[Translated title] Specialized cognitive interventions not superior for Long COVID symptoms, 2026, Michael Hüll

[Chandelier]

Spezialisierte kognitive Angebote bei Long-COVID-Symptomen nicht überlegen

Hüll, Michael

Abstract (Machine Translation)

Question: Do cognitive training and rehabilitation strategies improve cognitive symptoms (such as memory problems, impaired concentration, “brain fog”) in Long COVID?

Background: When various symptoms such as fatigue, shortness of breath, or cognitive dysfunction persist for more than twelve weeks after a SARS‑CoV‑2 infection, this is referred to as Long COVID.
Cognitive dysfunction in particular is common and burdensome because it substantially impairs patients’ ability to work and their quality of life.
To date there is no established, effective pharmacological or non‑pharmacological treatment. It is hypothesized that cognitive training programs, structured cognitive rehabilitation, and non‑invasive brain stimulation could promote neuroplasticity and alleviate symptoms.

Patients and Methods: The multicenter, randomized, controlled phase II trial enrolled 328 adults (median age 48 years; 73.5% female) with Long COVID.
Inclusion required self‑reported cognitive impairment. The study had five arms:

I)
Active control: unstructured computer puzzles and games (without an adaptive algorithm);

II)
BrainHQ: adaptive cognitive training targeting speed and accuracy;

III)
BrainHQ and PASC‑CoRE: a Long‑COVID–adapted cognitive rehabilitation program including mindfulness, goal management, and fatigue management;

IV)
BrainHQ and tDCS+: transcranial direct current stimulation, active;

V)
BrainHQ and tDCS–: transcranial direct current stimulation, sham (control arm for tDCS+).

Interventions were delivered over ten weeks, five times per week via telemedicine at home.
The primary endpoint was the change in the modified Everyday Cognition Scale‑2 (ECog2) score (self‑report covering the previous seven days) at the end of the intervention compared with baseline.

Results: None of the active interventions for cognitive training or cognitive rehabilitation showed a significant advantage over the active control or the other arms.
The adjusted differences in mean change were close to zero (e.g., BrainHQ vs. control: 0.0; 95% confidence interval –0.2 to 0.2).
All five groups showed some improvement over time (about 0.4 points on the ECog2 scale), but there were no differences between treatment groups.
A large proportion of participants (74%) reported subjectively that the intervention had helped, but this occurred equally across all arms (placebo/attention effect).

Improvement over time was inversely correlated with the duration of Long‑COVID symptoms (the shorter the symptom duration, the greater the spontaneous improvement), suggesting a natural course.

Conclusions: This large, rigorously conducted study found no efficacy for computerized cognitive training (BrainHQ), structured cognitive rehabilitation (PASC‑CoRE), or transcranial direct current stimulation (tDCS) in treating cognitive Long‑COVID symptoms.
The observation that all groups (including the active control) improved suggests that cognitive Long‑COVID symptoms often undergo spontaneous improvement or can be positively influenced by general attention and structure, regardless of the specific technique.

Comment by the author​

Spontaneous improvement more likely with shorter symptom duration
The study is of very good quality. However, blinding was only insufficiently possible for participants due to the nature of the intervention, which usually tends to favor false‑positive results. The use of an active control group underscores the scientific quality. Avoiding a biased comparison with a wait‑list control group, which for those affected would be perceived more as an adverse condition or a nocebo, was the right choice.

There is no indication in this negative study that specific interventions such as cognitive training or cognitive rehabilitation are effective.
Nevertheless, the study’s result shows that therapeutic optimism is justified when Long‑COVID has been present for a shorter time — regardless of which therapeutic measure is applied.
As with any symptomatology, chronification is a negative predictor for treatment response.
It is precisely in this area that non‑pharmacological and pharmacological interventions will have to be measured against each other.
Even though the demand for therapy is great, evidence on efficacy and side effects should determine the choice of treatment.

Spoiler: Original German Abstract

Fragestellung: Bessern kognitive Trainings- und Rehabilitationsstrategien die kognitiven Symptome (wie Gedächtnisprobleme, Konzentrationsschwäche, „Brain Fog“) bei Long-COVID?
Hintergrund: Sind verschiedene Symptome wie Fatigue, Kurzatmigkeit oder kognitive Dysfunktionen länger als zwölf Wochen nach einer SARS-CoV-2-Infektion nachweisbar, spricht man von Long-COVID. Insbesondere kognitive Dysfunktionen sind weit verbreitet und belastend, da sie die Arbeitsfähigkeit und Lebensqualität der Betroffenen erheblich beeinträchtigen. Bisher gibt es keine etablierte, wirksame medikamentöse oder nicht medikamentöse Behandlung. Es wird angenommen, dass kognitive Trainingsprogramme, strukturierte kognitive Rehabilitation und nicht invasive Gehirnstimulation die Neuroplastizität fördern und die Symptome lindern könnten.
Patienten und Methodik: In die multizentrische, randomisierte und kontrollierte Phase-II-Studie wurden 328 Erwachsene (Medianalter 48 Jahre; 73,5 % weiblich) mit Long-COVID aufgenommen. Einschlusskriterium waren selbstberichtete kognitive Beeinträchtigungen. Die Studie war fünfarmig angelegt:

