Does Lisdexamfetamine Improve Attention More Than Guanfacine? A 22-Participant Crossover Study

Quick answer: In a randomized, double-blind, three-treatment crossover experiment involving 22 young people with ADHD, a single oral dose of lisdexamfetamine produced faster average responses and less variable reaction times on a composite of computerized tasks than placebo or extended-release guanfacine. Guanfacine increased response-time variability relative to placebo. The two omnibus drug comparisons were significant (p < .001), but numerous individual-task outcomes—including CPT omissions, interference inhibition, and time-discrimination errors—were not significant. This is evidence about selected laboratory attention measures, not proof of a particular child’s global function, a blood-drug concentration, or safety to drive.

The original primary experiment is El Masri and colleagues, Frontiers in Psychiatry (2025), article 1676472. Read the original paper: DOI 10.3389/fpsyt.2025.1676472. Its meaningful contrast is among three treatment conditions within the same ADHD participants, not three independent cohorts. Every conclusion below preserves that design and differentiates directly measured cognition from biological chemical analysis that was never undertaken.

What question did the original experiment ask?

The investigators were not testing whether ADHD exists, whether medication should be prescribed to a particular patient, or whether an acute drug concentration proves impairment. They compared two pharmacologically different ADHD treatments: lisdexamfetamine, a prodrug converted to dextroamphetamine, and guanfacine extended release, an alpha-2A adrenergic agonist. The goal was to determine whether a single administered dose had distinguishable immediate effects on laboratory measures of reaction speed, variability, sustained attention, inhibitory control and time discrimination. A placebo condition provided an internal comparator. Separately, typically developing youth completed an assessment without these study drugs, allowing descriptive comparisons but not a second randomized medication trial. The contrast matters: reaction time is a measurable behavioral endpoint, while drug exposure, clinical response and forensic intoxication are different propositions requiring their own evidence.

Who was actually studied?

The ADHD treatment sample comprised 22 independently enrolled children, adolescents and young adults, aged 8–20 years, with a mean age around 13.5 years and six female participants. Participants met diagnostic criteria and underwent screening and baseline assessment. Each person attended three medication visits; multiple task trials during these visits are not new participants. Twenty-two age-, IQ- and sex-matched typically developing participants remained in the comparison analysis after two outlier exclusions. The typically developing cohort did not participate in the three-medication crossover. Some composites in the ADHD analysis used data from 21 participants, and the Mackworth Clock outcomes used 20. Consequently, neither 66 medication visits nor the task-level trial counts should be reported as 66 independent treated subjects, and denominator changes should not be concealed.

What was randomized and when was testing performed?

Each ADHD participant received matching capsules containing lisdexamfetamine, extended-release guanfacine or a vitamin-C placebo on separate visits in a counterbalanced, blinded sequence. There was at least one week between conditions. Neurocognitive testing took place about three hours after dosing; the broader protocol’s functional-MRI schedule was approximately four and a half hours after dosing. The authors explicitly noted that the cognitive assessment time was a compromise because the two drugs’ expected time-to-peak profiles differ. Thus even a clean statistical comparison at three hours cannot be treated as a complete 24-hour comparison of the products’ effects. Visit order, residual variability, time of testing and performance familiarity remain relevant methodological considerations.

What doses were given—and were they fixed for everyone?

The study used oral, weight- and age-adjusted doses, not a universal fixed lisdexamfetamine dose. Following mild transient effects in two participants, the initial lisdexamfetamine plan of 30 mg below 30 kg and 50 mg above that weight was changed. Eight- and nine-year-old participants were assigned 20 mg irrespective of weight; older participants received approximately 0.5 mg/kg with rounding to the study’s practical dose increments. Guanfacine extended release was dosed around 0.05 mg/kg, rounded down to 1, 2 or a maximum of 3 mg. Placebo consisted of 10 mg vitamin C. These numbers describe what was administered, not the mass of amphetamine or guanfacine recovered from a laboratory blood tube. Product dose is also not interchangeable with free-drug molecular mass or measured plasma exposure.

How did the researchers define mean response time?

The mean reaction-time composite pooled appropriately derived speed measures from the Go/no-go, Simon and continuous-performance tasks. Lower values represent faster responding in this context. The composite helps evaluate a broader domain without elevating a single favorable task as the whole answer, but it also combines tasks with distinct control demands. That is why the specific test scores must still be reviewed. The original table reports average composite times under placebo, lisdexamfetamine and guanfacine. The authors used repeated-measures analysis of variance, consistent with the same participants contributing observations under all three conditions. A statistically significant composite does not establish clinical effectiveness in every environment or the response of an untested individual.

