Can Lisdexamfetamine Improve Attention After Childhood Concussion? What a 16-Child Crossover Trial Measured

What is the defensible answer to the central question?

The original experiment supports a limited, preliminary treatment signal: among the 16 children with complete paired data, several parent-reported attention and executive-function measures favored lisdexamfetamine, and the investigators reported fewer continuous-performance-test omissions during active treatment. That is not the same as proving universal cognitive improvement. It is not a large pediatric trial, it was not designed to determine a forensic drug threshold, and it did not enroll an unselected population of children immediately after concussion.

A particularly important caution comes directly from the investigators’ printed numeric table. The Conners Continuous Performance Test (CPT) omission comparison is listed with t = −1.36 and p = 0.027 for the 16 completers. Those two values do not reconcile under the ordinary two-sided paired t-test described by the authors. The lower omission mean is a reported observation, but the strength of its inferential support cannot be accepted uncritically without the underlying paired data or clarification from the authors.

Who actually participated, and how many independent children were analyzed?

Thirty-four prospective participants completed screening; 20 were enrolled after eligibility review. Three enrolled children never began the protocol, mainly because of scheduling difficulties. Seventeen began randomized treatment, and one additional participant stopped after an anxiety complaint shortly after changing from placebo to active drug. The principal on-versus-off treatment comparisons therefore used 16 completed cases. These are not 32 or 40 independent children merely because each completer contributed observations under two treatments.

The enrolled group was aged 10–16, evenly divided by sex, and predominantly White as reported in the demographic table. Most injuries were classified as mild on available clinical history. Time since injury averaged 7.85 months and ranged from two to 29 months. The sources of injury were 55% sports-related, 25% motor-vehicle accidents, and 20% other causes. Those accident histories are recruitment descriptors, not incidents for which the study evaluated drug exposure or causation.

How was the twelve-week double-blind crossover organized?

The trial randomly assigned children to receive six weeks of lisdexamfetamine first followed by six weeks of placebo, or the reverse order. Participants, their families, and the study team were blinded; Vanderbilt’s investigational-drug service prepared visually matching capsules. Formal neurobehavioral assessments occurred at baseline, six weeks, and twelve weeks; intermediate weekly visits supported dose adjustments, vital-sign checks and side-effect monitoring. The 16 completers contributed within-person treatment comparisons after the two phases.

The investigators did not include a washout interval at the crossover. They explained that tapering or washout was considered unnecessary under the protocol, but its absence remains relevant when interpreting treatment sequence and practice effects. For some tests, treatment-by-order interactions were reported even though a main treatment-order effect was not. Randomization and blinding improve internal validity, but they cannot erase small-sample variability, carryover concerns, selective enrollment or uncorrected multiple testing.

What drug and dosage were actually administered?

The active intervention was oral lisdexamfetamine dimesylate, a prodrug converted in the body to dextroamphetamine. It was not methamphetamine, and the study did not administer illicit stimulants. Every active treatment phase began with 20 mg once daily during the first week. If tolerated, planned escalation was to 30 mg in week two, 50 mg in week three and as much as 70 mg in week four, with slower increments or reductions when side effects arose. Children remained at their tolerated dose for the rest of that six-week period.

Among the 16 completers, the average final active dose was about 53 mg, whereas matching placebo escalation averaged an apparent 69 mg. Seven of 16 attained 70 mg on active treatment in the standard sequence; some required a lower or more gradual dose. Those milligram values describe administered medication. They cannot be converted from this paper into a child’s measured whole-blood dextroamphetamine concentration, peak level, sampling-time concentration or legal impairment threshold.

Which outcomes were directly measured—and which were not?

The directly collected research outcomes were formal cognitive-test performance, child and parent symptom-rating scales, weekly body weight, heart rate, blood pressure and treatment-tolerability reports. The Conners CPT assesses sustained response performance through task responses and errors; its omission count is not a chemically measured drug concentration. Parent Conners-3, Child Behavior Checklist (CBCL), and Behavior Rating Inventory of Executive Function (BRIEF-2) scores are structured reports of observed behavior. Child-completed anxiety scales measure reported symptoms, not arterial or venous toxicology.

