Do Boys Respond Better to Racemic Amphetamine? What Classroom Scores in 97 Children Actually Showed

Answer: In a randomized crossover laboratory-classroom study of 97 children with attention-deficit/hyperactivity disorder (ADHD), optimized racemic amphetamine sulfate improved observed classroom behavior and mathematical work in both boys and girls. Boys had worse scores on some directly observed classroom behavior ratings, but that did not mean they experienced consistently greater medication benefit. Exploratory treatment-by-sex interactions appeared at one two-hour time point on the SKAMP Combined and Attention scales, while the academic performance tests showed no significant treatment-by-sex interaction. The authors did not adjust these multiple exploratory tests for multiplicity. This was a behavioral efficacy study, not a blood-amphetamine or individual-impairment validation study.

The evidence comes from Childress, Newcorn, and Cutler’s original secondary analysis of a clinical trial: Gender Effects in the Efficacy of Racemic Amphetamine Sulfate in Children with Attention-Deficit/Hyperactivity Disorder, Advances in Therapy (2019). The central forensic question is whether a statistically significant subgroup comparison on a classroom behavior scale licenses a conclusion about a particular person’s pharmacological response, intoxication, dishonesty, or dangerous behavior. It does not.

What question did the researchers actually investigate?

The study asked whether boys and girls with diagnosed ADHD differed in their response to optimized racemic amphetamine sulfate, using several outcomes that do not measure identical constructs. The ADHD Rating Scale-IV (ADHD-RS-IV) reflects symptom history reported through an investigator and caregiver. The SKAMP instrument represents trained observers’ ratings of behavior during structured classroom activities. The Permanent Product Measure of Performance (PERMP) counts attempted and correctly solved mathematics problems over a short testing period. A change on one scale cannot simply be substituted for a change on the others.

This paper was explicitly an exploratory secondary analysis of a previously reported multicenter medication trial, not an independently recruited experiment designed or powered in advance to adjudicate sex differences. Its subgroup analyses were intended in part to estimate effects for future studies. When a post hoc finding occurs among many comparisons, its interpretation must take account of both the observed effect and the testing strategy.

Who were the independent participants?

One hundred seven children entered the overall protocol; ten left before double-blind randomization. The intent-to-treat randomized population was 97 distinct children aged six through twelve: 59 boys and 38 girls. Forty-seven were assigned to medication followed by placebo, and fifty to placebo followed by medication. Those sequences are alternative orders for treatment in the same participants, not an additional set of children. A participant measured six times during a classroom day remains one participant.

The mean ages were 9.5 years for boys and 9.7 years for girls. The sex groups had similar clinical characteristics on many reported measures, although girls weighed more on average at entry. Eligibility required ADHD diagnostic and severity criteria and restricted several comorbid conditions. Children with severe behavioral or medical complications were consequently underrepresented. The sample cannot be described as a representative group of all children using stimulants, let alone adults being evaluated after a traffic stop or workplace incident.

Which stimulant was administered—and which was not?

The medication was an immediate-release 1:1 racemic mixture of d-amphetamine and l-amphetamine, described in the paper as racemic amphetamine sulfate (RA-AMPH; marketed as Evekeo). This is not dextroamphetamine alone, lisdexamfetamine, methamphetamine, methylphenidate, or a formulation containing a different ratio of amphetamine stereoisomers. Those substances and formulations should not be treated as interchangeable merely because they belong to a broad stimulant category.

Children first underwent eight weeks of individualized open-label dose optimization. The protocol began at 10 mg daily, divided into two 5 mg doses, and allowed weekly upward adjustments generally in 5 mg increments according to response and tolerability. The reported mean final daily optimized dose was approximately 23.4 mg, and the median was 20 mg; these are prescribed treatment amounts, not analytically measured concentrations. The subsequent blinded crossover compared the optimized medication with placebo.

Why were the classroom-day dosing conditions unusual?

During most treatment days, children took their medication twice daily. On each full-length laboratory-classroom assessment day, however, staff administered only the morning dose, allowing observation through the following ten hours without an afternoon booster. This matters when describing the apparent end-of-day response. A decline in a rating effect at ten hours under that schedule does not prove that a child taking a different prescribed schedule would experience an identical decline.

