Does Modafinil Reduce Persistent Fatigue After Stroke? What a 36-Person MIDAS Crossover Trial Proved—and Did Not Prove

Short answer: In a small randomized, double-blind crossover trial of 36 adults with persistent fatigue at least three months after stroke, six weeks of oral modafinil 200 mg daily was associated with lower self-reported fatigue and higher stroke-specific quality-of-life scores than placebo. The reported overall Multidimensional Fatigue Inventory treatment estimate was −7.38 points, and the Stroke-Specific Quality of Life estimate was +11.81 points. But the overall Montreal Cognitive Assessment and measures of depression, anxiety and stress did not improve significantly. The trial did not analyze modafinil concentrations in a physical specimen, directly test driving, or prove that any particular patient was unimpaired.

The underlying original experimental paper is Bivard and colleagues’ MIDAS (Modafinil in Debilitating Fatigue After Stroke): A Randomized, Double-Blind, Placebo-Controlled, Cross-Over Trial, published in Stroke in 2017. The primary article is accessible at doi:10.1161/STROKEAHA.116.016293. This analysis keeps the research’s actual comparator, population, measured endpoints, statistically nonsignificant findings and evidentiary limits separate rather than treating a favorable fatigue score as proof of objective alertness or safe performance.

Why does this particular trial matter?

Fatigue after stroke can affect participation in therapy, day-to-day activity and perceived well-being. A claim that a stimulant-like wakefulness medicine helps fatigue is narrower than a claim that it repairs all cognitive deficits following stroke. The MIDAS investigators deliberately selected a group with established, substantial, lingering fatigue instead of recruiting everyone during hospitalization. Their question was whether a modest daily dose of modafinil improved patient-reported fatigue relative to an indistinguishable placebo over a defined six-week course.

That design is informative for interpreting clinical benefit and its limits. It is less informative about an individual’s concentration of drug, instantaneous psychomotor capacity, cardiovascular response at a critical time or the cause of a particular accident. In a medical record review or litigation, evidence that a group rated fatigue as better at the end of six weeks should not be silently substituted for measured vigilance, legal impairment or a proven treatment response in an untested person.

Who actually participated, and how many independent people were studied?

The researchers screened 232 potentially eligible people with a history of stroke or transient ischemic attack. Thirty-eight met the criteria and consented, and two were excluded before trial medication began, leaving 36 randomized participants. The final 36 were a single biological sample: 22 men and 14 women, mean age approximately 63 years. They were about nine months after stroke on average, with the reported interval ranging from three to 38 months. Thirty-three had ischemic stroke histories and three hemorrhagic strokes.

Entry required an MFI score of at least 60 and a stroke history at least three months old. The baseline study-group mean MFI score was approximately 72. This is not a community screening sample of all stroke survivors, and it is not a study of naturally occurring sleepiness in otherwise healthy adults. It concentrates on people with considerable persistent symptoms who met a restrictive clinical profile.

Each randomized participant received both active drug and placebo in different periods. A six-week active course and a six-week control course do not create 72 independent persons. Nor do scores from several questionnaires multiply the number of patients. Repeated observations belong to the same 36 individuals and require a within-person interpretation.

Who was excluded, particularly regarding other causes of fatigue?

The eligibility criteria attempted to reduce alternative recognized causes of fatigue. Investigators excluded people with diagnosed or clinically suspected sleep apnea, narcolepsy, substantial renal impairment, dementia, recognized depression or other specified neuropsychiatric illness, and people using certain medicines, including benzodiazepines and antiepileptic drugs. A clinician could exclude someone for symptoms suggestive of sleep apnea even without an established diagnosis.

These choices matter because post-stroke fatigue is not a single biological mechanism. Findings in patients selected to have none of those recognized confounders cannot simply be generalized to patients with sleep apnea, significant depression, complex sedating polypharmacy or the acute hospitalization phase. Importantly, although suspicious sleep symptoms prompted exclusion, the researchers did not obtain specific formal sleepiness measurements at baseline and after each intervention, so unidentified sleep-disordered breathing or another contributor could not be completely ruled out.

How were the drug, placebo and crossover periods administered?

This was a single-center, phase-two randomized experiment conducted at John Hunter Hospital in Australia. Eighteen participants were assigned initially to modafinil and 18 initially to placebo. The active treatment was 200 mg of oral modafinil each morning for six weeks; the placebo tablet was manufactured to look the same and contained rice powder. After the first treatment period came a one-week washout, followed by six weeks on the alternative assignment. Outcome assessments took place around baseline, treatment-period end, washout and second-period end.

Randomization and dispensing were handled independently from the assessors, with the clinical trials pharmacist aware of allocation. This reduces expectation and allocation-related bias compared with an open-label series. It does not make the patients representative of all stroke survivors, nor does it remove the usual possibility of period effects and carryover in a crossover design. Investigators used within-participant, baseline-adjusted regression and conducted a treatment-sequence sensitivity check.