Although clinicians conducted screening and included a pregnancy urine test where appropriate, the publication does not provide a validated quantitative assay of lisdexamfetamine, dextroamphetamine or related metabolites in legal venous whole blood, centrifuged serum or plasma, saliva, hair, urine, vitreous humor or actual exhaled breath. There are no patient-specific chromatograms, lower limits of quantitation, specimen collection times tied to drug concentrations, or bioanalytical concentration–effect plots. A clinical dosing record is not a laboratory specimen.

What did the original numerical evidence show?

The following values are reorganized from the study’s printed results, not re-created participant-level measurements. A smaller standardized parent-problem rating generally indicates fewer reported difficulties; interpretation depends on the instrument. Numeric contrasts are within-person treatment periods in 16 completers, rather than independent parallel groups. The first row requires special caution because its displayed test statistic conflicts with its displayed significance level.

Why is the CPT omission result important but not self-verifying?

An omission in this task occurs when an expected response is not made; fewer omissions can be consistent with better sustained attention. The reported average declined from 1.35 during placebo to .625 during lisdexamfetamine. That direction is a legitimate observation from the printed table. However, statistical significance in a paired crossover sample depends on each participant’s within-person difference and its variability, not simply on comparing the two marginal standard deviations or assuming that the paired observations are independent.

The published CPT row pairs the number t = −1.36 with p = .027, while the stated analysis was a two-sided paired comparison. For 16 paired observations the conventional test has 15 degrees of freedom; a magnitude of 1.36 does not normally produce .027. A transcription or analytic-labeling mistake is possible, but the source does not resolve it. That numerical defect is an evidentiary reason for restraint, not permission to substitute an invented corrected result or declare the entire trial worthless.

Did parent ratings, child self-reports and task performance tell the same story?

No. Parents tended to report better functioning on multiple measures while a child was on active treatment. Beyond the table, the CBCL anxiety/depression rating was lower on active treatment (reported p = .027). At the same time, children reported higher Revised Children’s Manifest Anxiety Scale total symptom scores (p = .012) and physical symptom scores (p = .025) while receiving lisdexamfetamine. Those findings do not support a simplified statement that all observers perceived fewer anxiety symptoms.

Physical sensations such as tension, headache, restlessness, fatigue or sleep disturbance may contribute to the child’s questionnaire responses, and the authors specifically noted overlap with residual post-concussion complaints. Differences among informants are meaningful data, not something to average away without a prespecified validated method. A parent may observe classroom or home behavior while a child feels physiological arousal; both can be real. Neither rating alone supplies an objective toxicological dose, nor does it establish whether one particular report was accurate in litigation.

Which reported findings were genuinely nonsignificant?

Some results that look numerically favorable nevertheless fail the paper’s declared p ≤ .05 threshold. The BRIEF-2 task-monitoring comparison was p = .054, and its emotional-regulation index was p = .052. Those are nonsignificant, not smaller versions of proven effects. The text also describes a baseline anxiety/depression moderator with p = .052 as though it were significant; using the study’s own threshold, that individual result does not qualify. The exact printed numbers matter, especially when multiple dependent measures are screened for effects.

Other performance tests—including working-memory, processing-speed and memory measures—were administered, but their overall treatment main effects were not all positive. The study notes treatment-by-order interactions for Digit Span (p = .026), Coding (p = .039) and verbal learning (p = .004). That means the apparent effect depended on whether medication or placebo was encountered first, consistent with the authors’ consideration of practice or sequence effects. It does not justify labeling those tasks uniformly improved across both sequences.

How do safety findings change the scientific interpretation?

The trial’s weekly vital signs provide directly measured clinical outcomes, rather than mere theoretical concerns about a stimulant. On active medication, mean weight fell approximately 1.05 kg, heart rate increased about 7.46 beats per minute, and diastolic blood pressure rose approximately 3.37 mmHg in comparisons described by the investigators. The corresponding reported p-values were .006, .002 and .028. These averages are based on the 16 completers and should not be construed as predicted responses for an individual child with a different medical history.

One child stopped the protocol because of an anxiety complaint after switching from placebo to lisdexamfetamine. Reduced appetite was reported by 75% of completers during active medication. The fact that the investigators did not observe further predefined severe withdrawal events in this highly selected, short-duration sample cannot establish an absence of long-term or rare adverse events. More importantly, beneficial attention measures and physiological adverse effects can coexist; improved performance on a computer task does not logically cancel an increase in pulse or an anxiety symptom.

