Do Methylphenidate and Lisdexamfetamine Differ in 15-Hour Simulated Driving? A 44-Adult Crossover Trial

Short answer: In a randomized, double-blind crossover trial of young adults with attention-deficit/hyperactivity disorder (ADHD), optimized multilayer extended-release methylphenidate (PRC-063) was noninferior to optimized lisdexamfetamine (LDX) on a composite simulator-driving score assessed from one to 15 hours after the morning dose. The adjusted overall Tactical Driving Quotient (TDQ) difference, PRC-063 minus LDX, was −0.3 points (95% confidence interval −2.4 to +1.8), above the study’s prespecified noninferiority boundary of −7.5 points. That is a comparison between two actively treated conditions—not proof that either medicine eliminated impairment, prevented real crashes, or produced a particular whole-blood drug concentration. No quantitative forensic blood-drug concentration was reported.

The original research is Madaan, Bhaskar, Donnelly, and Cox (2024), A Randomized, Phase 3, Double-Blind, Crossover Comparison of Multilayer, Extended-Release Methylphenidate (PRC-063), and Lisdexamfetamine in the Driving Performance of Young Adults With ADHD, Journal of Attention Disorders 28(6):947–956, doi:10.1177/10870547241226634. This analysis distinguishes the trial’s directly recorded virtual-road performance from its treatment assignments, statistical estimands, adverse-event reports, and forensic propositions that were never tested.

What exact question did the trial answer?

The central question was comparative: after each participant received individually optimized doses of two ADHD medications, was average simulated driving with PRC-063 worse than with LDX by more than a clinically selected tolerance? Investigators prespecified a difference of −7.5 TDQ points, equal to one-half of the quotient’s standardized 15-point standard deviation. If the lower confidence limit for PRC-063 minus LDX stayed above −7.5, PRC-063 would meet the trial’s noninferiority definition. Noninferiority is narrower than proof that treatments are interchangeable for every person, identical in every outcome, or safer than not taking a drug.

The experiment did not ask whether an impaired driver would pass a roadside evaluation, whether any individual would avoid a collision, or whether one active drug was safer than placebo on the simulator. Both compared study conditions used active ADHD medication, and the study reported no medication-free simulator baseline. Its design is informative for the relative performance of these two optimized regimens in this setting, not for a broad claim that stimulant consumption improves road safety.

Who participated, and what are the actual denominators?

The disposition diagram reports 63 people evaluated at entry, 18 screen failures, and 45 independently randomized young adults with ADHD. The safety sample contained 24 men and 21 women. The sequence groups were 24 assigned to PRC-063 followed by LDX and 21 assigned to LDX followed by PRC-063. A single participant in the first sequence withdrew because of a work-scheduling conflict; 44 completed the crossover and formed the reported full-analysis population. Do not treat the 45-person safety denominator and the 44-person completed efficacy denominator as interchangeable.

Each completing participant experienced two treatment periods and repeated simulated drives within each period. Eight planned driving observations per person—four under each drug—are correlated repeated measures, not 352 independent patients. The efficacy model accounted for treatment, period, hour, sequence, and the treatment-by-hour term. Similarly, the number of adverse events and the number of subjects experiencing events are different quantities. The trial cannot support more statistical precision simply by multiplying the participant count by drives, individual driving variables, or adverse-event reports.

Were these drugs both amphetamine?

No. PRC-063 is a multilayer extended-release formulation of methylphenidate. Methylphenidate is a distinct molecular drug, not d-amphetamine and not methamphetamine. LDX is lisdexamfetamine dimesylate, a prodrug that yields active dexamfetamine after biotransformation; it is not equivalent to administering methamphetamine. The chemical distinctions matter whenever a clinical dose, a reported medication, or a forensic laboratory result is being interpreted. A test specifically detecting amphetamine or methamphetamine does not automatically quantify methylphenidate, and a generic positive screen cannot establish which prescription formulation was actually used without an appropriately targeted analytical method.

The study administered oral medication in a blinded crossover protocol. It did not administer a controlled recreational methamphetamine dose or evaluate a mixed-drug intoxication scenario. Differences between the optimized nominal doses, methylphenidate milligrams and lisdexamfetamine-dimesylate milligrams, should never be mistaken for equivalent molar doses, equivalent tissue exposure, or a particular quantity detectable in legal whole blood.