Classroom measurements were collected predose and at 0.75, 2, 4, 6, 8, and 10 hours. The two school assessment days belonged to the crossover periods, not six independent randomized trials. The reported model included treatment order and period, along with treatment, sex, and the treatment-by-sex term. Missing outcome data were not statistically imputed; the subgroup available for a particular analysis may therefore be smaller than the full randomized group.

What did eight weeks of dose optimization show?

Both sexes had large decreases in ADHD-RS-IV total symptom scores over the open-label period. This supports observed improvement during supervised treatment and titration, but the optimization phase was not a randomized placebo comparison: symptom change during it cannot be attributed to medication alone with the causal precision available from the later crossover.

Before optimization, boys had higher hyperactivity/impulsivity subscale scores than girls (20.8 versus 18.4; reported p = .038). At week eight, the corresponding scores were 5.9 versus 5.6 and their comparison was not statistically significant (p = .710). On inattention, the groups likewise showed no significant endpoint difference (p = .664). These results do not provide a basis for a categorical statement that one sex benefits more from racemic amphetamine across ADHD symptoms.

How did the blinded crossover change classroom behavior?

The SKAMP Combined scale assigns larger numbers to worse observed performance or behavior. At the two-hour postdose point in the overall intent-to-treat analysis, the estimated SKAMP Combined score was 10.3 under racemic amphetamine versus 18.1 under placebo, with standard errors of 1.09 for each reported least-squares mean and p < .0001. The difference measures observable classroom behavior under study conditions, not a chemical assay result or an assessment of criminal intent.

When the investigators examined boys and girls separately at two hours, the model-estimated medication-minus-placebo SKAMP Combined difference was −9.9 points for boys (n = 57 for this comparison) and −4.8 points for girls (n = 38). Both within-sex treatment contrasts were statistically significant (boys p < .0001; girls p = .0002). The two missing boys in this particular outcome comparison should not disappear from the overall trial denominator or be interpreted as a different enrolled cohort.

A larger point estimate in boys is not the same as finding that all boys respond better than all girls. Effects are group averages. A formal interaction asks whether the treatment contrasts differed beyond what would be expected from sampling variability, and it requires careful treatment of multiple subgroup tests.

Were treatment-by-sex interactions significant?

The authors’ cross-sectional fixed-effects analysis found nominal treatment-by-sex interactions on SKAMP Combined at two hours, F = 5.86, p = .0174, and on SKAMP Attention at two hours, F = 4.71, p = .0326. The corresponding two-hour SKAMP Deportment interaction was not significant (p = .1069). Neither of the two PERMP academic measures had a statistically significant treatment-by-sex interaction at that time (attempted problems p = .4122; correct problems p = .0925).

In the same model the treatment term itself was significant for multiple outcomes at numerous postdose time points, showing broad improvement under medication. The interaction term addresses a different question. A significant treatment effect does not automatically establish a treatment-by-sex interaction; a sex main effect does not show that medication works differently by sex.

These are interaction p values, not the p values for medication versus placebo within boys or girls. Two low p values at one time point among many unadjusted exploratory tests should not be reformulated as a consistently established sex-specific biological mechanism. The paper states that corrections for multiple testing were not performed for these comparisons.

What happened to the sex interaction at later time points?

At four hours, the interaction p values for the SKAMP Combined and Attention scales were .8978 and .6364; at ten hours they were .4787 and .5968. Across the measured classroom day, the paper did not report a repeating series of significant treatment-by-sex interactions. The apparent two-hour difference was therefore time-specific rather than a durable demonstrated separation in comparative pharmacologic benefit.

The direct behavioral scores nonetheless were not numerically identical between boys and girls. Boys tended to have higher SKAMP Combined scores, with differences driven mainly by the Deportment component. An overall difference in how behavior is recorded under observation is relevant to understanding the scale, but is not a reliable stand-in for personal stimulant metabolism, susceptibility, or a legally relevant action outside the classroom.

Did math performance show different medication responses?

The PERMP task requires children to attempt an individualized set of up to 400 mathematics problems in ten minutes. Its two primary counts are problems attempted (PERMP-A) and correct responses (PERMP-C). The article reports medication-placebo separation in both girls and boys at all measured postdose points, with improved attempted and correctly completed problems under treatment. The formal treatment-by-sex comparisons for these metrics were not statistically significant across the assessed time points.

