Did 30, 50, and 70 mg Lisdexamfetamine Help 76 Japanese Children With ADHD? What the Trial—and Its Safety Data—Show

Quick answer: In a four-week randomized, double-blind Japanese pediatric ADHD trial, participants assigned to oral lisdexamfetamine (LDX) 30, 50, or 70 mg daily had larger improvements in clinician-rated ADHD-RS-IV scores than participants assigned to placebo. At week four, the adjusted between-group score changes were −13.60, −15.32, and −13.69 points, respectively; all three comparisons were reported at p < 0.0001. That is evidence of short-term group efficacy, not evidence that 70 mg worked better than 50 mg, nor evidence of a child’s blood amphetamine concentration, individual impairment, or a forensic cause. The safety findings were also consequential: all 18 children in the 50-mg safety group reported at least one treatment-emergent adverse event, compared with 8 of 19 placebo recipients. The source is Ichikawa and colleagues’ original randomized trial, DOI 10.1089/cap.2019.0076.

What precise question did this clinical experiment answer?

The investigators wanted to know whether four weeks of assignment to one of three target doses of lisdexamfetamine would improve symptoms of diagnosed attention-deficit/hyperactivity disorder in a Japanese pediatric population, compared with an indistinguishable placebo. They also examined parent evaluations, clinician impressions, adverse events, body weight, pulse, blood pressure, electrocardiography, and routine clinical laboratory observations. This was a therapeutic efficacy and tolerability experiment, not a controlled driving study, drug-detection validation, or criminal forensic assessment. The primary endpoint was the change between baseline and week four on the Japanese version of the ADHD Rating Scale-IV, a clinician-administered symptom scale informed by parent or guardian reports about behavior at home and school.

Its statistical question was comparative and time-limited: how much did the randomized treatment groups improve relative to placebo during the monitored treatment period? An observed group mean decrease in ADHD symptoms is not the same thing as directly measuring synaptic dopamine, identifying an amphetamine molecule in a biological specimen, or determining how any particular individual would behave in a future setting. Those are distinct inferential tasks requiring different evidence. It also matters that these were patients with a defined diagnosis and eligibility restrictions, not a general sample of children, adult motorists, or people suspected of illicit stimulant exposure.

Who was actually studied: 76 people, not hundreds of observations?

The trial screened 103 patients and randomized 76 children or adolescents, ages six through seventeen, at 23 Japanese clinical sites. The four safety-population arms comprised placebo 19, LDX 30-mg target 19, LDX 50-mg target 18, and LDX 70-mg target 20. Seventy-one completed the four-week double-blind treatment. These figures describe distinct persons; the multiple weekly scores, clinician ratings, parent questionnaires, ECG traces, adverse-event entries, and repeated vital signs are observations from those same people rather than additional independent participants. An efficacy table at week four has different available sample counts—19, 18, 17, and 17—so the original randomized denominator should not be mechanically reused for every model estimate.

Most enrolled patients were boys. Across arms, the proportion male was approximately 72% to 90%, and average age was close to ten years. Baseline ADHD-RS-IV mean scores clustered between 37.1 and 38.1. The groups nevertheless differed descriptively in prior medical history and ADHD subtype composition. The primary adjusted model incorporated age categories and baseline symptom-score categories; randomization cannot eliminate all uncertainty in four groups this small. Its results do not automatically generalize to preschool patients, adults, patients with significant excluded cardiac disease, or individuals outside the diagnostic and treatment context. Children in a trial are also not the equivalent of human forensic blood specimens.

What drug was administered, and how did target-dose assignment work?