How were the dose schedules optimized?

Each participant had a treatment period lasting up to 21 days, with titration by a blinded clinician and a stabilization interval before driving assessment. PRC-063 dosing began at 45 mg/day; reported optimized doses were 45 mg in seven participants, 70 mg in 26, and 100 mg in 11, with a mean of 73.52 mg/day. LDX dosing began at 30 mg/day; the reported optimized categories were 30 mg for eight, 50 mg for 29, and 70 mg for seven, with a mean of 49.55 mg/day. These optimized-dose distributions correspond to the completion-analysis numbers, not two independently randomized populations.

Subjects attended a day of simulator assessments after morning dosing at approximately 7:30 a.m. The protocol then observed simulated driving at one, nine, twelve, and fifteen hours postdose. Thus the primary contrast represents treatment after titration to clinically selected doses; it is not a head-to-head test of fixed equivalent milligram doses or a single unadapted administration. Because the same person crossed into the second active medication without a medication-free driving comparison, the experiment does not isolate treatment efficacy against the person’s untreated driving behavior.

What did the driving simulator actually measure?

A stationary driving cabin and projected scenes created a virtual route combining about 2.6 miles of rural roadway, 4.3 miles of highway, and 2 miles of urban roads. The tactical task monitored 31 performance variables involving such domains as swerving, rolling stops, speeding, and collisions. Investigators selected 15 variables in advance for the composite TDQ, converted the constituent observations to standardized scores, and expressed the composite on a quotient scale centered at 100 with a standard deviation of 15. Higher scores indicated better performance relative to the study’s standardization procedure.

That numerical score is a derived simulator performance composite, not a direct laboratory assay of a drug. It is not an alcohol concentration, a measured reaction time in milliseconds, a recorded real-world crash rate, or a probability that a particular driver was safe. Its numerical scaling also means a −7.5 TDQ margin should be interpreted within this defined instrument; it does not translate automatically into a permissible change in steering deviation, an exact number of pedestrian impacts, or a statutory impairment limit.

What was the main noninferiority finding?

The model-adjusted mean TDQ was 99.9 (standard error 1.84) during PRC-063 treatment and 100.1 (standard error 1.84) during LDX treatment. The estimated PRC-063 minus LDX difference was −0.3 points (SE 1.08), with a two-sided 95% confidence interval from −2.4 to +1.8. Its lower bound was 5.1 points higher than the prespecified noninferiority boundary of −7.5. On that specified analysis, PRC-063 met the relative noninferiority criterion.

The trial separately reported p = .793 for a conventional comparison of treatment means. That nonsignificant result is not itself a proof of equivalence: a large p-value does not prove that two medicines cause exactly identical outcomes. The stronger, more disciplined interpretation is that the observed confidence interval excluded the prespecified magnitude of disadvantage for PRC-063, within this study’s selected participants, doses, simulator endpoint, and modeling assumptions. The findings cannot establish equivalence across every patient, adverse effect, time beyond 15 hours, or real-life traffic scenario.

How did the results differ at one, nine, twelve, and fifteen hours?

The investigators reported model-adjusted TDQ contrasts separately at four postdose assessments. Each interval spanned zero, and none of the four conventional pairwise comparisons reached statistical significance. At one hour the PRC-063 minus LDX difference was −2.4 points (95% CI −6.7 to +1.8; p = .258), at nine hours −2.2 (−6.4 to +2.1; p = .318), at twelve hours +3.1 (−1.1 to +7.4; p = .146), and at fifteen hours +0.3 (−3.9 to +4.6; p = .880). The treatment-sequence test was also nonsignificant (p = .4025).

Those observations are consistent with broadly similar TDQ averages across the measured interval, including the last simulated drive. They do not demonstrate the precise moment a drug started working, that every participant had sufficient attention at 15 hours, or that no late-day decrements could occur after the last observed time. This was not a continuous fifteen-hour road test. Nor should one calculate four independent clinical trials out of the four within-person measurement times; they were repeated, correlated outcomes from the same crossover participants.

What numerical evidence anchors the comparison?

Because the contrast is always PRC-063 minus LDX, a negative difference favors LDX and a positive difference favors PRC-063 on the TDQ scale. Rounding of component means explains why displayed 99.9 minus 100.1 does not exactly reproduce the model’s reported −0.3 contrast. The interval estimates, rather than the superficial similarity of rounded means, carry the noninferiority conclusion.