That result limits broad claims drawn from the two-hour SKAMP interactions. A difference in an observer-coded measure cannot be assumed to imply a corresponding difference in a direct output of short mathematics work. Conversely, success on that classroom task cannot certify competent driving, judgment under stress, or freedom from medication adverse effects. The PERMP estimates work product in a controlled environment; it was not constructed as an individual forensic performance test.

Why did SKAMP and ADHD-RS-IV disagree on baseline sex differences?

The ADHD-RS-IV covers parent- or caregiver-described patterns of ADHD symptoms, using an investigator interview about prior behavior. SKAMP uses near-real-time observations in a structured classroom and includes deportment, compliance, frustration-related conduct, and interpersonal interactions in addition to attention. Different respondents, time windows, and included behaviors can yield different profiles even when both scales are administered competently.

In this study boys’ higher SKAMP scores were principally associated with the Deportment domain. The investigators argued that this domain includes behaviors not fully represented in a core-symptom interview measure. That is a measurement-content observation—not a finding that boys were less honest, more dangerous, or morally culpable. Examples of items within an observational instrument do not establish the occurrence of a specific act in a particular child, much less a diagnostic or legal conclusion.

Can the results identify a child’s blood amphetamine concentration?

No. The source reports a prescribed amount of medication and repeated rating or task scores; it does not present a contemporaneous quantitative analytical series measuring d-amphetamine or l-amphetamine in each child’s venous whole blood, separated plasma, serum, urine, saliva, or hair. A prescribed milligram dose is not an actual measured nanogram-per-milliliter concentration, and classroom performance scores are not chemically validated proxies for either enantiomer.

For genuine toxicology evidence, the relevant question includes which specimen was physically collected, how its analytes were identified, whether enantiomers were separated, its calibration and quality-control conditions, and the collection interval. General principles of specimen comparability are discussed in the published biological matrices guide; the distinction between a screening indicator and an analyte-specific analytical result is explained in the published confirmatory drug testing guide.

Nothing in the paper validates converting a rating improvement into a whole-blood value, applying a plasma-to-whole-blood factor, reconstructing the last dose from SKAMP scores, or identifying whether another person ingested a particular product. The measurement gap is substantive, not a detail that can be repaired by making an assumption about average absorption.

Could an observed response prove individual intoxication or impairment?

The trial studied symptom management in diagnosed children selected for tolerability and clinical benefit, not intoxication. The fact that a medicine improves task performance in a treatment setting cannot establish the opposite proposition that a given blood finding implies a particular pattern of performance deterioration. ADHD symptom control, treatment response, stimulant side effects, recreational exposure, and forensic impairment are separate questions requiring different evidence.

No participant was evaluated in standardized on-road driving, a witnessed crash, a workplace accident, a criminal event, or a legal credibility examination. There was no individually calibrated equation linking sex, classroom scores, measured blood amphetamine, and accident probability. Claims about case-specific impairment, aggression, dishonesty, or causation would therefore be outside the scope of this paper regardless of how small an exploratory p value appears.

How should a forensic examiner treat the 0.0174 and 0.0326 p values?

Each number arose from a treatment-by-sex test at one observed time point. The comparisons were exploratory, the sample was modest for subgroup inference, and multiple scales were evaluated repeatedly across the day without multiplicity correction. Under these circumstances a low unadjusted p value is evidence about that specific fitted contrast, not a demonstration that a sex-specific response rule has been validated or will generalize.

The investigators themselves recognized the secondary analysis was underpowered for its moderation question. Lack of statistical significance elsewhere also does not prove exact equivalence between boys and girls: a nonsignificant test may reflect limited precision. The defensible interpretation preserves both directions of uncertainty—neither declaring an enduring difference from one nominal interaction nor declaring perfect equality because other tests were nonsignificant.

What design choices constrain the conclusions?

The sample passed through eight weeks of medication optimization before crossover, selecting children able to tolerate and benefit from the treatment protocol. That enrichment matters when estimating response rates in an unselected clinical population. Children with some substantial psychiatric or medical comorbidities were excluded. No independent trial was powered prospectively around the two-hour sex-interaction claim, and a much larger sample could have altered those subgroup estimates.

