Does Lisdexamfetamine Improve Sleep or Cause Insomnia in Adult ADHD? What a 420-Person Trial Found

Answer first: In a four-week randomized, double-blind adult ADHD trial, lisdexamfetamine (LDX) improved one self-reported daytime-functioning component of the Pittsburgh Sleep Quality Index (PSQI), but it did not significantly improve the overall PSQI sleep-quality score relative to placebo. At the same time, participants assigned LDX reported insomnia more often as a treatment-emergent adverse event: 69 of 358 (19.3%) compared with 3 of 62 (4.8%) assigned placebo. Both findings can be true because an adverse-event report of insomnia and a questionnaire summarizing the preceding month are different measurements. Neither outcome establishes that every treated person sleeps better, that reported insomnia is clinically unimportant, or that medication makes an individual fit to drive.

The primary report is Adler and colleagues’ 2009 analysis, Effect of lisdexamfetamine dimesylate on sleep in adults with attention-deficit/hyperactivity disorder, Behavioral and Brain Functions 5:34, DOI 10.1186/1744-9081-5-34. Its direct observations concern a short randomized clinical treatment period, subjective sleep ratings, symptom measures and reported adverse events. They are not drug concentration measurements, a laboratory-based sleep study, or forensic observations of an accident.

What is the central scientific distinction?

The question “Does this medication cause insomnia?” is not identical to “Did the average global sleep-quality score worsen?” Insomnia recorded as an adverse event describes a new problem reported during treatment, coded using the trial’s adverse-event system. The PSQI, by contrast, asks about multiple aspects of sleep during the preceding month and combines seven component scores into one global number. Changes in one component can offset changes in others; several individuals can report meaningful difficulties even when the group mean remains similar. Therefore, the correct reading of this study must maintain two independent tracks: sleep-related event frequencies and adjusted between-group questionnaire comparisons.

A further distinction is between the subjective daytime-dysfunction component of the PSQI and measured psychomotor vigilance. The component asks about trouble remaining awake during activities and maintaining enthusiasm to complete tasks. It was not a sustained-attention reaction-time test, a driving simulator, or an objective measure of alertness at the moment of a specific event. A reduced questionnaire score can support a group-level account of perceived daytime function without establishing an individual performance threshold.

What was the study design?

Researchers analyzed data from a four-week, phase 3, parallel-group, randomized, placebo-controlled trial of adults aged 18–55 with clinically diagnosed ADHD. The treatment was double-blinded and used a forced-dose escalation scheme. Random assignment placed 62 adults on placebo, 119 assigned to an eventual 30-mg daily LDX dose, 117 to 50 mg, and 122 to 70 mg. Together, the active-dose arms comprised 358 participants. Study treatment was taken orally in the morning, not intravenously, intranasally or as methamphetamine.

All active-group participants initially received 30 mg/day; participants assigned higher target doses escalated during the first one or two weeks. Consequently, a nominal target of 50 or 70 mg is not an accurate description of every tablet taken during the entire trial. The design compared randomized treatment strategies, not a single standardized pharmacokinetic exposure. It also did not randomize participants to an objective sleep intervention or compare LDX against another active ADHD medication.

The paper is a focused sleep analysis of an existing ADHD efficacy trial, not a separate new cohort of 420 people. A later or earlier paper derived from the same clinical trial can ask another legitimate research question, but independent reports must not be counted as independent recruited populations when considering the amount of clinical evidence.

How many independent adults contributed information?

The safety population was the 420 randomized and exposed adults shown in the original disposition figure. The parent study’s ADHD symptom efficacy population comprised 414 people meeting its applicable data requirements. For the PSQI sleep analyses, valid endpoint questionnaire information was available for 402 participants. These are related but different denominators. The same adult may supply a baseline score, a week-four score, multiple component scores and more than one event report; those repeated observations cannot be counted as additional independent humans.

The study reports 349 completers and 71 early discontinuations overall. Although an endpoint PSQI observation was available for most randomized subjects, missing questionnaire data and early withdrawal still matter. The analysis used least-squares adjusted mean changes for sleep scores, and event percentages used each assigned treatment group’s safety denominator. These are not interchangeable denominators, and the article should not silently combine them into a single fictitious “420 complete sleep measurements” statement.

What did the PSQI actually measure?

The self-rated Pittsburgh Sleep Quality Index includes questions addressing subjective quality, time until sleep begins, reported duration, habitual sleep efficiency, sleep interruptions, sleep medication use and daytime dysfunction. Each of those seven component scores ranges from 0 (less difficulty) to 3 (more difficulty), and the global score sums the components across a 0–21 scale. Higher global scores indicate worse reported sleep. The analysis administered the PSQI at baseline and after treatment or earlier discontinuation, rather than continuously measuring nightly sleep architecture.