What did the primary reaction-time results show?

For the 21 ADHD participants with composite reaction-time data, the source reports a placebo mean of 392.65 ms (SD 57.84), a lisdexamfetamine mean of 370.31 ms (SD 54.08) and a guanfacine mean of 389.24 ms (SD 60.77). The drug-condition omnibus analysis was F = 8.61, p < .001. The paper also reports false-discovery-rate-corrected pairwise evidence favoring lisdexamfetamine over placebo and over guanfacine, with stated p = .003 for each comparison. These are within-person laboratory averages, not prospective prediction intervals for another person. The magnitude and precision of an individual effect cannot be reconstructed simply by subtracting condition means without individual paired data.

Why does reaction-time variability matter separately from speed?

Two people may share the same average response speed while one makes highly inconsistent responses, which may be especially relevant to maintaining attention. The coefficient of variation (CV) expresses response-time dispersion relative to mean speed, rather than being another measurement in milliseconds. The study’s composite CV means were 0.23 under placebo, 0.21 under lisdexamfetamine and 0.25 under guanfacine. The omnibus drug contrast was F = 9.00, p < .001. The authors reported lower CV under lisdexamfetamine and higher CV under guanfacine relative to placebo. Importantly, the printed CV dispersion values and some paired test-statistic/p-value combinations deserve independent checking before anyone uses them for precision calculations; the original table’s mean ordering is not, by itself, an independently validated clinical cutoff.

Original source Table 2. Values with ± are reported means ± SD, not confidence intervals. CV SDs in the original are subject to reporting questions and are not silently repaired here. Omnibus p values test drug conditions, not individual diagnosis or legally validated impairment.

Did guanfacine improve attention in the same way?

Not on the core reaction-time-variability measure in this acute protocol. The average CV under guanfacine was higher than under placebo, and the authors reported an adjusted pairwise comparison favoring placebo. The reaction-time mean under guanfacine was 389.24 ms, close to the 392.65-ms placebo mean, whereas lisdexamfetamine was faster. This supports a narrow single-dose, three-hour, task-composite difference; it does not prove that guanfacine generally worsens daily ADHD symptoms or fails as a longer-term treatment. Guanfacine’s intended benefits, expected pharmacokinetics, dose titration, and clinical tolerability can differ over weeks of use. One cannot import results from chronic treatment trials into this acute crossover or vice versa.

Which computerized outcomes did not show a significant drug effect?

Several outcomes did not pass the primary omnibus significance threshold. The original CPT omission-error drug comparison had p = .054; time-discrimination errors, p = .080. The Go/no-go inhibition measure had p = .121, the Simon interference reaction-time effect p = .860, CPT commission errors p = .259, and the Mackworth Clock omissions and commissions p = .311 and .313 respectively. The premature-response composite had p = .367. These results are essential to the interpretation: the primary supported treatment differences concern selected composites, not an improvement of every inhibition, vigilance, or timing measure. A result close to .05 remains statistically nonsignificant under the stated threshold rather than becoming a proven effect because its effect-size estimate looked sizable.

Why should exploratory favorable trends be called trends?

The paper’s authors performed exploratory pairwise comparisons for time discrimination and CPT omissions despite nonsignificant omnibus drug effects. For example, the reported lisdexamfetamine-versus-placebo pairwise comparisons had p = .078 for time-discrimination errors and p = .103 for CPT omission errors. Some other pairwise comparisons approached but did not cross .05. Descriptive means can guide hypotheses, but they do not supply independent confirmation. Presenting those findings as established improvements would systematically overstate the evidence. The same caution applies when many related task outcomes and planned or exploratory contrasts are considered together: statistical correction reduces, but does not erase, risks of selective emphasis and sample-specific exaggeration.

Did medication normalize the children to a typically developing benchmark?

The investigators compared ADHD participants with a separate typically developing sample after identifying the measures affected by treatment. These are not additional randomized drug conditions. Their reported group-comparison tests did not show significant differences for composite mean reaction time (p = .093) or coefficient of variation (p = .432) in the selected analyses. Importantly, failure to detect a between-group difference is not a formal demonstration of equality, equivalence or normalized functioning; those conclusions require prespecified margins, power and a suitable inferential design. The study also recruited a specific clinical sample in which autism spectrum diagnoses or elevated traits were common, and its sensitivity regressions did not show a significant additional explanation of the main composite results.

What did parents report about adverse experiences?