Lisdexamfetamine dimesylate is a prodrug: the administered molecule is converted to active dexamphetamine after absorption. It must not be called methamphetamine, and a nominal lisdexamfetamine capsule strength is not a measured dexamphetamine concentration. Treatment was administered by mouth once each morning. Although subjects were randomized to three target-dose groups of 30, 50, or 70 mg/day, all active groups began at 30 mg/day; doses were increased in 20-mg increments at weekly intervals until the assigned target was reached. Thus, at the early visits, the group labelled “70 mg” was not necessarily receiving 70 mg of lisdexamfetamine. This detail sharply limits any simplistic conclusion that an apparent early response was caused specifically by 70 mg rather than by common initial treatment.

The program included a four-week double-blind treatment phase, then a one-week single-blind placebo phase for those following that portion of the protocol, with follow-up or entry into a longer-term extension. The efficacy comparisons highlighted in this article concern the original four-week blinded phase. The study does not report an independently measured ng/mL plasma peak for each child, serial pharmacokinetic curves, or contemporaneous forensic whole-blood drug levels. Prescribed amount, estimated delivered dose, generated active metabolite, and chemically determined biological concentration are separate quantities. Assuming one from another would overstate what the investigators actually tested.

What did the prespecified primary ADHD symptom outcome show?

On the clinician-rated ADHD-RS-IV total scale, improvement is represented by a decrease in score. The week-four adjusted mean change was −2.78 points for placebo and −16.38, −18.10, and −16.47 for the 30-, 50-, and 70-mg target groups. Corresponding adjusted LDX-versus-placebo contrasts were −13.60, −15.32, and −13.69 points; each was statistically significant at p < 0.0001. The article’s analysis used a repeated-measures mixed-effects model and a serial gatekeeping strategy for controlling the family of primary pairwise comparisons. The statistical design gave priority to the 70- and 50-mg versus placebo comparisons before proceeding to the 30-mg versus placebo comparison.

This is a coherent positive finding for four-week symptom control in the enrolled population. It is not a randomized comparison demonstrating that 50 mg is statistically superior to 70 mg, even though its numerical improvement was largest. Nor does a significant score difference reveal whether a particular patient achieved a specific pharmacokinetic exposure, whether the drug would improve a tested driving skill, or whether every child responded. Subgroup interpretation, long-term safety, and the benefit-versus-harm decision for an individual require further information. In quantitative discussions, report the unit as ADHD-RS-IV score points, not percent improved, milligrams, blood concentration, or a validated measure of motor impairment.

Primary numerical evidence: randomized groups and week-four symptoms

Because the model estimated adjusted least-squares means and used repeated observations, these figures should not be treated as arithmetic scores for each child. Available participants at week four differ from initial arm assignments. Results are contrasts with placebo, not active-dose head-to-head tests. This organization is independent commentary on the numerical findings, rather than a reproduction of the source’s published table or figure. No new confidence interval, direct active-dose superiority test, or participant-level response rate has been computed from these aggregates.

Did the trial show a reliable dose-response relationship?

No formal monotonic dose-response relationship should be inferred from the three active-arm means alone. The 50-mg target had the numerically greatest adjusted ADHD-RS-IV decrease; the 70-mg target did not produce an even larger one. The arms were small, and the trial’s planned multiplicity-adjusted primary question was superiority to placebo, not a sufficient powered gradient or pairwise ranking between all active-dose levels. Individual factors such as developmental stage, timing, tolerability, comedication exclusions, and symptom-score variability would further complicate a dose-response claim. A change of a few score points between small trial arms is not automatically a pharmacological dose-response curve.

The weekly titration protocol reinforces the point. During the initial period every active participant started on the same nominal 30-mg capsule amount, notwithstanding the eventually assigned target arm. The early score trajectory therefore answers whether assignment to an eventual regimen was associated with benefit, not what happened after four weeks of uninterrupted exposure to a fixed 70-mg dose. The completed four-week data support short-term effectiveness against placebo across assigned arms. They neither prove that larger doses always provide greater benefit nor establish that increasing doses would be appropriate for a particular child. Prescribing remains an individual clinical matter outside the evidentiary scope of the paper.

What changed in inattention and hyperactivity subscores?