Baseline mean global PSQI scores were 6.3 for placebo, 5.4 for the 30-mg arm, 5.9 for the 50-mg arm and 6.0 for the 70-mg arm. Thus sleep problems were present before the study treatment. The investigators noted that global scores above five commonly indicate poor self-reported sleep. A high initial score matters because a slight later reduction is not evidence that the study medicine abolished existing insomnia or normalized objective sleep.

Evidence table 1: which sleep outcomes were statistically different?

Evidence table notes: Values are drawn from the study’s Results, Tables 1–2 and accompanying figures. Negative global PSQI and daytime-component changes indicate improvement; an increase in sleep-onset latency does not. Standard errors describe adjusted change estimates, not person-specific confidence intervals. The last two rows are reported nonsignificant without an invented exact p value. These measures were questionnaires, not quantitative stimulant assays or objective polysomnography.

Did lisdexamfetamine improve overall sleep quality?

The original answer is no significant improvement relative to placebo was established. The adjusted mean global PSQI change was −0.8 across active groups compared with −0.5 for placebo, producing a reported between-treatment p value of .33. A small numeric difference in the favored direction is neither a proven group-level advantage nor proof that two treatments are identical. A nonsignificant test means the data did not establish a difference at the chosen conventional threshold under that analysis.

The individual active-dose comparisons on the global score were also reported as nonsignificant. The six other PSQI components besides daytime dysfunction—subjective sleep quality, onset latency, duration, efficiency, interruptions and sleep-medication use—were not significantly different from placebo in the study’s component analyses. That pattern is important because a positive isolated component is not a positive global sleep study.

What improved in daytime functioning?

The study’s clearest adjusted PSQI result was in the component labeled “daytime dysfunction.” Across active treatment arms, this component changed by −0.4 points versus 0.0 in placebo, with p = .0001. Each of the three assigned dose groups also met a reported p ≤ .01 against placebo for this component. This is consistent with adults reporting less daytime trouble staying awake or generating enough enthusiasm to complete usual activities after receiving LDX in this short protocol.

It would be inaccurate to substitute “better objective vigilance” for “lower PSQI daytime dysfunction.” The two questionnaire items do not assess every aspect of alertness and were not designed to adjudicate whether a driver responds quickly to a hazard. They do not tell us whether all participants improved, whether a subgroup became risk-free or whether the change was large enough for an individual functional decision. The authors themselves questioned the clinical significance of a component improvement when global PSQI scores and other component scores did not similarly show improvement.

Why were insomnia adverse events more frequent?

Participants and investigators reported events arising during the four-week treatment period using the trial adverse-event procedure. With LDX, spontaneous reports of insomnia were more frequent than with placebo, and some participants stopped study medication because of sleep-related complaints. The report describes sleep-related events as generally mild or moderate, but its severity table also includes severe insomnia in eight active participants. These facts must not be hidden beneath the nonsignificant global PSQI outcome.

It is also important to distinguish reported incidence from proven mechanism. Randomization strengthens an interpretation of treatment-group differences, but an individual event label does not establish its complete causal pathway, nightly duration, electroencephalographic correlate or precise onset relative to a measured drug concentration. The paper gives descriptive frequencies for adverse events, not an inferential p value for every event type. A reader should not invent a significance test or adverse-event hazard ratio that the authors did not publish.

Evidence table 2: sleep-related treatment-emergent events

Evidence table notes: The first five rows reproduce source-defined treatment-emergent event categories from the original Table 3; the discontinuation row uses original Table 5. Categories can occur in the same person and may not be summed to obtain a number of unique affected adults. These figures are not blood concentrations, nights with insomnia or measured sleep-stage minutes. No unreported hypothesis-test p values are assigned to these descriptive proportions.

Did reported insomnia diminish during the four weeks?

The researchers recorded 50 active-treatment participants (about 14% of the combined LDX safety population) with insomnia events in week one, compared with nine in week two, ten in week three and seven in week four or later. This temporal distribution was one reason the authors described reported insomnia as decreasing after the first treatment week. But it should not be transformed into a controlled claim that every insomnia problem resolved by day eight or that risks vanish during long-term maintenance. Some events may continue for more than a single week, and forced-dose escalation changes exposure across the same period.

The overall insomnia category included several people with more severe events, and nine LDX-assigned participants withdrew due to sleep-related adverse events (with ten such event reports across those nine people). The article documents eight discontinuations attributed to insomnia, one to a sleep disorder, and one to hypersomnia; the event categories are not all different people. Any assessment of tolerability must describe both the group averages and the people for whom the adverse effect was consequential enough to stop participation.

