Were Mixed Amphetamine Salts Noninferior to Modafinil for Sleepiness? A 44-Adult Trial

Answer first: A 12-week randomized, blinded clinical trial in 44 adults with idiopathic hypersomnia or narcolepsy type 2 found that both modafinil and immediate-release amphetamine-dextroamphetamine were associated with improvements in reported sleepiness. However, the trial did not establish that amphetamine-dextroamphetamine was noninferior to modafinil on its prespecified primary Epworth Sleepiness Scale outcome: the noninferiority test returned p = 0.11 against a two-point margin. Noninferiority was supported for selected secondary self-reported symptoms, including a sleep-inertia questionnaire and a hypersomnia severity index. These results do not identify an individual person’s blood amphetamine concentration, establish that medication restored objective vigilance, or validate safe driving. The drugs were administered therapeutically; the study did not perform quantitative forensic drug analysis on blood, urine, oral fluid, or other bodily specimens.

The original investigation is Trotti and colleagues, Modafinil Versus Amphetamine-Dextroamphetamine For Idiopathic Hypersomnia and Narcolepsy Type 2: A Randomized, Blinded, Non-inferiority Trial, published in CNS Drugs (2024), DOI 10.1007/s40263-024-01122-y. Its most important interpretive distinction is not whether the mean Epworth scores happened to look similar; it is whether the uncertainty around their difference excluded a clinically important disadvantage for the amphetamine group. It did not.

What scientific question did the investigators actually test?

The investigators examined whether a traditional stimulant could perform closely enough to an established treatment to meet explicitly defined noninferiority criteria in adults with central disorders of hypersomnolence. This is a therapeutic comparative-effectiveness question. It is not an analytical-toxicology validation, a driving-simulator experiment, or a study of recreational stimulant use. The comparator was modafinil, while the experimental regimen was the immediate-release combination amphetamine-dextroamphetamine. Calling that combination “methamphetamine,” “lisdexamfetamine,” or pure dextroamphetamine would misidentify the administered substance.

The primary endpoint was reduction in the patient-reported Epworth Sleepiness Scale (ESS) from baseline to week 12. Secondary endpoints concerned patients’ impressions of improvement in overall illness, sleepiness, brain fog, and difficulty waking, plus numerical symptom questionnaires. A finding on any one of those instruments answers only its designated endpoint. It does not automatically demonstrate improvement in reaction time, sustained attention under sleep deprivation, divided attention while driving, or every dimension of hypersomnolence.

Who participated, and what is the correct biological sample size?

Forty-four adults took at least one randomized study dose: 22 assigned to modafinil and 22 to amphetamine-dextroamphetamine. Their average age was 35.4 years (standard deviation 8.9), and 37 were women. Thirty-three had idiopathic hypersomnia and eleven had narcolepsy type 2. This diagnostic mix matters because treatment responses in the relatively small narcolepsy subgroup cannot be estimated with the same precision as results in the larger combined sample.

Each participant was assigned to one drug for twelve weeks; this was a parallel-group study, not a four-condition crossover. Week-zero, week-four, week-eight, and week-twelve questionnaire scores represent repeated observations of those people, not independent cohorts. There were 44 independent participants in the modified intention-to-treat analysis, even if a regression used several measurements per participant. Initial screening of 93 people likewise does not make 93 treated study subjects.

What doses were actually administered?

Modafinil was supplied in 100 mg capsules and immediate-release amphetamine-dextroamphetamine in 10 mg capsules. Both regimens started with one capsule in the morning for one week. Participants could titrate upward at approximately weekly intervals to the highest tolerated, clinically useful dose, with optional morning and lunchtime dosing. The highest permitted total was modafinil 200 mg twice a day, 400 mg/day, or amphetamine-dextroamphetamine 20 mg twice a day, 40 mg/day. A participant who developed an adverse effect could return to a lower dose.

Among participants completing treatment, the average final daily dose was 360 mg (SD 74) modafinil or 34 mg (SD 7.5) amphetamine-dextroamphetamine. Those are amounts prescribed and taken as study medication; they are not ng/mL measurements in blood. The investigators pooled different tolerated doses into each randomized treatment arm, which answers a pragmatic regimen question but does not supply a controlled dose-response comparison for any one concentration.

Why does the trial’s design matter to interpreting its claims?

Participants and investigators were blinded to the randomized medication identity, and dispensing and assignment were managed by a separate investigational pharmacy. This improves protection against expectation bias relative to an unblinded drug comparison. But the study included no placebo-only arm and intentionally used self-reported clinical endpoints. Thus it cannot isolate every observed within-person improvement from expectation, natural variation, or regression to the mean, nor prove comparable effects on objectively measured wakefulness.