At week four, adjusted inattention subscale contrasts were −7.69, −10.66, and −9.05 points for the respective 30-, 50-, and 70-mg targets compared with placebo, with all source p values < 0.0001. On the hyperactivity/impulsivity subscale, the analogous differences were −6.25 (p = 0.0005), −5.13 (p = 0.0046), and −5.21 (p = 0.0038). Those results help show that the four-week total-score benefit was not limited to just one subdomain in the original fitted models. The measures remained questionnaire-based symptom assessments, not chemical measurements or direct tests of high-risk real-world performance.

One especially useful caution appears earlier in follow-up: the authors report that at week one, the hyperactivity/impulsivity improvement for the eventually 70-mg-assigned group versus placebo was not statistically significant, with p = 0.1099. The study otherwise describes significant comparisons at many other time points and subscores. This specific null result should be retained rather than omitted to make every result look uniformly positive. It also occurred while the target-70-mg arm was in the dose-escalation protocol, not necessarily at a stable 70-mg daily exposure. Neither p = 0.1099 nor an untested outcome means a zero biological effect: the former is a reported nonsignificant comparison, while the latter supplies no corresponding experiment at all.

What did the parent questionnaires and clinician impressions add?

The Conners 3 parent scale offered a separately reported viewpoint, rather than a chemical corroboration of drug exposure. At week four, the combined inattention-plus-hyperactivity/impulsivity subscale differed from placebo by −9.96 points for the LDX 30-mg target (p = 0.0076), −12.41 for the 50-mg target (p = 0.0015), and −10.04 for the 70-mg target (p = 0.0082). The publication describes statistically significant results for many Conners 3 subscales at weeks three and four, while expressly exempting the conduct-disorder subscale from that generalization. Without a reported exact p value for every excluded comparison, it would be unsound to invent a null estimate or describe an unreported effect as disproven.

At week four, clinician global improvement rated “much” or “very much” improved was reported in about 61%–71% of active-treated participants, compared with about 11% of placebo participants. Parent global assessment gave approximately 56%–65% active versus 16% placebo. These are clinically meaningful descriptions of perceived and observed symptom change within the trial. The numerators and percentages depend on who completed each particular assessment, and should not be reinterpreted as an independent blinded driving or biochemical outcome. A parent’s observation can support a clinical symptom scale but does not prove a child’s measured drug concentration, future injury risk, or capacity to perform a specific forensic task.

What safety signal appears when the adverse events are organized by arm?

The safety population consisted of 76 patients. For at least one treatment-emergent adverse event, the reported group proportions were 42.1% with placebo, 68.4% with the 30-mg target, 100.0% with the 50-mg target, and 70.0% with the 70-mg target. Especially prominent were decreased appetite and initial insomnia, with headache also reported in active groups. Because these groups contained only 18 to 20 children each, the exact proportion in one arm is not a reliable ranking of longer-term risk across all dose strengths, but it is still part of the actual observed safety record and should not be omitted from a benefit-focused explanation.

The source reported no deaths, severe adverse events, or serious treatment-emergent adverse events during the short study. That is a statement about what was seen in this selected cohort over this observation period, not a demonstration that serious outcomes are impossible or that long-term pediatric treatment is risk free. Three active-treated patients stopped treatment because of adverse events, according to the reported discontinuation details. A clinician making a prescribing decision would weigh these events alongside symptom benefit and individualized monitoring. A forensic scientist should likewise avoid inferring either the presence or absence of a particular adverse effect in an unstudied person from these group frequencies.

Numerical safety evidence from the same original experiment

TEAE means treatment-emergent adverse event and does not necessarily establish causal attribution for each occurrence. The table displays observed proportions and means, with no invented inferential p values for differences between safety arms. The 50-mg target’s 100% proportion is striking but came from 18 children, and the variation across targets should not be presented as a statistically confirmed inverse dose effect. Weight changes were observed clinical measures; they are not measures of amphetamine in blood, excretion, metabolism, or a proven mechanism of psychiatric response. Comparisons beyond the source report would need the underlying individual data and a prespecified statistical analysis.