Why do adverse-event reporting and PSQI scores diverge?

Several methodological differences explain why the two measurements should not be expected to coincide. First, the adverse-event system asked about new symptoms during visits, while the PSQI summarized perceived sleep over roughly a month. Second, a single sleep problem may meaningfully trouble a participant yet contribute only part of a composite index. Third, the global score incorporates daytime functioning; an improvement there can counterbalance a deterioration in another domain at the aggregate level. Fourth, some reports may reflect transient onset while the questionnaire integrates many nights.

The authors conducted an additional descriptive examination of PSQI scores among participants with sleep-related adverse events. Their group scores did not appear appreciably worse at endpoint despite those reported events. That exploratory comparison cannot prove individual insomnia was imaginary or clinically irrelevant. Participants entered the trial with different sleep histories, the PSQI is subjective, and the analyses did not replace a prospective objective sleep laboratory assessment. The most accurate synthesis preserves apparent tension instead of pretending the endpoints answer the same question.

Was sleep objectively measured?

No objective sleep architecture endpoint was reported for this analysis. There was no polysomnography measuring respiratory events, oxygen desaturation, EEG-defined sleep stages or arousal index, and no study-reported actigraphy-derived nightly sleep estimates. The PSQI sleep-onset and duration values arose from participant recollection. Investigators also collected general clinical safety observations, but the present publication did not present objective physiological sleep measures capable of confirming whether a specific patient’s sleep had improved or deteriorated overnight.

This matters especially when someone wants to extrapolate to untreated obstructive sleep apnea, circadian disorders, shift work, occupational fatigue or a particular missed driving hazard. These were not experimentally tested causal endpoints. Reports of greater enthusiasm or reduced perceived daytime difficulty should not be converted to evidence about apnea severity, legal alertness standards or individual workplace safety. A separate protocol with objective overnight measurements and validated performance tasks would be needed for those specific questions.

What did the investigators actually measure in biological specimens?

The relevant exposure was the randomized oral LDX dosing strategy: 30, 50 or 70 mg/day target, with titration. The sleep paper did not quantify the concentration of lisdexamfetamine or its active metabolite dextroamphetamine in collected legal venous whole blood. It did not provide a parent-drug or metabolite concentration in separated plasma or serum, a chromatographic blood assay, a validated reporting limit, sample time relative to the dose, or a concentration-effect model.

It also did not chemically quantify stimulant exposure in urine, oral fluid, sweat, hair, vitreous humor or exhaled breath. General clinical laboratory safety assessments are not equivalent to quantitative forensic toxicology. LDX is a prodrug that yields d-amphetamine; it should not be mislabeled methamphetamine. A prescription dose of 30–70 mg is not a concentration in ng/mL, and no defensible formula can convert this randomized assignment into a person-specific venous whole-blood value.

For the general differences between matrices, the site’s independently published biological matrices guide explains why whole blood, plasma, serum and other specimens must be interpreted separately. Its confirmatory testing guide describes the analytical requirements that this sleep questionnaire study did not attempt to satisfy. Those resources offer methodological context rather than additional data from this randomized clinical trial.

Can these findings support a blood-concentration or impairment opinion?

No. A study about self-rated sleep during randomized therapeutic treatment can help frame what was measured in a population, not tell an examiner which drug amount was present in the blood of a particular person at an incident. No original specimen-specific analytical calibration, dosing adherence confirmation, individual concentration-time measurement, field-sobriety examination, reaction-time test battery or on-road performance endpoint links these questionnaire outcomes to a legally contested episode.

A clinical finding may nevertheless help identify questions for a forensic investigation: Was the medication prescribed and taken as directed? Was a blood specimen actually collected, from which compartment and when? Did a scientifically validated laboratory determine drug identity and concentration? Was there independent evidence of actual sleep loss, another sedative or stimulant, fatigue, medical illness or performance deficit? Even when those records exist, the clinical trial alone supplies neither a chemical impairment cutoff nor the causal explanation for a crash, injury or another individual’s behavior.

What limitations affect the clinical inference?

The follow-up was short—four weeks—and the dose-escalation plan was forced, so target doses were not necessarily tailored to an optimal efficacy–tolerability balance for each individual. Participants with significant psychiatric comorbidity, several medical risks and substance-use histories were excluded. People with serious baseline sleep problems might have been less likely to volunteer even though insomnia was not formally an exclusion criterion. Those features constrain generalization to more medically complex patients.