1. I)
   aktive Kontrolle: unstrukturierte Computer-Puzzles und Spiele (ohne adaptiven Algorithmus);
2. II)
   BrainHQ: adaptives kognitives Training, das auf Geschwindigkeit und Genauigkeit trainiert;
3. III)
   BrainHQ und PASC-CoRE: auf Long-COVID angepasstes kognitives Rehabilitationsprogramm mit Achtsamkeit, Zielmanagement und Fatigue-Management;
4. IV)
   BrainHQ und tDCS+: transkranielle Gleichstromstimulation, aktiv;
5. V)
   BrainHQ und tDCS–: transkranielle Gleichstromstimulation, „sham“ (Kontrollarm für tDCS+).

Die Durchführung der Interventionen erfolgte über zehn Wochen, fünfmal pro Woche per Telemedizin zu Hause. Der Hauptendpunkt umfasste die Veränderung des modifizierten Everyday-Cognition-Scale-2(ECog2)-Scores (Selbstauskunft über die letzten sieben Tage) am Ende der Intervention im Vergleich zum Ausgangswert.
Ergebnisse: Keine der aktiven Interventionen zum kognitiven Training beziehungsweise zur kognitiven Rehabilitation zeigte einen signifikanten Vorteil gegenüber der aktiven Kontrolle oder den anderen Armen. Die adjustierten Unterschiede in der mittleren Veränderung waren nahe Null (z. B. BrainHQ vs. Kontrolle: 0,0; 95 %-Konfidenzintervall –0,2 bis 0,2). Alle fünf Gruppen erreichten über die Zeit eine gewisse Verbesserung (ca. 0,4 Punkte auf der ECog2-Skala), aber es gab keinen Unterschied zwischen den Behandlungsgruppen. Ein großer Teil der Teilnehmenden (74 %) berichtete subjektiv, dass die Intervention geholfen habe, jedoch geschah dies in allen Armen gleichermaßen (Placebo-/Aufmerksamkeitseffekt).
Die Verbesserung über die Zeit war invers mit der Dauer der Long-COVID-Symptome korreliert (je kürzer die Symptome, desto größer die spontane Verbesserung), was auf einen natürlichen Verlauf hindeutet.
Schlussfolgerungen: Diese große, rigoros durchgeführte Studie konnte keine Wirksamkeit für computergestütztes kognitives Training (BrainHQ), strukturierte kognitive Rehabilitation (PASC-CoRE) oder transkranielle Gleichstromstimulation (tDCS) bei der Behandlung kognitiver Long-COVID-Symptome nachweisen. Die Beobachtung, dass sich alle Gruppen (inklusive der aktiven Kontrolle) verbesserten, deutet darauf hin, dass kognitive Long-COVID-Symptome oft einer spontanen Besserung unterliegen oder durch allgemeine Aufmerksamkeit und Strukturierung positiv beeinflusst werden, unabhängig von der spezifischen Technik.
Knopman DS et al. Evaluation of interventions for cognitive symptoms in Long COVID: a randomized clinical trial. JAMA Neurol. 2026;83(1):49-59

Kommentar​

Spontanverbesserung bei kurzer Symptomdauer wahrscheinlicher​

Die Studie ist von sehr guter Qualität. Die Verblindung ist jedoch aufgrund der Natur der Intervention für die Teilnehmenden nur unzureichend möglich, was zumeist aber falsch-positive Ergebnisse begünstigt. Die Verwendung einer aktiven Kontrollgruppe unterstreicht die wissenschaftliche Qualität. Die Vermeidung eines verfälschenden Vergleichs mit einer Wartekontrollgruppe, die für die Betroffenen eher als adverse Bedingung oder Nocebo zu sehen ist, war richtig. Es gibt in dieser negativen Studie keinen Hinweis darauf, dass spezifische Interventionen wie kognitives Training beziehungsweise kognitive Rehabilitation wirksam sind. Nichtsdestotrotz zeigt das Ergebnis der Studie, dass bei kürzerem Bestehen von Long-COVID therapeutischer Optimismus herrschen darf – egal welche Therapiemaßnahme erfolgt. Wie bei jeder Symptomatik ist die Chronifizierung ein negativer Prädiktor für ein Therapieansprechen. Gerade hier werden sich aber nicht medikamentöse und medikamentöse Interventionen messen lassen müssen. Auch wenn der Ruf nach Therapie groß ist, sollten Evidenz zu Wirksamkeit und Nebenwirkungen die Therapieauswahl bestimmen.