What was the study’s actual treatment and safety denominator?

The denominator depends on the claim. Enrolled participants numbered 20; 17 began a randomized treatment period; 16 completed both phases and supplied the principal comparative efficacy data. The adverse event that caused withdrawal occurred outside the 16 completers. A statement that “only one child experienced anxiety” would not be justified because the paper reports various anxiety-related scale results and subjective side-effect complaints across different denominators and contexts. Nor should 75% be projected automatically onto all 20 enrolled children.

Similarly, repeated weekly weights, blood pressures, pulse readings and questionnaire administrations increase the number of measurements, not the biological sample size. Counting two six-week medication periods as 32 unrelated children would ignore crossover correlation. A forensic expert reviewing the design would also ask whether the source’s chosen analysis properly handled treatment order, attrition, carryover and multiplicity; these considerations influence how confidently group averages may generalize to other people.

How strong is the evidence given sample size and multiple comparisons?

The authors explicitly stated that they did not conduct a formal sample-size power analysis and applied no formal multiplicity correction to the many paired comparisons. They described the approach as exploratory, intended not to miss possible signals in a first study of this kind. With only 16 completers, one participant’s data can move an average or a significance test appreciably. Some p-values are close to .05, and a large family of simultaneous tests naturally presents a risk of chance-positive findings even when no overall correction is made.

This does not negate the trial’s prospective randomization, placebo masking, repeated structured assessment or carefully described titration. It does limit what the evidence proves. Confidence is greater for the specific recorded contrasts than for a broad therapeutic claim about all pediatric post-concussion symptoms. It is weaker still for outcomes the study did not measure, such as later return-to-school functioning, academic grades over years, on-road driving, legal competency, or causal attribution of injury symptoms in one disputed event.

Why are clinical cognitive improvements not forensic intoxication measurements?

Clinical outcome measures answer how participants performed or felt under randomized treatment phases. Forensic toxicology asks different questions: which analyte was detected, in what physical specimen, at what time, by which validated method, at what concentration, with what quality control, and how that evidence may relate to timing or effect. Neither a 20-mg starting dose nor a 53-mg average final dose is a measured dextroamphetamine concentration. The active prodrug’s identity also does not establish the presence of methamphetamine.

To make a concentration-based inference, an actual specimen and laboratory measurement are needed. Venous whole blood and separated plasma or serum are distinct matrices, and neither can be assumed from unspecified clinical monitoring. A negative pregnancy screen is not an amphetamine quantitation. A child with faster or steadier responses on a CPT does not become a calibration standard for blood amphetamine; a drug-positive sample on another date would require its own method and context.

What evidence would be necessary for a legally useful drug interpretation?

A defensible case-specific review begins with the contemporaneous chain of custody, identified matrix, actual drug and metabolite results, method validation, calibration, detection and quantitation limits, internal quality controls, collection time, prescribed regimen and clinical circumstances. If breath is sampled, it must be a real exhaled-breath analyte measurement rather than an inference from a cognitive test. If urine, saliva, hair, vitreous or tissue is tested, that result must be interpreted for that matrix rather than translated to a whole-blood number without validated support.

The additional question of effect may require reliable contemporaneous observations, clinical findings, a validated task closely matched to the proposed function and assessment of alternative causes such as TBI sequelae, fatigue, pain, anxiety, sleep deprivation or concurrent medications. An individual driver’s legal impairment is not an automatically transferable consequence of a small crossover group’s average pediatric symptom score. For general background on specimen distinctions and method-specific evidence, see the existing biological matrices guide and confirmatory drug testing guide.

What did the source establish about head injury mechanisms and causation?

It did not allocate children randomly to traumatic injury or to no injury, and it did not isolate the cause of any single child’s persistent symptoms. Participants entered because of prior TBI and ongoing attention complaints; injuries were predominantly mild and involved differing mechanisms. The trial’s randomized exposure was the treatment capsule, not the original accident. Clinical history included sports injuries and motor-vehicle accidents, but the paper did not reconstruct those events, estimate their forces, test fault, evaluate a witness, or connect measured medication levels to any injury.