Why is noninferiority not the same as superiority or equivalence?

A superiority test asks whether there is sufficient evidence for a nonzero difference. A noninferiority test instead asks whether evidence rules out a disadvantage worse than a specific prespecified margin. Here, the investigator-defined loss that had to be ruled out was −7.5 TDQ points for PRC-063 relative to LDX. Because the confidence interval’s lower bound was −2.4, it excluded that chosen disadvantage, even though it still allowed a small PRC-063 disadvantage or advantage. This is statistically coherent: the same primary comparison can be nonsignificant for superiority while meeting noninferiority.

Noninferiority also depends on whether the margin has clinical meaning, whether participants and methods are appropriate, and whether the model handles incomplete observations and correlated trials. It should not be inflated into equivalence across all safety variables. The trial’s TDQ standardization and mixed-model framework answer a defined comparative question. They do not make the noninferiority margin a toxicological decision limit; a result of 7.5 TDQ points is neither a drug concentration nor a threshold separating sober from impaired drivers.

Did the study prove that either medication improved driving compared with untreated ADHD?

No. Although the authors described similarly maintained performance and clinicians saw symptomatic improvement during treatment, both experimental driving conditions involved optimized active medications. There was no study-arm simulator score collected under placebo or no medication for direct comparison. Without that comparator, the difference between PRC-063 and LDX does not quantify the improvement either produced relative to an untreated driving baseline. Statements suggesting that scores around 100 prove safe driving would likewise confuse an internally standardized quotient with an objective road-safety criterion.

The experiment is nonetheless more informative than a simple self-report comparison: participants performed a structured virtual driving task across several times after dosing, and investigators analyzed paired treatment observations. That comparative strength should be preserved without inventing a placebo effect. The reported clinical improvement ratings are not themselves an independent on-road safety validation. An expert assessment of whether medication influenced a particular driver’s behavior would still require contemporaneous observations and specific circumstances, not merely the treatment-arm means of this research cohort.

What did participants report about risky driving?

The secondary Cox Assessment of Risky Driving–3 (CARDS-3) questionnaire asked about how often participants had engaged in listed risky maneuvers during the preceding week. The adjusted treatment-period means were 20.7 (SE 1.32) with PRC-063 and 19.0 (SE 1.32) with LDX. The estimated difference of +1.8 points (SE 1.86; 95% CI −2.0 to +5.5) was not statistically significant (p = .351). The measure thus did not establish a treatment difference in reported risky behavior within the study, even though individual behaviors and recollections might vary.

Self-report and simulation are different evidentiary streams. A participant’s questionnaire about the past week’s driving is not a police crash record, a telematics trace, or a contemporaneous road examiner’s observation. Neither a nonsignificant questionnaire difference nor similar virtual TDQ means can establish that a specific person had no dangerous error at a particular later time. Careful interpretation names the measured endpoint and does not silently replace it with accident occurrence or crash prevention.

What do adverse-event figures show?

The study separately monitored treatment-emergent adverse events (TEAEs), clinical examinations, routine laboratory safety measures, and ECGs. The source reported 89 TEAE instances during PRC-063 treatment and 106 during LDX treatment; those are event counts, not the number of distinct people affected. The safety table expresses category frequencies as people out of 45 in the safety population for each active treatment. For decreased appetite the counts were 13/45 (28.9%) with PRC-063 and 23/45 (51.1%) with LDX; for initial insomnia 14/45 (31.1%) and 8/45 (17.8%); for headache 12/45 (26.7%) and 14/45 (31.1%).

These descriptive patterns deserve attention but should not be labeled statistically significant differences in the absence of a suitable stated test. Even an adverse event potentially relevant to alertness, such as insomnia, is not an observed driving-error outcome in a given participant. No serious adverse events or withdrawals attributed to treatment-emergent adverse events were reported. The one trial departure arose from a scheduling conflict, not a demonstrated medication-induced collision or toxicity.

What do dose and safety figures show together?

A dose-frequency table and an adverse-event frequency table have different denominators and interpretations. Dose categories describe completers after optimization; the study reports its safety results against the 45 randomized-and-exposed people. Counts within each treatment are not independent randomized arms, because the study crossed the same adults between the medications. The table therefore organizes the relevant source evidence without turning it into an unsupported causal relative-risk comparison.