The primary time point was week twelve. To handle missing questionnaire scores or dropout, the investigators specified a modified intention-to-treat analysis with last observation carried forward. This preserves assigned participants in the main comparison but can make a final-week estimate depend on a person’s earlier score rather than a genuine completed week-twelve observation. The authors also performed per-protocol analyses. Both approaches have limitations when discontinuations are appreciable and treatment tolerance differs between arms.

What does noninferiority mean, and how is it different from a nonsignificant difference?

In a superiority test the usual null hypothesis is no difference, and a nonsignificant result means the evidence did not establish a difference at the chosen threshold. It does not demonstrate equivalence. A noninferiority study begins with an explicit maximum acceptable disadvantage. For continuous ESS change, this study specified a two-point margin. The relevant uncertainty interval had to rule out a decrement of that magnitude for amphetamine-dextroamphetamine relative to modafinil.

The observed improvements were close in absolute size, but the interval around the difference remained compatible with a clinically meaningful disadvantage. The primary p = 0.11 therefore means the required noninferiority criterion was not met. A separate ordinary comparison gave p = 0.64, which likewise does not establish sameness. The primary interpretation is consequently precise: both arms improved on average, but noninferiority of amphetamine-dextroamphetamine on ESS was not proven in this study. It would be equally wrong to claim that failure to prove noninferiority proves the drug was inferior.

What exactly happened to Epworth scores?

The reported mean ESS improvement was 5.0 points (SD 2.7) in the modafinil arm and 4.4 points (SD 4.7) in the amphetamine-dextroamphetamine arm. An improvement is a decrease in reported sleepiness. The standard deviations summarize variation among individuals; they are not assay uncertainty, randomization errors, or standard errors of the mean. Similar-looking averages should be considered alongside substantially different variability and the 22-person sample in each group.

For the binary outcome of at least two-point ESS improvement, the trial used its prespecified categorical margin and also failed to demonstrate noninferiority (p = 0.57). It is important not to conflate the continuous two-point ESS margin with the proportion-based 0.30 margin used for categorical responders. Both were research choices about what disadvantage would be acceptable, rather than a physiologically validated threshold for vehicle operation or impairment.

Which secondary measures supported noninferiority?

The study found noninferiority on the Hypersomnia Severity Index (HSI) improvement: means 8.0 (SD 5.6) for modafinil and 10.4 (SD 6.8) for amphetamine-dextroamphetamine, with noninferiority p = 0.003. On the Sleep Inertia Questionnaire (SIQ), improvements averaged 19.0 (SD 15.9) versus 18.2 (SD 18.4), noninferiority p = 0.04. The study also met selected noninferiority criteria for patient global impression categories, including much/very much improvement in cognitive dysfunction and sleep inertia.

These outcomes refer to reports about symptoms over the study period. A person’s perceived ability to wake up, concentrate, or function is important clinically, but it differs from a measured vigilance lapse or brake-response time. Furthermore, a favorable noninferiority test for brain-fog symptoms does not convert the earlier failed ESS result into a successful primary trial. The endpoints are related but not identical; each requires its own test and interpretation.

What does the original quantitative outcome evidence show?

The strongest caution from this table is that the patient global impression entries have a different amphetamine denominator from the full randomized arm. For those categorical questions, 10/20 should not silently be rewritten as 10/22. The distinction changes the proportion and the reported decision rule. Nor should “good” and “any” improvement categories be interchanged: the cognitive outcome met noninferiority for the stricter category but missed it for the broader category.

Why is the cognitive-dysfunction result easy to overstate?

For much or very much improvement in cognitive complaints, ten of twenty rated amphetamine-dextroamphetamine as beneficial, compared with five of twenty-two rated modafinil that way. The trial reported a small p value for noninferiority. That is not a prespecified test of whether amphetamine-dextroamphetamine was statistically superior to modafinil at restoring cognition. The authors expressly distinguished these analyses.

Moreover, the term cognitive dysfunction here described patient-reported difficulty thinking, focusing, or experiencing brain fog. Participants did not receive a comprehensive controlled psychomotor and on-road driving battery tied to simultaneous quantitative blood-drug concentrations. It would be scientifically indefensible to translate a favorable questionnaire category into a measured increase in complex divided attention for a specific driver on a specific day.

Why were some noninferiority tests positive while the primary test failed?

Each endpoint has its own clinical scale, noise, distribution, responder definition, missing-data pattern, and selected margin. The authors used a two-point ESS margin; the categorical PGI-C endpoints used an absolute 0.30 noninferiority difference; the SIQ sum used ten points, and the HSI used three points. Those choices determine how far apart the treatments can be and still support noninferiority. The study did not ask one all-purpose question that can be answered by choosing whichever secondary result produces the smallest p value.

Clinical noninferiority also depends on assumptions about the active comparator’s established effects and an appropriate margin. The authors noted that the selected two-point ESS threshold might have been stringent; that possibility is a design limitation, not permission to retrospectively change the threshold until significance appears. For the primary planned analysis, the result remains what was reported: an unsuccessful demonstration of noninferiority, with more and larger trials needed to resolve treatment differences.