It is therefore inappropriate to claim that a favorable treatment response establishes the initial injury’s biomechanics or demonstrates that specific current symptoms arose solely from that injury. Conversely, a child who did not improve on a particular rating scale would not thereby be shown to have no clinically meaningful post-concussion condition. The trial is evidence about selected symptoms under trial conditions and carries the ordinary limitations of clinical intervention research when applied to individual causation disputes.

What numbers summarize clinical benefit, statistical limits and safety?

This second table keeps different outcome types separate. Percentages and laboratory units are not interchangeable, and the study’s reported treatment-associated vital-sign changes are not estimates of a blood drug concentration. A p-value above .05 is labeled nonsignificant, even where an author’s surrounding prose sounds more favorable. Only outcomes with actual published data are listed; absence of an assay or driving test is recorded as unmeasured, never as a fabricated zero.

How should a reader reconcile a promising signal with unresolved questions?

The most faithful reading preserves two facts together. First, this carefully designed but small pediatric crossover study described potentially useful treatment responses among children with persistent attention problems after TBI. Second, the important CPT inferential-statistic mismatch, the number of unadjusted tests, several genuine null endpoints, informant disagreement, physiologic treatment effects, selection criteria and short follow-up materially qualify the strength and scope of those results. A good answer does not erase positive data, and it does not turn near-threshold p-values into confirmed findings.

The study’s investigators disclosed manufacturer funding, a factor relevant to critical appraisal but not a substitute for reviewing methods and data. Their initial findings may support further controlled research with larger and more diverse samples, prospectively defined primary endpoints, corrected analyses and explicit exploration of treatment sequence. The source alone does not resolve individualized drug concentration, diagnosis, safety or causation. Those questions require different evidence, not a more confident restatement of this same trial.

Frequently asked forensic-science questions

1. How many children actually completed both treatments?

Sixteen. Twenty were enrolled, 17 began treatment and one participant withdrew after an anxiety complaint. Treatment phases and weekly data points are repeated observations of the same biological participants; they do not double the number of independently studied children.

2. Did the research show that Vyvanse always improves concentration after concussion?

No. It found selected improvements in a small group of carefully screened children with persistent TBI-related attention complaints, and some outcomes were nonsignificant. The published CPT significance row is numerically inconsistent. Broader or long-term benefit to every child was neither demonstrated nor the target of the trial.

3. Was a 70-mg dose given to all participants?

No. Active treatment began at 20 mg daily and escalated cautiously only when tolerated, up to a 70-mg ceiling. The average final active dose in the 16 completers was approximately 53 mg; some remained much lower. The placebo period used blinded dose escalation without active drug.

4. Does lower CPT omission performance establish a particular dextroamphetamine blood level?

No. Omission errors are task responses. The paper did not publish a quantitative drug assay on venous whole blood or a different biological drug matrix. Administered medication, measured behavior and assayed chemical concentration are three separate kinds of evidence.

5. Were all executive and emotional outcomes statistically significant?

No. For example, the parent-rated BRIEF task-monitoring p = .054 and emotional-regulation index p = .052 failed the study’s threshold. Child-reported anxiety symptoms increased on active treatment even as some parent ratings improved. Treatment sequence also mattered for selected cognitive measures.

6. Can the study identify impairment or injury causation for an individual driver or witness?

No. It involved pediatric treatment and computer-based clinical tests, not on-road driving, validated individual impairment cutoffs or accident reconstruction. A particular case requires its own toxicology, examination, documented conduct, temporal evidence and analysis of alternatives.

7. Does the absence of additional serious withdrawals prove the medication has no safety concerns?

No. A small six-week exposure period in selected children cannot exclude uncommon or long-term adverse outcomes. The source recorded one anxiety-related withdrawal, frequent appetite effects, an average decline in body weight and increases in pulse and diastolic pressure. See the existing therapeutic drug monitoring guide for distinctions between dosing and measured concentrations.

Primary study and citation

Tramontana MG, Prokop JW, Williamson E, Duffie T, LaFever H. Traumatic Brain Injury-Related Attention Deficits in Children: A Controlled Treatment Trial with Lisdexamfetamine Dimesylate (Vyvanse). Brain Sciences. 2021;11(1):117. View the original complete primary empirical publication. The research reports group-level clinical outcomes, exploratory statistics and safety observations; it does not report quantitative forensic blood amphetamine measurements or validate a person-specific impairment conclusion.