Did the investigators measure methylphenidate or dexamfetamine in blood?

No quantitative forensic drug-level evidence is reported for this trial. There are nominal prescriptions, supervised morning dosing, safety hematology and urinalysis, and instrument-recorded driving observations. The paper does not supply directly assayed participant methylphenidate, lisdexamfetamine, or dexamfetamine concentrations in legal venous whole blood, centrifuged plasma, or serum at one, nine, twelve, or fifteen hours. Routine urinalysis in a clinical safety panel is not the same thing as a quantitative LC–MS/MS analyte-versus-time study. The study does not report oral-fluid, hair, vitreous, tissue, or breath stimulant concentrations either.

These missing data constrain any pharmacokinetic assertion. A 73.52 mg/day mean PRC-063 oral regimen is not a measured 73.52 ng/mL blood result, and the 49.55 mg/day LDX mean is not an active dexamfetamine plasma exposure estimate. Because the laboratory did not quantify the drugs in these persons, no concentration–TDQ relationship, elimination model, last-dose reconstruction, blood-to-plasma conversion, or per-se impairment threshold can legitimately be extracted from this experiment.

Could an observed TDQ value be used as a forensic impairment cutoff?

A TDQ value is an internally standardized task score produced by a defined virtual setting. The authors describe a quotient of 100 as the study mean, with scores near 85–115 representing roughly one standard deviation around that reference. Those conventions describe distribution on the chosen simulator scale. They are not validation results establishing an individual’s probability of a collision, sensitivity and specificity for legally impaired driving, or a decision boundary for an unobserved driver outside the laboratory.

For a forensic cutoff one would need to define the target condition, establish reliable measurement procedures and reference standards, identify relevant populations, and quantify classification error around a prospective threshold. This trial was designed around a between-medication noninferiority margin, not around identifying impaired individuals. It supplies no diagnostic confusion matrix, roadside specificity, causal crash risk ratio, or link to an individual’s contemporaneous whole-blood stimulant concentration. Moving its −7.5 margin into legal testimony as if it were an impairment threshold would misuse the study.

What are the strongest design features and most important limitations?

Its strengths include the randomized treatment sequence, blinded active-drug comparison, within-person crossover, standardized dosing-day procedures, predefined primary TDQ, and mixed-model treatment of repeated simulated drives. Individual dose optimization resembles clinical medication practice more closely than arbitrarily treating every person with the same fixed milligram amount. Recording measurements at four postdose intervals also makes the duration of the observed comparative effect more transparent than a single brief task.

Important limits remain. Only 44 young adults completed both arms, one participant withdrew for scheduling reasons, and the sequence groups had differing prior-stimulant proportions. There was no placebo or medication-free simulator arm and no direct chemical concentration measurements. The simulator’s virtual environment cannot reproduce the full complexity of public roads, unexpected hazards, personal driving history, fatigue, distractions, co-medications, or weather. The primary TDQ is a compound derived score, meaning component-level patterns may differ even with close overall means. The study was also sponsored by entities with interests in the studied drug development; appropriate interpretation requires reading its disclosures and independently evaluating both design and outcomes.

How does this relate to a measured stimulant result in an actual case?

A scientifically defensible case review begins with the actual specimen, collection time, analytical target, confirmatory instrument, units, measurement uncertainty, and chain of custody. A laboratory reporting methylphenidate, amphetamine, lisdexamfetamine, or methamphetamine may be answering different chemical and temporal questions. An observed medication history may help contextualize a result, but nominal dose and urine-screen status are not quantitative substitutes for a validated result in the correct matrix. For foundational distinctions between matrices, see the site’s published biological-matrices guide.

Specificity matters particularly for stimulants with different structures or metabolites. An immunoassay screen and a definitive confirmatory determination are not interchangeable, and a treatment study reporting no drug levels cannot validate a particular police laboratory assay. The published confirmatory-testing guide addresses the broader need to connect the analyte, method, and question. Therapeutic dosing and timed quantitative concentrations are also separate forms of evidence; the site’s therapeutic-drug-monitoring guide explains why actual sampling history is necessary before making exposure inferences.

Can the trial establish what caused an individual collision?