How often did participants report adverse events?

Seventeen of twenty-two participants in each randomized arm experienced at least one treatment-emergent adverse event, 77.3% per group (p = 1.0). However, seven of twenty-two participants assigned modafinil discontinued because of an adverse event, versus two of twenty-two assigned amphetamine-dextroamphetamine. This was a numerically large contrast, 31.8% versus 9.1%, but the original between-group test was p = 0.13. It is not proof of a true difference in discontinuation risk.

The source additionally reported two severe events and one serious event in the modafinil arm, with zero in the other arm, but a study with 22 people per condition cannot reliably estimate rare-event risk. Adverse events were clinical reports, not evidence that subjects reached a particular circulating drug concentration, had misused medication, or suffered a forensic injury attributable to one measured molecule.

What were the source-accurate adverse-event comparisons?

Appetite reduction was more frequently reported on amphetamine-dextroamphetamine, while anxiety symptoms appeared more frequently on modafinil in this small sample. These exploratory event-by-event tests should not be taken as definitive long-term comparative risks, especially given multiple comparisons and modest sample sizes. More importantly for forensic interpretation, an observed adverse-event frequency is not a laboratory cutoff, a measured individual’s blood level, or a proof that an adverse event occurred because of a particular medication in a particular case.

Did the trial measure objective alertness or operational performance?

No standardized Maintenance of Wakefulness Test result or directly administered on-road driving comparison formed the primary outcome. The investigators explicitly described their preference for patient-reported endpoints and the consequent limitations. Respondents assessed how they felt and functioned across weeks; these instruments can support clinical treatment evaluation without quantifying sudden vigilance lapses during a legally relevant event.

A person can report meaningful relief of sleepiness while still experiencing residual symptoms, and different tasks can challenge distinct cognitive domains. Indeed, only about half of the participants in the amphetamine group and roughly two-thirds of the modafinil group described much or very much improvement in sleepiness. Consequently, the study neither shows universal wakefulness restoration nor permits an individual’s clinical ESS change to stand in for a driving fitness examination.

Was human whole blood or plasma analyzed for either medication?

No validated quantitative parent-drug study matrix was reported for the randomized outcome analyses. The source did not present LC–MS/MS or GC–MS results for parent modafinil, amphetamine, or their metabolites in legal human venous whole blood, separated serum, plasma, oral fluid, urine, hair, vitreous humor, sweat, or exhaled breath. There were no measured ng/mL drug concentration curves tied to the questionnaire observations, and no quantitative pharmacokinetic dose-to-concentration model was validated by the study.

This limitation is categorical, not a p-value failing to reach 0.05. A prescribed 20 mg capsule is an administered mass; “34 mg/day” is a mean regimen; neither equals a plasma measurement. Even if plasma had been measured, it would be necessary to distinguish it from whole blood, which has different specimen composition and cannot simply be relabeled by convention. See the published biological-matrices guide for why specimen identification precedes any inference about concentrations.

Can questionnaires distinguish therapeutic use from illicit stimulant exposure?

Not by themselves. The trial used supervised prescription assignment and clinical symptom tracking; its questionnaires were not chemical identification tests. Forensic confirmation of a substance, when clinically or legally indicated, requires an actual collected specimen, validated method, chain-of-custody practices where applicable, and analytical data identifying relevant parent drugs or metabolites. This clinical trial supplies none of those person-specific forensic results.

In particular, mixed amphetamine-dextroamphetamine is an amphetamine-containing prescription preparation; it must not be silently renamed methamphetamine. A symptom improvement after an amphetamine prescription does not establish the source, enantiomeric profile, concentration, dose timing, or legal significance of an unrelated patient’s positive screen. The laboratory distinction between presumptive screens and specific analyte confirmation is addressed in the published confirmatory drug-testing guide.

What does adverse-event discontinuation say about blood concentrations?

The higher numerical discontinuation frequency in the modafinil arm may be clinically important and worth follow-up. But a symptom such as anxiety, a report of reduced appetite, or a participant’s decision to discontinue does not yield an analytically meaningful parent-drug blood level. The trial did not collect parallel drug concentrations to characterize the variability in absorption, metabolism, timing, or tissue distribution among people who experienced these events.

Nor can the average maximum doses be used to interpolate a single subject’s serum or whole-blood exposure. Such an interpolation would require a validated pharmacokinetic framework and carefully recorded sampling conditions. For further context on what blood-drug measurements can legitimately indicate and what they cannot, see the published therapeutic-drug-monitoring guide.

What features limit clinical and forensic generalization?