Parents or carers supplied short-term symptom reports after each condition. The study found a significant overall medication-condition difference in total reported adverse effects, with the greatest total burden following lisdexamfetamine; the source reports adjusted p = .002 versus both placebo and guanfacine. Its table listed frequent complaints such as reduced appetite and difficulty falling asleep under lisdexamfetamine. For example, the original count of poor-appetite reports is 13 under lisdexamfetamine, 3 under guanfacine and 2 under placebo; trouble getting to sleep is 10, 1 and 2, respectively. These are observations during a small acute trial, not long-term incidence estimates. The original table labels N = 21 yet its displayed percentages often follow a denominator of 20; these inconsistencies mean the count and percentage should not casually be treated as independently verified rates.

Counts reproduce original source Table 3 without converting counts into purported population incidence rates. That source labels N=21 for the table although many printed percentages correspond to denominator 20; this reporting mismatch limits independent percentage reconstruction.

Were fatigue and drowsiness themselves statistically different?

No significant omnibus differences were demonstrated for the individual ordinal symptom items evaluated with Friedman tests. Reported drowsiness had p = .285 and fatigue p = .262. This does not establish that neither drug can cause drowsiness, nor that one treatment is definitively safer. Parents’ short symptom questionnaires, small denominators, dose adjustment, three-hour task timing and lack of repeated long-term safety assessment limit the inference. Distinguish a frequency table that merely describes an observed count from a separately tested treatment effect with its own inferential result. The paper reports an omnibus treatment difference for the total adverse-effect tally, but that is not the same hypothesis test as each individual complaint.

Were task performance and clinical ADHD symptoms interchangeable?

No. A brief laboratory cognitive battery measures performance under standardized tasks and specific testing circumstances. ADHD diagnosis and clinical function encompass symptoms, school and home functioning, treatment tolerability, development and many contextual variables that a three-hour postdose test session does not capture. A child might perform more quickly in a computerized task yet experience adverse effects, or have symptoms affected by sleep, setting, anxiety or comorbid conditions. Similarly, a mean result cannot establish how a particular person would function in a traffic environment, workplace or courtroom. These distinctions are particularly important when the finding is used in forensic consultation: the research endpoint must be described exactly rather than converted into a different clinical or legal endpoint.

Did this study chemically measure lisdexamfetamine or dexamphetamine?

No quantitative drug concentration is reported for legally collected human venous whole blood, separated plasma or serum, urine, oral fluid, hair, sweat, vitreous or exhaled breath. The intervention was administered orally in blinded capsules. The measurements used for the report were cognitive responses, demographic information, screening, clinical histories and parent symptom questionnaires. Neither chromatographic parent-drug concentrations nor validated metabolites, detection limits, quantitation limits, timed blood curves or pharmacokinetic ratios were presented as original results. The correct statement is not that all concentrations were zero; it is that drug-specific chemical quantitation was not reported or established by this primary experiment. No concentration-response equation can be extracted from an unmeasured specimen.

Why are whole blood and plasma or serum distinctions legally important?

Forensic blood analysis must specify whether the laboratory analyzed an unseparated venous whole-blood specimen, or plasma or serum after separation. Matrix choice affects the composition and interpretation of a reported drug concentration. But here there was not even a reported drug assay in one of these matrices: the data concern administration of a product and observed task behavior. A saliva screening result, if available in a different context, would not become a venous drug concentration, and a nominal milligram capsule does not imply a measured nanogram-per-milliliter result. For methodological background, compare the published guide to biological specimen matrices and confirmatory drug testing. These guides provide general methodology rather than substitute evidence for the participants in this experiment.

Could these data establish safe driving or a person’s impairment?

No driving task, simulated road hazard, actual roadway incident, collision, eyewitness identification, or individualized fitness-to-drive cutoff was validated in this study. The composite reaction-time score is obtained under specific computerized conditions and does not directly sample visual scanning, vehicle control, sustained divided attention on the road, navigation, risky decisions or changing traffic events. Even if a task’s mean difference is statistically clear, it is insufficient to prove that a particular medicated child was fit or unfit to drive at a particular time. Such an opinion would also demand facts concerning clinical condition, drug history, contemporaneous behavior, any actually measured and validated analyte concentration, measurement uncertainty and the precise legal question.

What confounding and study-design limits remain?

Randomization and within-person contrasts reduce certain differences in baseline characteristics, but they do not eliminate all uncertainty. The small sample reduces precision; the single-dose design and dose adjustments complicate generalization; the task session at about three hours precedes one drug’s approximate peak exposure; two participants reportedly chewed extended-release capsules despite instructions; some outcome data are missing; and baseline diagnoses, prior medication exposure and developmental age vary. Nonparametric and sensitivity checks help, but do not replace a larger confirmatory trial. Effects reported for the selected composites therefore should not be treated as established for every child, dosing regimen, task environment or long-term outcome.