No. The trial established a group-based relative effect on a virtual simulator after two clinician-optimized ADHD regimens. It did not observe a particular traffic collision, reconstruct an accident, measure the driver’s contemporaneous drug concentration, or test interactions with alcohol, sleep deprivation, multiple sedating prescriptions, injury, illness, or roadway conditions. A decision about an individual incident would need a separate event-specific evidentiary record and appropriately validated reasoning, including the possibility that medication treatment, baseline ADHD symptoms, fatigue, external hazards, and chance contributed in different proportions.

Conversely, this study does not prove that prescription stimulants necessarily caused impairment or increased crash risk. The correct conclusion is bounded: at group level on this virtual task, optimized PRC-063 met the prespecified noninferiority criterion relative to optimized LDX through the observed 15-hour period. Both overclaiming impairment and overclaiming safety would transform a relative controlled trial into a finding it was not designed to make.

Scientific FAQs

1. Did methylphenidate outperform lisdexamfetamine overall?

No superiority difference was established. Adjusted overall TDQ means were 99.9 versus 100.1, with a PRC-063-minus-LDX difference −0.3 and p = .793. Noninferiority was supported because the lower 95% confidence limit, −2.4, exceeded the chosen −7.5 boundary. Those are distinct statistical propositions; neither proves identical performance for every subject.

2. Does this study show the medicines remain active in everyone for 15 hours?

No. It reports comparative group-level simulator observations at one, nine, twelve, and fifteen hours after supervised morning dosing. The final observed between-treatment contrast was +0.3 TDQ points, p = .880, but individual duration of effect and pharmacokinetic concentrations were not measured. It cannot establish coverage in all drivers beyond the sampled conditions.

3. How many people were actually compared?

Forty-five participants were randomized and entered the safety analysis; forty-four completed both treatment sequences and were described as the efficacy/full-analysis population. Two medication periods and four drives per period do not create new independent participants. Attrition and the analysis denominator should be disclosed alongside any quoted p-values.

4. Were both study treatments forms of amphetamine?

No. PRC-063 is methylphenidate in an extended-release formulation; LDX is a lisdexamfetamine prodrug converted to dexamfetamine. Neither exposure should be mislabeled as methamphetamine. A legal laboratory result must identify the actual measured target through a method validated for that target and biological matrix.

5. Does a normal-looking Tactical Driving Quotient prove lawful driving ability?

No. TDQ is a standardized virtual-task composite, not a validated impairment classification or personal crash-risk score. The study had no medication-free driving comparator, no direct roadside evaluation, and no confirmed person-level accident outcome. The −7.5-point noninferiority criterion is not a statutory driving threshold.

6. Can the optimized oral dose reconstruct venous whole-blood stimulant concentration?

No. The paper provides prescription milligrams and dosing-day procedures, but no quantified methylphenidate, LDX, or dexamfetamine in collected whole blood, plasma, or serum. Without an actual assay, patient pharmacokinetic data, validated model, and appropriate sampling times, an individual’s concentration or time of ingestion cannot be inferred reliably from this trial.

7. Do the adverse-event percentages establish a difference in simulator crash danger?

No. Decreased appetite and initial insomnia had different descriptive percentages in the safety table, but the trial did not equate these event reports with incident-specific crashes or demonstrate a statistically tested safety-risk contrast for each. Adverse-event occurrences, simulator-driving composites, and actual collision causation require separate analyses and evidentiary support.

What conclusion can this primary study actually support?

The central supported statement is precise: in 44 completed young-adult ADHD crossover participants after individualized dose optimization, multilayer extended-release methylphenidate PRC-063 satisfied a −7.5-TDQ-point noninferiority criterion relative to lisdexamfetamine on a four-timepoint laboratory virtual-driving composite; the adjusted difference was −0.3 points (95% CI −2.4 to +1.8). Other observed treatment contrasts and the self-report driving-risk questionnaire were nonsignificant. Distinct chemicals, participant and repeated-measure denominators, model-derived TDQ units, and the absence of quantified whole-blood drug results must stay separate.

For forensic purposes, the experiment supports relative performance under these experimental conditions, not a chemical test, a universal stimulant impairment rule, individual competence behind the wheel, or the cause of a particular collision. The primary source is Madaan and colleagues, 2024, DOI 10.1177/10870547241226634. Any application beyond the trial should identify the additional evidence and assumptions needed rather than treating the trial’s group means as an individualized legal finding.