What did pulse, blood pressure, body weight, and ECG actually show?

At last evaluation, mean diastolic blood pressure changed by −2.54 mmHg in placebo recipients and rose by 2.51, 7.76, and 2.65 mmHg in the respective 30-, 50-, and 70-mg targets. Mean pulse-rate change was −3.77 beats/min with placebo and +7.49, +11.56, and +7.45 beats/min with the active targets. Mean body weight increased by 0.5 kg on placebo and decreased by the active-group values in the safety evidence table. The authors reported that blood pressure and pulse changes were not classified as treatment-emergent adverse events, not that these physiological variables lacked measurable differences or that those shifts must be benign for everyone.

For QTcF, one participant each in the 50-mg and 70-mg groups exceeded 430 milliseconds at some point, and none exceeded 500 milliseconds. No ECG abnormalities were classified as treatment-emergent adverse events. Those are the study’s observations under its entry restrictions and monitoring rules; they are not broad individualized safety thresholds for forensic interpretation. Pulse, blood pressure, weight, and electrical intervals are physiological measurements with their own units and error structures. None is a substitute for a chemically specific assay of amphetamine, and none by itself establishes misuse, illicit source, driving fitness, a particular dose, or a causal explanation for an external incident.

Did investigators actually measure blood dexamphetamine concentrations?

No quantitative chemical concentration of administered LDX or resulting dexamphetamine in human forensic venous whole blood is reported as a study outcome. The source describes administration of oral capsules, clinical laboratory safety evaluations, vital signs, symptom scales, and ECG assessment. It does not supply a validated LC–MS/MS or GC–MS assay for parent dexamphetamine in whole blood, matrix-matched calibration, identification tolerances, limit of detection, limit of quantitation, drug concentrations in ng/mL, sampling times relative to each person’s clinical assessment, or metabolite-to-parent ratios. It would be an error to turn “50 mg capsule” into “50 ng/mL blood” or to infer a peak concentration from group mean symptom scores.

Forensic human venous whole blood is not synonymous with separated plasma or serum; those matrices can differ in drug partitioning, preparation, and concentration reporting. Urine is commonly informative about exposure or excretion, oral fluid about a different compartment and collection context, hair about extended deposition history, and breath about yet another analytical problem. None of those was the chemically quantified drug exposure endpoint in this therapeutic trial. The absence of such testing is a study-design boundary, not a nonsignificant pharmacokinetic result: no quantitative blood-drug hypothesis was tested in the reported analysis.

Can a laboratory infer an individual’s impairment from a successful ADHD trial?

No. Randomized group improvement in a diagnosed pediatric population does not test whether one individual was impaired on a roadway, at school, during a particular altercation, or in any other contemporaneous legally disputed event. The trial did not implement validated on-road driving, simulated driving, divided-attention field sobriety tests, accident reconstruction, eyewitness reliability tests, crash observation, or person-specific blood amphetamine concentration–effect analyses. Those endpoints must be marked unmeasured, rather than labelled negative or assigned a fabricated p value. An ADHD symptom rating over weeks cannot be mapped directly onto a momentary blood level or a particular behavior.

Equally, evidence that clinicians rated symptom improvement after supervised treatment does not establish that a positive amphetamine class screen in someone else was due to prescribed LDX, methamphetamine, or a specific route of administration. Chemical identification, biological compartment, source interpretation, medical records, administration timing, pharmacokinetics, alternative causes, and actual behavioral evidence all matter. Clinical medication efficacy may be relevant to a different medical question, but it does not validate legal impairment cutoffs. Any expert opinion that jumps from the four-week mean symptom difference to certainty about one person’s dose, use time, subjective experience, intent, or incident causation exceeds the original experimental observations.

Which null results deserve as much attention as the positive ones?