Returned study-pill bottles served as an adherence check. Tablet returns can show what was left in the container but cannot prove every unreturned dose was swallowed. There was no pharmacokinetic blood sample presented as an independent adherence assay.

What exactly was the primary endpoint?

The primary endpoint was the 20-item Multidimensional Fatigue Inventory (MFI), a self-report instrument representing general fatigue, physical fatigue, mental fatigue, activity and motivation. Its aggregate score ranges from 20 to 100, with greater values reflecting worse fatigue. It asks people to characterize their own recent experience; it does not time a reaction to an unexpected vehicle, assay a chemical, establish sleep duration or certify a person’s capacity to operate equipment.

The prespecified hypothesis concerned a mean treatment difference in MFI after modafinil compared with placebo. The reported baseline-adjusted overall estimate was a 7.38-point reduction on modafinil. The authors reported this overall difference as statistically significant at the study’s primary threshold. A second self-reported instrument, the Fatigue Severity Scale (FSS), also favored modafinil, with an estimated mean difference of −6.31 points and reported p = .0048.

These results provide a signal of improvement in the symptoms the subjects themselves rated. The distinction between perceived fatigue and objective cognitive speed matters: a person may feel less fatigued without showing a statistically discernible change in a global cognition test. Conversely, a test can detect selective cognitive changes without proving an improvement in fatigue or the person’s real-world function.

What were the central significant and nonsignificant numerical findings?

The following table rearranges selected primary-trial results by their evidentiary question. A negative fatigue estimate generally favors modafinil because lower fatigue scores are better; a positive quality-of-life estimate favors modafinil because higher quality-of-life scores are better. The values are treatment contrasts, not measured modafinil concentrations, and the same 36 participants contributed crossover observations.

Statistical significance describes a test under its assumptions; it does not by itself specify a minimum clinically important improvement for every individual. With just 36 randomized people, outcome estimates may be sensitive to baseline adjustment and other modeling decisions. The publication’s reporting of its principal MFI confidence interval and several domain scores is internally difficult to reconcile; the private source audit records that issue. It is inappropriate to manufacture a corrected interval or turn the reported point estimate into a precise individual prognosis.

Did the benefit persist when participants changed treatment periods?

The researchers reported that the initial placebo group had a mean MFI near 58 at the first treatment endpoint, while the initial modafinil group had a mean near 50. Among the participants who received modafinil first, the mean fatigue rating rose toward 59 after washout. At the second period endpoint, those taking modafinil after placebo were reported near 48, compared with approximately 59 in those taking placebo after modafinil.

These period-specific descriptions are directionally consistent with a treatment-associated fatigue improvement that attenuates when the course ends. However, the figures are not a blinded rechallenge of every possible causal explanation. Participants experienced an extended trial, repeated assessments and their own rehabilitation trajectory. In a formal carryover check, the treatment-by-sequence term was not significant (p = .299). That result did not detect a meaningful interaction; it does not prove carryover impossible or guarantee that washout was perfect for every patient.

How much did quality of life improve, and which domains were unaffected?

The total Stroke-Specific Quality of Life (SSQoL) score rose by an estimated 11.81 points with modafinil. Several domains showed nominally significant positive contrasts: energy, mobility, social roles, vision and thinking. Others did not, despite some numerical trends. A favorable total score therefore should not be described as restored self-care, improved language, improved family function or a return to work unless those endpoints are actually supported.

Quality-of-life questionnaires are valuable for understanding patient experience, but their dimensions are not direct proof of any one functional capacity. The ability to self-report feeling better about mobility differs from passing a standardized gait examination; the perception of clearer thinking differs from a validated general cognitive advantage. The study’s domain analyses were multiple secondary comparisons, which require particular care before an isolated nominal p-value is treated as robust confirmation.

Numerical estimates in this table are points on their own respective scales, not standardized effect sizes, milligrams or laboratory concentrations. Comparing +2.07 mobility points directly with +0.20 cognitive subscore points as if they shared units would be meaningless. Subdomain results should be read with the broader nonsignificant total cognitive score and the many tests performed.

Why was overall cognition not significantly better?

Global cognition was assessed with the Montreal Cognitive Assessment (MoCA), a brief composite of memory, visuospatial ability, language, attention, executive processing and related functions. In the reported adjusted model, the overall treatment difference was +0.32 point, with p = .365. That does not establish a global cognitive improvement. The study nevertheless identified nominal positives for the abstraction and visuospatial subscores.

A few positive components within a nonsignificant total are not logically impossible. Components can respond differently, tests have different variance and precision, and many comparisons create opportunities for findings near conventional thresholds. The supported conclusion is more discriminating than either “modafinil restored cognition” or “nothing changed”: the primary symptom result favored treatment, global MoCA did not, and two exploratory subdomains were reported as better.