The trial had 44 treated participants in total, with relatively few participants diagnosed with narcolepsy type 2. Drug doses were individualized, rather than one uniform exposure. Some outcome entries had missing measurements, and the primary modified intention-to-treat model used last observation carried forward. The trial included no inert-placebo control, no objective wakefulness test, no drug-level concentration measurements, and no road or simulated driving performance experiment. These are characteristics of the design, not evidence that the participants were impaired or unimpaired.

The investigators’ findings have legitimate clinical relevance: both regimens were associated with reported symptom improvements, and some secondary endpoints satisfied noninferiority criteria. But scientific precision requires respecting the failure of the primary ESS criterion, the event-level and person-level sample sizes, the limits of the subjective scales, and the distinction between therapeutic drug administration and forensic measurement. Those boundaries are particularly important when research is cited in litigation or individual medical decisions.

Could these findings establish intoxication, safe driving, or the cause of an incident?

No. The trial was not designed to ascertain an individual driver’s actual drug concentration at a given time, determine whether a dose had been taken before an incident, measure field sobriety, classify legal impairment, or calculate the probability that a medication caused a crash or other injury. The study’s results are group comparisons of treatment strategies. A single individual’s clinical disease, medication response, co-medications, sleep debt, and relevant event circumstances would need separate evidence; the original article does not supply them.

Equally, a nonsignificant superiority p-value, a positive noninferiority result for sleep inertia, and a failure of primary ESS noninferiority are all different statistical statements. None is a validated affirmative finding of unimpaired driving. The appropriate forensic use of this paper is to describe its tested therapeutic endpoints accurately, state the evidentiary limits, and avoid substituting a self-report measure for an unperformed drug analysis or incident reconstruction.

Frequently asked questions about this study

1. Did the trial prove that mixed amphetamine salts and modafinil are equally effective?

No. Its primary noninferiority comparison on Epworth improvement did not succeed (p = .11), and the ordinary mean-difference test was nonsignificant (p = .64). Failure to show a difference is not proof of equivalence. The study did support noninferiority on several secondary symptom outcomes, but that is a narrower conclusion.

2. Does a noninferiority p-value below .05 mean the amphetamine was superior?

No. The alternative hypothesis in a noninferiority test concerns ruling out a clinically unacceptable disadvantage. It does not test whether the treatment has a superior effect. The strong result for the “good cognitive dysfunction” response category cannot be described as proof that amphetamine-dextroamphetamine outperformed modafinil on objectively tested cognition.

3. Was this a trial of methamphetamine or lisdexamfetamine?

Neither. The trial directly compared immediate-release amphetamine-dextroamphetamine with modafinil. Methamphetamine is a different chemical entity, and lisdexamfetamine is a prodrug with different pharmacokinetics. The study cannot be cited as their controlled head-to-head test.

4. How many patients actually contributed to the outcomes?

Forty-four participants took at least one randomized medication dose: 22 per group. Some patient global impression responses used a denominator of 20 in the amphetamine arm. The trial’s four assessment visits and multiple questionnaires are not additional independent patients; dropped-out participants could contribute earlier scores through last observation carried forward.

5. Was a blood amphetamine concentration measured after dosing?

No quantitative parent-drug whole-blood, plasma, or serum concentration series was reported. The prescribed doses and average final daily regimens were administration details, not specimens or measured ng/mL. The study cannot establish a forensic concentration cutoff for either medication.

6. Did modafinil cause more treatment discontinuation?

Discontinuation attributed to adverse events was observed in seven of 22 modafinil participants versus two of 22 amphetamine-dextroamphetamine participants. The original between-group test returned p = .13, so a reliable population-level difference was not established. This small trial cannot prove individual causation or settle rare safety questions.

7. Can the results determine whether a medicated person could safely operate a vehicle?

No. Neither actual on-road performance nor a validated driving simulator was the study’s endpoint; the principal instruments were subjective symptoms. The findings cannot substitute for person-specific clinical assessment, any relevant contemporaneous evidence, or legally appropriate drug testing in a particular situation.

Bottom line for careful scientific interpretation

Forty-four randomized adults reported improvement with both regimens, but amphetamine-dextroamphetamine did not meet the planned Epworth noninferiority criterion. Some secondary disease-symptom and patient-impression measures did meet noninferiority criteria, with important differences in missing-data denominators and variation among individuals. Adverse events were common in both groups, and specific reactions differed descriptively. The study did not directly quantify blood concentrations of either administered medication or validate individual driving capability, intoxication, or the cause of any incident.

Primary research: Trotti LM, Blake T, Hoque R, Rye DB, Sharma S, Bliwise DL. Modafinil Versus Amphetamine-Dextroamphetamine For Idiopathic Hypersomnia and Narcolepsy Type 2: A Randomized, Blinded, Non-inferiority Trial. CNS Drugs. 2024;38:909–920. doi:10.1007/s40263-024-01122-y. Interpret all estimates within the original trial design and measured endpoints.