Open-label symptom change, blinded classroom response, and descriptive sex comparisons came from different phases and measures. The open-label phase used twice-daily medication, whereas the laboratory outcome day used one morning dose; the ADHD-RS-IV relied on reported recent symptoms and SKAMP on observed classroom behavior. The source was sponsored by the drug’s manufacturer, and some authors reported financial relationships. These features do not negate the recorded observations, but they belong in an assessment of external validity and evidentiary weight.

What would be needed to evaluate a different person’s stimulant response?

A person-specific assessment should begin with the actual drug and formulation, medically documented indication and dose schedule, timing, concurrent treatments, relevant baseline functioning, and the performance question at issue. If a toxicology result is available, determine the physical matrix, analytes, laboratory method, units, uncertainty, and specimen integrity. If a task outcome is being asserted, establish that it measures the legally relevant function rather than just a surrogate under artificial conditions.

When considering therapeutic exposure, a timed concentration can inform dose interpretation in an appropriate context, but it cannot replace observed clinical and task evidence. The published therapeutic drug monitoring overview separates concentration interpretation from the clinical question. A case-specific conclusion would still require contemporaneous facts and validated methods—not extrapolation from this pediatric trial’s average SKAMP and PERMP scores.

Frequently asked questions

1. Did racemic amphetamine work in girls as well as boys?

Both groups showed substantial improvement on the study’s ADHD symptom and classroom measures. The sex groups did not differ significantly in optimized ADHD-RS-IV total score, and PERMP academic performance did not show consistent treatment-by-sex interactions. One exploratory two-hour classroom behavior interaction does not override that broader pattern.

2. Why was the male two-hour SKAMP treatment difference numerically larger?

The fitted medication-minus-placebo SKAMP Combined difference was −9.9 points among 57 boys with that assessment and −4.8 among 38 girls. This is a group comparison on an observer-rated scale; it may reflect differences in rated behavior, available variation, or other study characteristics. It does not mean every boy benefits more than every girl.

3. Were the two significant interactions corrected for multiple comparisons?

No. The source explicitly states that the exploratory time-point and subgroup analyses did not apply multiple-testing adjustments. The SKAMP Combined p = .0174 and Attention p = .0326 were nominal two-hour interactions, while later corresponding interactions were nonsignificant.

4. Did the investigators directly measure amphetamine in whole blood?

No patient-specific quantitative whole-blood, plasma, or serum amphetamine concentration was a reported efficacy endpoint. The investigators administered a prescribed medication and collected ratings and test scores. The medicine’s dose in milligrams is not a measured biological concentration.

5. Is a worse SKAMP Deportment score proof a child lied or acted dangerously?

No. SKAMP is an observational behavioral rating instrument containing multiple coded items. A group score or a category label does not identify a particular dishonest act, violent event, motive, or causal mechanism in an individual.

6. Does a nonsignificant PERMP sex interaction prove biological equality?

No. A nonsignificant interaction indicates that a difference was not demonstrated at the selected significance level in this dataset. It does not prove identical responses, particularly in a secondary analysis not powered specifically to establish equivalence.

7. Can these findings support a DUI threshold or reconstruct the time of someone’s last dose?

No. The trial contained neither validated actual driving measures nor paired person-level amphetamine concentrations needed for those inferences. It cannot yield a lawful impairment cutoff, time-since-ingestion algorithm, or explanation of a specific incident.

Primary research and bottom line

Original empirical study: Childress AC, Newcorn JH, Cutler AJ. Gender Effects in the Efficacy of Racemic Amphetamine Sulfate in Children with Attention-Deficit/Hyperactivity Disorder. Advances in Therapy. 2019;36:1370–1387. https://doi.org/10.1007/s12325-019-00942-5. This discussion independently evaluates the original clinical data analysis; the study’s figures are not reproduced.

The reliable synthesis is narrower than the headline question: under this protocol, optimized racemic amphetamine benefited both sexes. Baseline observed deportment differed, a small number of unadjusted two-hour sex interactions emerged, and objective mathematics performance did not corroborate a stable differential medication response. The study measured pediatric symptoms and classroom tasks, not enantiomer-specific biological concentrations or individualized forensic impairment. Those boundaries are essential to avoid turning a group-level secondary analysis into unsupported testimony about a different person.