The report did not administer a detailed forensic psychomotor impairment battery, measure lane position while driving, score collision-avoidance response under hazard, or establish an age-specific medication-related reaction-time threshold. Its MoCA results cannot resolve those unmeasured matters, whether the cognitive total was statistically significant or not.

Were depression, anxiety or stress significantly changed?

The investigators used the Depression, Anxiety and Stress Scale (DASS) to examine mood-related complaints. The reported total DASS contrast was −0.76 with p = .7688. Anxiety was reported as +0.87 with p = .2916; depression −0.06 with p = .9608; and stress −1.43 with p = .0684. None reached the conventional threshold in the original table.

This matters because fatigue, depression, apathy and stress may coexist but are not identical endpoints. One cannot infer that a drug improved depression because a fatigue measure became better. Nor can one infer a mood disorder was caused or excluded by the drug because a small trial’s group-level DASS treatment differences were statistically nonsignificant. Those questions require appropriate clinical assessment and, where relevant, individual longitudinal evidence.

What did the researchers report about adverse events?

Across the approximately 13-week study, investigators reported 12 adverse events: five assigned to the modafinil period and seven to placebo. Listed events included four headaches, one nausea episode, two anxiety events, three agitation events and two dizziness events. No serious adverse events were reported during this small trial. These are counts of reported events, not necessarily 12 distinct participants and not exposure-adjusted rates suitable for a large-population risk calculation.

It would be an error to claim that modafinil cannot cause serious adverse reactions because this trial recorded none among only 36 participants. Its design was intended to explore symptom efficacy, not to exclude rare adverse effects or establish long-term cardiovascular, neuropsychiatric or dependence safety. Similarly, a numerically smaller count of events in one arm cannot establish an adverse-event protective effect without a suitable analysis and a much larger safety dataset.

Was modafinil measured in whole blood, serum or plasma?

No measured modafinil concentration appears in the reported trial. The intervention was verified operationally as a tablet assignment and pill-return count, with clinical outcome questionnaires. The investigators did not report an LC–MS/MS, GC–MS or other drug-specific quantified modafinil result from venous whole blood, separated serum, plasma, urine, oral fluid, hair or breath. They did not present a drug-concentration-time curve, a concentration–effect regression or an individual exposure cut point.

A dose of 200 mg is the amount administered in a tablet, not the concentration circulating in a patient’s blood at a later time. A physical measurement in plasma, even if it existed, could not be relabeled as a whole-blood result without an appropriate conversion and supporting validation. This paper supplies neither measure. The site’s published guide to forensic biological matrices explains why specimen distinctions are not merely terminological.

Confirmatory chemistry answers a different question from patient-reported fatigue. If a case requires confirmation that modafinil was present at a particular level, the laboratory’s actual specimen type, sampling time, validated target analyte and method become essential; see the separately published confirmatory drug testing guide. MIDAS cannot be retrofitted into such an assay simply because its study was randomized.

Can the reported improvements establish driving or workplace safety?

No. The patients were not recruited as drivers for an on-road assessment, and the study did not measure braking latency, standard deviation of lane position, pursuit tracking, surprise hazard avoidance, occupational injuries or aircraft operation. Although improved fatigue may matter for daily life, the relationship between feeling less fatigued and successfully managing a complex, fast-changing environment was not tested in this protocol.

A case-specific assessment would need to ask whether the person had a relevant neurological impairment, what task was undertaken, the contemporaneous level of sleepiness or attention, which medicines were taken, and what observed behavior or event evidence exists. The trial gives no calibrated function that accepts a modafinil pill history and outputs a patient’s probability of a collision or impairment at a specified clock time.

The same restraint applies to witness credibility or intent. A self-reported fatigue score says nothing about the truthfulness of a person recounting events. The study included no witness-identification, memory accuracy, malingering or credibility outcome. Presenting reduced fatigue as proof of an accurate or inaccurate witness would confuse entirely different constructs.

How strong is a 36-person double-blind crossover design?

Random assignment, matching placebo, blinding and within-person comparison are important strengths. They decrease the likelihood that the observed group difference is merely a contrast between two nonrandom groups with different baseline stroke severity. The crossover structure can also improve efficiency where a symptom is sufficiently stable and the intervention’s effects are reversible during washout.

There are nevertheless material limits. The study was conducted at one center, involved only 36 selected stroke survivors, and followed each course for six weeks. It could not establish effects after years of treatment. The one-week washout and sequence-sensitivity test provide reassurance but do not eliminate every period effect. Many secondary domain tests were performed. The participants used subjective scales for the most favorable findings, so the results are not equivalent to an independent objective function or laboratory assay.