The most directly relevant genuine nonsignificant efficacy result is the LDX 70-mg-target group versus placebo comparison for hyperactivity/impulsivity at week one, p = 0.1099. It should not be quietly replaced with the significant week-four result or mislabeled a primary trial failure. Separately, the baseline age comparison (p = 0.9963), baseline body weight comparison (p = 0.6416), and baseline sex distribution (p = 0.5576) did not reach conventional significance. These help characterize enrolled groups, but null baseline tests do not prove the samples were physically or clinically identical, nor do they increase the underlying randomized sample size.

The publication also reports no serious or severe treatment-emergent adverse events within the trial; that is an observed zero event count, not a statistical finding that long-term serious risk is zero. Conduct-disorder-subscale outcomes were not uniformly within the reported significant Conners comparisons, but source details should control their exact interpretation; do not invent an exact inferential value. Conversely, not measuring forensic whole-blood drug concentration, drug metabolites, psychomotor driving capability, or future incident causation does not constitute any p > 0.05 null result. The disciplined scientific habit is to separate genuinely tested nonsignificance, descriptive zero events, and endpoints never assessed.

Why do brief duration and eligibility restrictions matter?

Four weeks of blinded comparative treatment is informative about short-term symptom efficacy, but too brief to establish durable effectiveness across years, developmental trajectories, uncommon adverse reactions, dependence potential in the general population, or long-term growth and cardiovascular effects. One-week placebo follow-on and later extension options do not automatically convert the four-week primary controlled comparison into a long-term randomized result. The sample of 76 was selected from 103 screened children, and key comorbid conditions and clinical exclusions reduced the relevance of the findings for patients outside those criteria. The predominance of younger boys also constrains subgroup generalization, particularly to older adolescent girls.

Industry funding and involvement were transparently reported: Shionogi and Shire supported the trial, and named authors included pharmaceutical-company personnel. That fact should be considered in ordinary conflict-of-interest review without being treated as proof that the observed results were fabricated or necessarily invalid. A better assessment examines prespecified endpoints, allocation, masking, statistical gatekeeping, participant flow, missingness, independent replication, and outcome ascertainment. The significant primary contrasts can remain genuine while clinical transportability and safety in an individual remain uncertain. Responsible interpretation preserves both the trial’s therapeutic evidence and the limits of its design.

What does this add to forensic toxicology evidence assessment?

The paper illustrates a general chain of inference that experts should not collapse. The first link is exposure assignment: a randomized clinician administers a known capsule amount. The second is biochemical exposure: the compound is absorbed and converted, but this study did not quantify that process in a forensic matrix. The third is a measured endpoint: here, clinician and parent ADHD ratings, vital signs, weight, and adverse events. The fourth is generalization: estimating how those outcomes relate to an unstudied individual, time, disease state, task, or event. This clinical trial supports some links and leaves others entirely open. A forensic interpretation needs the physical and temporal evidence for the inference it is asked to support.

For specimen-based interpretation, begin with the biological-matrices guide rather than equating a dose with a result; review the principles in confirmatory drug testing before treating a screening category as molecular identification; and distinguish the monitoring problem described in therapeutic drug monitoring from a retrospective legal impairment question. These are existing published contextual resources, not citations claiming that they independently verified the Japanese pediatric trial. The trial source, methods, and actual measured endpoints remain the controlling evidence for its own results.

Bottom line for clinicians, scientists, and courts

A real 76-person Japanese pediatric randomized experiment found substantial four-week ADHD symptom-scale improvements in three lisdexamfetamine target-dose arms compared with placebo. A real source-specific early nonsignificant hyperactivity comparison, meaningful observed adverse-event burdens, and uneven completion should accompany any summary of those positive primary results. The randomized arms were small; target doses were achieved through escalation; there was no demonstrated statistical ranking among active doses; and some clinically important longer-term or subgroup questions remained unanswered. This is more informative than saying merely that a stimulant “worked” or “did not work.”

