normally need actual quantitative samples from individual participants, with precisely documented sampling times, analyte identity, matrix, calibration, precision and other analytical details. An outcome score measured at a scheduled time cannot substitute for the missing chemical numerator. The administered milligrams are not the same variable as nanograms per milliliter in a sampled compartment.

Here the dependent variables were SKAMP-C observer scores and PERMP worksheets. No model links individual drug concentrations to those scores because no postdose drug concentrations were produced. Statements about amphetamine distribution, metabolism or blood-based detection windows may appear in other literature, but they were not directly tested in the original 18-child cohort.

How does the crossover design strengthen and limit the inference?

Within-person placebo comparison can reduce some stable between-child variation. The randomized sequence of active/placebo visits further helps separate treatment from order effects. All 18 completed the two conditions, avoiding the particular problem of substantially different observed sample sizes across arms. These strengths support a cautious group-level conclusion about the measured endpoints.

Nevertheless, the intervening days, earlier open-label optimization, practice session, selection of children already judged likely to benefit and the exclusion of known amphetamine nonresponders all constrain generalizability. The paper was single-site and did not enroll adults, impaired motorists or otherwise unselected community children. Blinding addresses some expectation bias but cannot guarantee that teachers or participants were unaware of effects during each period.

Can the pilot establish everyday academic performance?

Not directly. PERMP is a standardized short worksheet measure under controlled laboratory conditions, not a semester grade, school attendance record, independently adjudicated learning outcome, or a long-term curriculum assessment. By three hours the task favored active medication, but at 30 minutes the two math comparisons remained statistically nonsignificant. Whether that contrast generalizes to a different classroom period or a child with different baseline academic skills requires other evidence.

The study also reported varying worksheet difficulty matched to ability. The dependent variable is the number of correctly solved or attempted items under the assigned test conditions, not a universally comparable “intelligence gain.” A forensic statement concerning a specific person’s work, recollection or behavior would need examination of that individual’s records and circumstances.

Does the study test driving, witness memory or a crash?

It does not. There was no simulator or real-world on-road performance, no measurement of stopping distance or lane control, no eyewitness memory protocol, and no adjudication of a collision or injury. The participants were children with diagnosed ADHD completing laboratory classroom assessments. Translating their SKAMP-C ratings into legal driving ability would be both a population extrapolation and a domain-of-measurement extrapolation.

Even a genuine clinical benefit and a genuine adverse-event count cannot establish what happened in an independent, specific event. Exposure history, contemporaneous biological specimens when available, other medical conditions, environment, task demands and the actual event record are separate requirements. Here these case-specific inputs were unmeasured.

What should a scientifically sound expert report distinguish?

A defensible explanation begins by naming the original studied population (18 children), the actual compound and administration route (oral extended-release amphetamine), the three assay times relevant to each condition (predose, 30 minutes and three hours) and the fact that the same children crossed over. It should then distinguish primary SKAMP-C change from postdose level and the separate PERMP outcomes. The early behavioral result was statistically significant, while both 30-minute mathematics outcomes were not.

The explanation should further identify the actual biological specimen question: this paper contains no analyte-specific forensic blood assay. A nominal dose, a teacher rating, a timed work result and an amphetamine blood concentration are not interchangeable. When a proposed legal conclusion concerns intoxication, driving performance, timing of use or cause of injury, this paper provides neither the needed measured endpoints nor the individualized observations.

What did the researchers themselves acknowledge about uncertainty?

The investigators expressly discussed small sample size, potential underpowering of the early mathematics endpoints and limitations in applying their results to a more heterogeneous population. They also identified the possible impact of testing mathematics unusually early in the morning, as distinct from classroom rating. Those alternatives are plausible interpretations, not separately randomized explanations proved by this experiment.

Selection and dose optimization matter as well: children with known insufficient response to amphetamine were excluded, and the small open-label optimization stage preceded the blinded tests. A conditional clinical finding in this setting should not be restated as a guaranteed response in all children or an individual expert determination about a later legal event.

Frequently asked questions

Did amphetamine improve classroom behavior at 30 minutes?

In this 18-child crossover, the prespecified SKAMP-C predose-change difference favored the active oral suspension over placebo by −8.6 points (SE 3.01; p=.0118). That is a measured group-level observer-rating result, not proof of equivalent improvement in every other classroom outcome or in an individual outside the experiment.

Were children already completing more math at 30 minutes?

The 30-minute PERMP estimates favored active drug numerically, but problems attempted (p=.1361) and problems correct (p=.1128) were genuinely nonsignificant at the stated threshold. The null is a failure to demonstrate a treatment difference in this pilot, not proof of no possible effect.

When did the math performance comparison become significant?

At three hours, the estimated active-minus-placebo treatment differences were +60.3 problems attempted and +61.6 correct, both p=.0003. This is a short, controlled, difficulty-matched worksheet comparison. It cannot by itself validate sustained everyday academic success.

How much amphetamine was actually administered?

The optimized double-blind oral suspension doses were nominally 15, 17.5 or 20 mg once daily; the mean blinded-period dose was 15.8 mg/day. There is no study-specific quantitative assay showing each child’s circulating concentration, absorbed fraction or blood drug exposure.

Was blood amphetamine or an amphetamine metabolite measured?

No quantitative postdose parent amphetamine, metabolite, drug plasma, serum or forensic venous whole-blood result is reported. The study monitored clinical vital signs and adverse events rather than performing a drug-concentration-to-performance analysis.

Can a 30-minute response establish the time of last use or impairment?

No. The test sampled group classroom endpoints at scheduled times; it did not measure the biological onset moment, an individual last-use clock, driver performance, traffic incidents or legal impairment. A significant rating difference is not a validated forensic timeline.

Are these findings proof of safe driving or absence of intoxication?

No. The study enrolled children with diagnosed ADHD in a laboratory classroom, not motorists. It produced no driving tests, no impairment threshold, and no event-specific observations. The proper conclusion is limited to the original group measurements and their uncertainty.

Original primary research

Childress AC, Kando JC, King TR, Pardo A, Herman BK. Early-Onset Efficacy and Safety Pilot Study of Amphetamine Extended-Release Oral Suspension in the Treatment of Children with Attention-Deficit/Hyperactivity Disorder. Journal of Child and Adolescent Psychopharmacology. 2019;29(1):2–8. Original peer-reviewed study, DOI 10.1089/cap.2018.0078. This analysis paraphrases the original study and constructs new summary tables; it does not reproduce the original artwork.

Relevant same-site background

For additional method-level context, see Biological matrices and why results differ and Confirmatory drug testing and matrix-specific evidence. These guides explain how matrix selection and confirmatory procedures constrain forensic interpretations; they do not supply missing specimens for this trial.

Primary research identifier: 10.1089/cap.2018.0078. Educational scientific analysis; not individualized clinical, toxicological or legal findings.