Subjective instruments can be influenced by recall, expectations and symptom fluctuations. The PSQI was a secondary sleep analysis of a trial primarily organized around ADHD symptoms, not an experiment powered to settle every sleep disorder diagnosis or severe adverse effect. Several author disclosures identify research support, honoraria or other relationships with manufacturers; such interests should be recorded as part of critical appraisal without replacing scrutiny of the actual numbers. The outcome of an exploratory component analysis should not be elevated over the nonsignificant global measure merely because its p value is small.

How does this study differ from other lisdexamfetamine trials?

Other publications may use the same parent randomized ADHD trial or study a different sample, dose schedule, clinical domain or pharmacokinetic question. The correct unit of independent clinical replication is not the number of publications. This 2009 report’s distinctive contribution is its PSQI global versus daytime-dysfunction contrast alongside reported insomnia events in the four-week 420-person randomized program. It does not independently repeat another clinical comparison simply because its author list or DOI differs from a post hoc efficacy paper.

Forensic readers should also distinguish this clinical questionnaire question from analytical methods that quantify dextroamphetamine, drug-effect experiments in healthy volunteers, pediatric ADHD efficacy trials, or simulator studies that do not measure actual on-road injuries. Sharing a medication name does not make such experiments interchangeable. An answer-first source analysis should follow the actual study outcome rather than recycle the same general stimulant advice under another headline.

What is the defensible bottom line?

In these 420 randomized adults, LDX treatment over four weeks was associated with a significant improvement in self-reported PSQI daytime dysfunction, while the adjusted global PSQI improvement relative to placebo was nonsignificant. Meanwhile, insomnia was reported more often under active treatment than placebo, and some individuals withdrew because of sleep-related problems. Consequently, saying either “LDX improves sleep” or “LDX invariably ruins sleep” loses important distinctions present in the original data.

The scientific conclusion is narrower and stronger: this publication documents a difference in perceived daytime function, no proven group-level advantage in overall PSQI sleep quality, and a notable imbalance in insomnia adverse-event frequencies during short-term clinical treatment. It does not supply an individual’s blood concentration, objectively measured sleep physiology, a legal impairment threshold, or a causal explanation for a specific accident or witness’s actions.

Frequently asked questions

1. Did all 420 adults complete the sleep questionnaire at the endpoint?

No. The trial’s safety population contained 420 randomized and exposed adults, but 402 had valid endpoint PSQI assessments. A separate 414-person population was used for the parent ADHD-symptom efficacy analysis. Those denominators answer different questions and should not be blended.

2. Was the global PSQI improvement statistically significant?

No. The active-group adjusted global change was −0.8 versus −0.5 for placebo, with p = .33. That nonsignificant result neither establishes treatment superiority nor proves equivalence between active medicine and placebo.

3. Did people taking lisdexamfetamine report more insomnia?

Yes. Insomnia was recorded in 69/358 active participants (19.3%) and 3/62 placebo participants (4.8%). These descriptive treatment-emergent event frequencies coexist with the nonsignificant global PSQI result. The original report did not provide a separate exact hypothesis-test p value for this event row.

4. What was significantly improved?

The PSQI daytime-dysfunction component showed an adjusted change of −0.4 under combined LDX compared with 0.0 under placebo, p = .0001. It is a self-report component, not a direct driving, attention-reaction, or sleep-apnea measurement.

5. Did the investigators verify better sleep with a laboratory sleep study?

No. The analyzed sleep outcomes came from the participant-completed PSQI and clinical adverse-event reports. Polysomnographic stages, objective apnea indices, actigraphy-based nightly sleep and measured reaction-time vigilance were not reported endpoints of this sleep paper.

6. Did the trial measure dextroamphetamine concentrations in forensic whole blood?

No. Participants were assigned oral LDX doses, but the sleep analysis did not report quantitative LDX or d-amphetamine testing of legal venous whole blood, plasma or serum. Dose assignment cannot substitute for a time-stamped analytical toxicology result.

7. Does an improved daytime score establish that a particular driver was unimpaired?

No. Individual impairment requires event-specific clinical, behavioral and—when relevant—validated analytical evidence. A four-week self-rated population average does not validate a blood concentration cutoff, last-dose inference, crash reconstruction, witness reliability determination or personal driving-safety conclusion.

Original research citation and interpretation boundary

Adler LA, Goodman D, Weisler R, Hamdani M, Roth T. Effect of lisdexamfetamine dimesylate on sleep in adults with attention-deficit/hyperactivity disorder. Behavioral and Brain Functions. 2009;5:34. Read the complete original primary article and its methods and tables for the full design, adverse-event definitions and exact reported contrasts. This analysis is an independent educational interpretation of the published research; it is not an assessment of any named individual or disputed incident.