Forensic extrapolation requires even more restraint. Oral capsule amount is not an assayed whole-blood amphetamine concentration; a pediatric ADHD rating is not a validated roadside psychomotor assessment; and a therapeutic clinical trial is not an experiment on individual intoxication, last dose, culpability, collision cause, or witness reliability. Preserving those boundaries strengthens rather than diminishes the legitimate therapeutic evidence. The correct question is what the original investigators measured, what comparison their model actually supported, and which proposed legal or toxicological conclusions would require entirely different data.

Frequently asked questions about the 76-person Japanese LDX study

1. Did the study prove lisdexamfetamine improves ADHD symptoms at all three target doses?

It provided statistically significant four-week adjusted ADHD-RS-IV improvements against placebo for each randomized target-dose arm: 30, 50, and 70 mg/day, with reported primary between-group p values below 0.0001. Because every active group began at 30 mg and increased by protocol, target assignment and actual dose on each early day must not be confused. The evidence is about these groups during a monitored short trial, not a universal guarantee of individual response.

2. Was 70 mg statistically better than 50 mg?

No such superiority follows from the published primary comparisons. The 50-mg target had the largest numerical adjusted score improvement, but the trial’s planned hypothesis tests compared each active target against placebo. A numerical difference among active groups is not itself a powered, multiplicity-adjusted head-to-head result. Dose optimization in an individual requires clinical assessment, not ranking this table’s means as if they were an established pharmacological gradient.

3. Which outcome was genuinely nonsignificant?

The source specifically reported p = 0.1099 for the eventually 70-mg-assigned arm versus placebo on the hyperactivity/impulsivity subscale at week one. That is a tested comparison with insufficient evidence of a difference at the conventional threshold in that early measurement. It does not undo the significant week-four contrasts, and it does not imply that drug blood concentrations or driving outcomes were tested and found null.

4. Why were adverse events most frequent in the 50-mg group?

All 18 participants in that safety arm reported at least one treatment-emergent event, compared with 13/19 at the 30-mg target and 14/20 at the 70-mg target. The trial does not demonstrate from those descriptive counts that a 50-mg target is inherently riskier than 70 mg in all pediatric patients. Very small groups, event definitions, and four-week follow-up limit dose-risk inference. Appetite loss, insomnia, headaches, pulse, and weight changes nevertheless warrant attention.

5. Did the researchers test whole blood, plasma, or urine for amphetamine?

The publication does not report a validated quantitative drug assay or individual amphetamine concentrations in legal venous whole blood, separated plasma or serum, urine, oral fluid, hair, breath, or vitreous. Clinical laboratory safety testing is not the same as chemical quantitation of the administered prodrug and active metabolite. No limit of detection, limit of quantitation, or concentration-time pharmacokinetic profile should be invented for this experiment.

6. Can these results predict an individual child’s future driving impairment or a legal incident?

No. The subjects were diagnosed pediatric patients studied for symptom control, and the measured outcomes were clinical scales and safety observations. There was no contemporaneous legal whole-blood concentration-effect model, validated individual driving test, crash analysis, or event-causation assessment. Those proposed conclusions were unmeasured; it would be scientifically wrong to convert them into nonsignificant p values or affirmative legal opinions.

7. Where is the original study, and what information should a reader verify?

The source is Ichikawa H, Miyajima T, Yamashita Y, Fujiwara M, Fukushi A, and Saito K (2020), Phase II/III Study of Lisdexamfetamine Dimesylate in Japanese Pediatric Patients with Attention-Deficit/Hyperactivity Disorder, Journal of Child and Adolescent Psychopharmacology 30(1):21–31, DOI 10.1089/cap.2019.0076. Check participant disposition, actual titration, mixed-model adjusted week-four effects, the early hyperactivity null, safety denominators, and limits of the short selected cohort. The paper reports therapeutic outcomes, not forensic specimen validation.