Identifying EEG biomarkers of fatigue and brain fog in long COVID: insights from rest and movement-related brain activity, 2026, Casado Sánchez et al.

[Chandelier]

Identifying EEG biomarkers of fatigue and brain fog in long COVID: insights from rest and movement-related brain activity

Spoiler: Authors

Casado Sánchez, M.; Morales Fajardo, K.; Long, S.; Janaudis-Ferreira, T.; Duclos, C.; Boudrias, M.H.

Highlights​

• Long COVID showed reduced post-movement beta rebound in somatomotor cortex.
• Slower sensorimotor alpha frequency at rest reflected fatigue-linked excitability.
• EEG changes tied to fatigue and brain fog suggest disrupted E/I balance.

Abstract​

Background​

Fatigue and brain fog are among the most reported symptoms in Long COVID (LC), yet their underlying neural mechanisms remain unclear.
In this context, we aimed to identify neural correlates associated with LC symptoms.

Methods​

Electroencephalography (EEG) was used to quantify brain oscillatory activity during sustained handgrips and rest in the left central region overlying the primary hand motor (C3) and somatosensory areas (C5).
A sample of twenty LC participants reporting fatigue, brain fog, and post-exertional malaise (PEM) was analyzed, along with 20 age- and sex-matched controls.
All participants were right-handed, and EEG activity was recorded on the dominant left hemisphere over motor and somatosensory areas.
The main LC symptoms were assessed using the Fatigue Severity Scale (FSS), the Cognitive Failures Questionnaire (CFQ), and the DePaul Symptom Questionnaire for PEM (DSQ-PEM).

Results​

LC participants showed a smaller post-movement beta rebound (PMBR) during handgrips in the left central region (electrode C3, p = 0.02) compared to controls, which negatively correlated with cognitive symptoms (r = −0.35, p = 0.02).
They also showed a slower alpha central frequency (CF) at rest (electrode C5, p = 0.003), which correlated with fatigue (r = −0.264, p = 0.019).

Conclusions​

Both the lower PMBR and slower alpha CF suggest impaired inhibitory control and reduced excitability in the sensorimotor cortex of LC individuals.
Together, these features, which were associated with fatigue and brain fog, highlight altered dynamics in the sensorimotor cortex of LC individuals.
These findings offer preliminary mechanistic insight into LC symptoms and suggest that PMBR and alpha CF warrant further investigation as candidate EEG markers of this condition.

Web | DOI | Journal of Clinical Neuroscience | Open Access

 

[Chandelier]

Also see previous doctoral thesis thread:
Identifying electroencephalography (EEG)-based biomarkers of Long COVID, 2026, Casado Sánchez

 

[Chandelier]

Summary by Vipin M. Vashishtha on Twitter:

---

#LongCOVID “brain fog” is beginning to reveal a measurable electrical signature in the brain.

EEG recordings from people with LongCOVID showed abnormal activity in the sensorimotor cortex, including a weaker post-movement beta rebound and slower alpha rhythms. 1/

Strikingly, these changes correlated with cognitive difficulties and fatigue, offering a potential biological explanation for symptoms that are often dismissed as subjective. 2/

The brain’s electrical rhythms may provide a new window into the biology of Long COVID—and potentially a future objective biomarker of “brain fog.” 3/3

---

 

[Siebe]

The paper uses a very mild Bonferroni correction. The thesis Appendix III has the False Discovery Rates with Benjamin-Hochberg correction.

Raw and FDR-adjusted p-values for every EEG–symptom correlation:

• PMBR C3 vs CFQ: raw p = 0.02, FDR p = 0.59.
• Alpha CF C5 vs FSS: raw p = 0.01, FDR p = 0.59.
• Alpha CF C5 vs CFQ: raw 0.04, FDR p = 0.81.

Looks like the typical 'test many things and then hypothesize after results are known'. Like, check out these long lists!

All credit goes to Claude Opus 5.0 + my personal peer review guide. I checked the figures myself.

 

[forestglip]

The paper uses a very mild Bonferroni correction. The thesis Appendix III has the False Discovery Rates with Benjamin-Hochberg correction.

I think that Bonferroni is usually (always?) more strict than BH correction.

- https://www.statsig.com/blog/controlling-type-i-errors-bonferroni-benjamini-hochberg

We generally recommend the Benjamini-Hochberg Procedure as a less severe measure than the Bonferroni Correction, but which still protects you from some amount Type I errors.

The Benjamini-Hochberg Procedure helps most when there are a large number of hypotheses you’re testing and want a moderate reduction in false positives, whereas the Bonferroni Correction is more conservative and will be most useful when there are a smaller number of hypotheses being tested concurrently.

 

[Siebe]

I think that Bonferroni is usually (always?) more strict than BH correction.

- https://www.statsig.com/blog/controlling-type-i-errors-bonferroni-benjamini-hochberg

Right, but they are applied differently here. They used Bonferroni while defining the family as 2 electrodes. The BH correction in the thesis, on the other hand, uses all the correlations as a family so it's much stricter.

Regarding what's the appropriate family to choose, I've learned it as follows:
> A family should contain every test that would have been reported as a finding had it come out significant.

Maybe not all the tests in the correlation table would match that, but I think 14 or 28 would (2 electrodes x 7 variables x [maybe?] pre- and post-task). Not the 2 electrodes.

 

[forestglip]

Right, but they are applied differently here. They used Bonferroni while defining the family as 2 electrodes. The BH correction in the thesis, on the other hand, uses all the correlations as a family so it's much stricter.

I see what you mean. I haven't read the paper in detail, but in the methods section, they say they applied Bonferroni to the hypothesis-driven tests, and BH to the rest:

Bonferroni correction was applied to hypothesis-driven tests, and FDR correction was applied to exploratory analyses (reported in the Supplementary Material).

I think that's probably valid, as long as they fully reported all the pre-planned "hypothesis-driven tests".

 

[Siebe]

I see what you mean. I haven't read the paper in detail, but in the methods section, they say they applied Bonferroni to the hypothesis-driven tests, and BH to the rest:

I think that's probably valid, as long as they fully reported all the pre-planned "hypothesis-driven tests".

They'd need to actually clarify beforehand which tests were hypothesis driven. Instead, the thesis states that were chosen based on the results being positive? (If I'm reading this correctly??)

Hypothesis-driven correlations were restricted to EEG metrics that showed significant
effects between groups and tested against the two main symptom scores (CFQ and FSS).
As these analyses followed significant group differences and were interpreted
independently per electrode and behavioural score, no further correction for multiple
comparisons was applied to these correlation tests. (p.73)

Even if it was like you said, that still need to define which tests those are, and they're more likely to be 14 or 28 than 2.