Web | DOI | PDF | InFo Neurologie + Psychiatrie

 

[Eleanor]

This is this Recover-Neuro trial, I think? (same numbers of patients etc)

This bit made me laugh.

There is no indication in this negative study that specific interventions such as cognitive training or cognitive rehabilitation are effective.
Nevertheless, the study’s result shows that therapeutic optimism is justified when Long‑COVID has been present for a shorter time — regardless of which therapeutic measure is applied.

I often use "Nevertheless" as a keyword search when trying to get a quick sense of a paper - often an indicator that the author's beliefs aren't going to be impacted by evidence.

 

[Turtle]

Or the paper was written by a foreigner.
I was taught in England to use words like this.

 

[Chandelier]

I often use "Nevertheless" as a keyword search when trying to get a quick sense of a paper - often an indicator that the author's beliefs aren't going to be impacted by evidence.

I was taught in England to use words like this.

"Nichtsdestotrotz" is a word formerly invented by students (German Wikipedia) and comes with some interesting history:

(Machine translation from Duden.de)

The origin of the (non‑)word nichtsdestotrotz

The hybrid nichtsdestotrotz, formed from nichtsdestoweniger (“nevertheless”) and trotzdem (“however”), is a thorn in the side of some speakers.
Here you can learn more about its origin and how its usage has developed.

For many people, the word is something like a “non‑word,” since it is essentially a meaningless blend of nichtsdestoweniger and trotzdem.
For others, however, it is a completely “normal” word that they use without hesitation.
And because its usage has been increasing year by year, it is now included in the Duden.
If one takes the view that the meaning of a word is determined by its usage, then nichtsdestotrotz simply means “nevertheless,” and there would be nothing more to say.
The Duden, however, labels the word as “colloquial” for good reason.

The word is attested since the late 19th century, especially in the Berlin and Leipzig regions.
When Alfred Döblin used it in his 1918 novel Wadzek’s Battle with the Steam Turbine, this does not mean that nichtsdestotrotz was already accepted in formal written language — it only shows that Döblin was particularly close to the pulse of spoken language. It has often been speculated that a language critic like Kurt Tucholsky or the comedian Heinz Erhardt picked up the word and turned its absurdity into a sketch.
However, no evidence for this has ever been found. It is quite possible, though, that a rarely used word like nichtsdestotrotz became unintentionally popular only after being criticized publicly.

For many people, it remains a taboo word today in contexts that are not clearly humorous.
Anyone who uses it still risks being laughed at.
Whether one wants to use this word is something each person must decide for themselves — in light of its history.
But the example shows, in any case, that language is not only a means of communication, but also plays a role in how people are classified and judged within society.

 

[rvallee]

Improvement over time was inversely correlated with the duration of Long‑COVID symptoms (the shorter the symptom duration, the greater the spontaneous improvement), suggesting a natural course.

I think this is the first paper to dare say it plainly, even though they fumble it at the end with the weird statement about optimism. Most academics and clinicians have chosen instead to use this as evidence that their ineffective junk works, using odds in their favor, which they tip further on purpose.

This has been known for years, there is no credible excuse against it. It's very much like prosecutors choosing to hide evidence simply because otherwise they wouldn't be able to convict. Which is the whole point. Medicine has far lower standards of evidence than fair legal systems, and that's a low bar already. In a fair and competent health care system, doing so should lead to harsh penalties. Instead it leads to promotions and an endless stream of cash.

I disagree that there is a hypothesis here, though. This is a model. It doesn't make any sense, in fact makes a mockery of known facts. It's an assertive model, it simply says "this must work", and then tries to prove that it works, indifferent that it doesn't. There is no credible basis for it, it simply derives from the traditional generic conversion disorder model, and is assumed to work on this basis alone.

The adjusted differences in mean change were close to zero (e.g., BrainHQ vs. control: 0.0; 95% confidence interval –0.2 to 0.2).
All five groups showed some improvement over time (about 0.4 points on the ECog2 scale), but there were no differences between treatment groups.
A large proportion of participants (74%) reported subjectively that the intervention had helped, but this occurred equally across all arms (placebo/attention effect).

This is why most pragmatic trials choose subjective questionnaires, even when there are objective measures that can be used. Because the subjective questionnaires offer a permission structure to lie. It's why no evidence relying on them can ever be trusted, and everyone who trots them out as evidence know this, and argues it when they don't believe in what's being falsely claimed to be effective.

What's the difference between alternative and real medicine? Real medicine works. Psychobehavioral ideology is alternative medicine, it does not work. The difference no longer exists much, but it remains that the near totality of the achievements of medicine are purely biomedical, the rest combined is almost entirely irrelevant as far as improving outcomes. The streak continues: science works, pseudoscience does not.