How should the evidence be used in an expert report?

An evidence-grounded expert should start with the actual N, comparator, exposure method and measured endpoint; quote material results with their tested denominator and uncertainty, and present the nonsignificant outcomes alongside the significant ones. Where a primary table contains internal reporting discrepancies, avoid silently inventing corrected statistics or computing purportedly exact individual effects. A useful analysis can explain that an acute lisdexamfetamine condition was associated with faster and more consistent laboratory responses while certain other domains did not achieve significant differences. It must not convert treatment assignment into drug concentration, medication response into exoneration or impairment, or abstract group evidence into cause of a specific accident or witness act. Site-specific scientific conclusions require a separate case record and validated relevant methods.

What is the strongest bounded scientific conclusion?

The source supports a treatment-condition difference in pooled computerized mean reaction time and variability under a single-dose blinded crossover in a small ADHD cohort. It also reports a greater total caregiver-observed short-term adverse-effect burden under lisdexamfetamine, an unfavorable variability trend under guanfacine, and multiple nonsignificant test-specific outcomes. Each proposition applies to the stated protocol and source-reporting quality. Nothing in these data directly determines a forensic whole-blood drug concentration, the exact timing of consumption, an individual’s overall neurocognitive ability, a definitive medication choice, or causation of an event. The distinction between direct observation, statistical association and unmeasured legal inference is the central scientific safeguard.

Seven questions readers commonly ask about this trial

1. Was lisdexamfetamine directly compared with guanfacine?

Yes. The 22 ADHD participants were assigned to both medication conditions and placebo across counterbalanced visits, with at least seven days between visits. The principal analysis used repeated measures; it was not a comparison of three independent sets of children.

2. Did lisdexamfetamine make every attention task significantly better?

No. Mean reaction time and its relative variability showed significant condition effects. The original CPT omissions p=.054, time discrimination p=.080, Go/no-go inhibition p=.121 and Simon interference p=.860 did not show statistically significant omnibus medication effects.

3. Was the same stimulant dose given to every participant?

No. The initial lisdexamfetamine dosing plan was revised following mild effects in two participants and dosing depended on age and body mass, with younger participants at 20 mg and older participants around 0.5 mg/kg subject to practical rounding. Guanfacine was also weight-based, capped at 3 mg.

4. Did the investigators assay dexamphetamine in venous whole blood?

No. The primary report does not provide chemical quantitation of lisdexamfetamine, dextroamphetamine or guanfacine in legal venous whole blood or separated serum/plasma, and it offers no direct chromatographic drug concentration data in urine, oral fluid, sweat, hair, vitreous or exhaled breath.

5. Did guanfacine prove harmful or ineffective as a chronic ADHD treatment?

No. It increased response-time variability on this single-dose, approximately three-hour computerized task protocol. A short acute crossover cannot establish the total long-term benefit-risk balance of a medication, especially when titration and expected peak times differ.

6. Were higher side-effect counts equivalent to serious injury?

No. The source recorded parent-observed short-term complaints. Counts and percentages in the original Table 3 have a denominator inconsistency, and the study did not establish an individual serious injury, a long-term incidence rate or a case-specific causal event.

7. Can faster mean response time establish safe driving or a credible witness?

No. The trial did not validate driver impairment thresholds, measure on-road safety, assess eyewitness reliability or test a causal event. A group-based task result cannot substitute for contemporaneous case facts and validated forensic analysis.

Original primary citation and the limits of inference

El Masri S, Kowalczyk OS, Chiu TH, Criaud M, Lukito S, Mazibuko N, Salazar de Pablo G, Perez Rodriguez V, Makos O, Lam S-L, Westwood S, Eaton-Turner A, Bozhilova N, Conti A, Santosh P, Roessner V, Kohls G, Mehta MA, Rubia K. Differential effects of a single dose of Lisdexamfetamine and Guanfacine on cognitive function in children with ADHD. Frontiers in Psychiatry. 2025;16:1676472. Original empirical research DOI: 10.3389/fpsyt.2025.1676472.

Experimental assignment, reported cognition and caregiver complaints are distinct categories of evidence from quantitative drug testing. None of the source’s means or statistical comparisons establishes a particular person’s drug concentration, last use, safe driving, impaired driving, courtroom credibility or the cause of a specific event. For a different kind of measurement question, the separately published therapeutic drug monitoring guide discusses why even measured concentration-time data require specimen and dosing context; such concentrations were not available here.