Another limitation is the narrow enrollment. Investigators excluded clinically recognized or suspected sleep apnea and several psychiatric or medical confounders, which makes the findings less directly applicable to the many stroke survivors with those conditions. Medication compliance was monitored with returned tablets rather than chemical assays. These limitations define where the trial is useful; they do not erase the within-study fatigue and quality-of-life findings.

What information would a case reviewer need beyond MIDAS?

A responsible reconstruction begins with the actual medical and laboratory record for the person in question. A prescription, dispensing record or self-report may show intended medication use. It does not automatically demonstrate ingestion on a disputed date, timing of absorption, circulating concentration or impairment. If a concentration is relevant, review the specimen, draw time, matrix, analytical validation, identification criteria, calibration and units rather than substituting a dose from an unrelated trial.

If the disputed issue is function, obtain contemporaneous clinical observations and appropriate functional testing. The MIDAS MFI and SSQoL measured weeks-long patient experience; they were not standardized accident-reconstruction tools. A mental-status assessment and a real task-performance evaluation answer other questions. The site’s independently published therapeutic drug monitoring overview distinguishes clinical concentration interpretation from broader forensic claims, without implying that MIDAS supplied such concentrations.

Finally, separate population association, a plausible pharmacological mechanism and individual causal attribution. This experiment supports a within-person symptom advantage under a defined treatment protocol. It cannot independently identify why a specific stroke survivor was fatigued, why a particular fall happened, whether a vehicle collision would have occurred without modafinil, or which of several interacting medical factors caused an outcome.

Frequently asked questions about the MIDAS trial

1. Did 200 mg of modafinil per day significantly improve fatigue after stroke?

Yes, for the selected 36-person sample under the reported six-week crossover protocol: the main self-reported MFI treatment estimate favored modafinil by 7.38 points, and the companion Fatigue Severity Scale also favored treatment. These are questionnaire findings, not proof that all stroke survivors benefit or that objective vigilance universally improved.

2. Did the study prove modafinil improved overall memory or cognition?

No. The overall Montreal Cognitive Assessment difference was +0.32 point, p = .365, a nonsignificant result. The researchers reported nominally improved abstraction and visuospatial subscores, but those exploratory findings do not override the null global cognitive result or establish an individual memory-capacity change.

3. Was the placebo actually another stimulant such as caffeine?

No. The MIDAS placebo was a matching rice-powder tablet. This was a modafinil-versus-inert-placebo clinical trial, not a modafinil-versus-caffeine or modafinil-versus-amphetamine study. Drug-specific comparisons from different research populations should not be fused into a fabricated head-to-head result.

4. Does an MFI fatigue reduction mean a patient could safely drive?

No. Neither on-road driving nor simulated braking, lane weaving or collision avoidance was measured. A lower fatigue rating cannot be converted into a validated safe-driving threshold for an individual stroke survivor. Neurological deficits, time of day, medicines and actual performance would require independent assessment.

5. Did the researchers measure modafinil in whole blood or plasma?

No. They administered a defined oral tablet dose and measured subjective symptoms and brief clinical test scores. The paper did not report confirmed venous whole-blood, serum, plasma, urine, hair, saliva or breath modafinil concentrations. There is therefore no study-specific therapeutic or forensic concentration cutoff to extract.

6. Were mood and adverse events clearly better on modafinil?

No. Overall DASS and its depression, anxiety and stress domains were not significantly improved. Five adverse events occurred in modafinil periods and seven in placebo periods, with none classified as serious. Those small event counts do not prove a meaningful safety advantage or rule out rare harms.

7. Can the trial prove modafinil caused or prevented a particular accident?

No. MIDAS did not study accidents, eyewitness accuracy or individual forensic causation. It tested a group of selected patients’ reported fatigue and quality of life during two six-week clinical treatments. A litigated event calls for direct evidence of that person, the relevant time, mechanism, performance and competing causes.

Bottom line for clinical and forensic interpretation

In 36 selected adults with persistent post-stroke fatigue, six weeks of modafinil at 200 mg daily produced a reported improvement in self-rated fatigue and stroke-specific quality of life compared with matching placebo. Overall cognition and mood outcomes were not statistically significant. The small crossover experiment was useful as preliminary clinical evidence, not a chemical concentration test, individual impairment classifier, collision study or personalized causal certificate.

The key scientific discipline is to preserve what was actually measured: assigned oral tablets; repeated self-report fatigue and quality-of-life scales; brief cognitive and mood assessments; and adverse-event reports. It is equally important to preserve what was not measured: modafinil concentrations in legal whole blood or any other matrix, validated driving fitness, witness credibility and incident causation. The full original primary paper, Bivard et al., MIDAS, doi:10.1161/STROKEAHA.116.016293, supports those boundaries without needing an invented extrapolation.