---
title: "A prospective observational study of post-COVID-19 chronic fatigue syndrome following the first pandemic wave in Germany and biomarkers associated with symptom severity"
authors: "Kedor C, Freitag H, Meyer-Arndt L, Wittke K, Hanitsch LG, Zoller T, Steinbeis F, Haffke M, Rudolf G, Heidecker B, Bobbert T, Spranger J, Volk HD, Skurk C, Konietschke F, Paul F, Behrends U, Bellmann-Strobl J, Scheibenbogen C"
journal: "Nature Communications"
year: 2022
doi: "10.1038/s41467-022-32507-6"
url: "https://www.nature.com/articles/s41467-022-32507-6"
conditions: ["ME/CFS", "Long Covid"]
studyType: "Observational cohort, cross-sectional at 6 months, n=42 post-COVID plus 19 ME/CFS comparison patients"
kind: paper
reviewedOn: 2026-10-19
guideVersion: "0.8.1"
model: "claude-opus-5-5"
effort: "high"
read:
  - "Full text"
  - "Figs 1 to 5, viewed as images (Figs 1, 2 and 4 added after publication)"
  - "Supplementary information"
  - "Peer review file (Supplementary Information 2)"
  - "Source data spreadsheet, used for the women-only reanalysis of Fig. 5"
  - "https://doi.org/10.1038/s41467-022-33784-x"
score: 5
verdict: "Careful early evidence that some people meet strict ME/CFS criteria after mild COVID-19, but the headline blood-marker links to grip strength appear to pool men and women, and most weaken or vanish in women alone."
confidence: "moderate"
importance: "moderate"
strengths:
  - "Strict criteria, with a real medical work-up"
  - "Mostly confirmed, mostly non-hospitalised infection"
  - "Effect sizes and source data published"
  - "Limitations stated openly"
weaknesses:
  - "Self-referred patients; cannot show how common"
  - "No healthy or recovered controls measured"
  - "Blood-marker correlations not adjusted for sex"
  - "Mental health claims exceed the measures used"
guideNotes:
  - "Correlations in mixed-sex groups: check for sex adjustment, since sex alone drives many blood values and strength measures"
  - "Check whether a title's headline finding matches the analysis that survived correction"
design: observational
topics: ["diagnosis-biomarkers", "course-outcomes"]
humanInput: none
---

## Study design

This is a single-centre observational study from the Charité Fatigue Center in Berlin. It is a substudy of the Pa-COVID-19 cohort. The paper calls itself "prospective", but what it reports is a cross-sectional look at one time point: about six months after infection. Follow-up results are not part of this paper.

The study has three groups:

- 42 people with post-COVID-19 syndrome (PCS) who had COVID-19 between March and June 2020 and still had "moderate to severe fatigue and exertion intolerance" six months later. They were seen between August and November 2020.
- These 42 were split by the Canadian Consensus Criteria (CCC) for myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). 19 met them (called PCS/ME/CFS) and 23 did not (PCS/non-ME/CFS).
- 19 people with ME/CFS after a non-COVID infection, picked from 123 such patients seen at the clinic in the same period. They were matched for age and sex, and were the 19 with the shortest illness: median 13 months, range 7–19.

There is no healthy control group, no group of people who recovered from COVID-19, and no group with fatigue from another cause. The authors list all three as limitations. Grip strength and lab values are compared with published reference values instead.

The authors describe the study as "purely exploratory with hypothesis generating character". That is the right frame. The question is whether the abstract, title and discussion stay inside it.

## Sample selection

### How people got into the study

Recruitment was by self-referral. The clinic announced the study on its website, and a German long COVID patient group shared it. 81 people made contact. 24 were screened out by questionnaire, 57 were seen in clinic, and 42 were included. Of the 15 excluded in clinic, 5 had no evidence of infection, 9 had evidence of another illness, and 1 had a history of one (Supplementary Fig. S2).

This is a selected sample at a specialist ME/CFS centre, filtered for the symptoms that define ME/CFS. So the share who met the criteria (19 of 42) says nothing about how often COVID-19 leads to ME/CFS. The authors say so: "we do not know how representative they are."

### Infection confirmation

The study required a positive polymerase chain reaction (PCR) test or positive antibody test (serology). The paper does not say how many had which. Three people with no PCR test and negative serology were still included because they had lost their sense of smell and taste. That is reasonable for the first wave, when testing was scarce, but it sits oddly with the 5 people excluded for "no evidence for SARS-CoV-2 infection". The paper does not explain where the line was drawn.

### Acute illness severity

32 had mild COVID-19 and 10 had moderate illness with pneumonia. Three were treated in hospital, and none needed oxygen. This is a mostly non-hospitalised cohort, which is the most useful kind for this question.

### Case definition and post-exertional malaise

ME/CFS was diagnosed with the CCC, with one change: post-exertional malaise (PEM) had to last at least 14 hours rather than 24. The 14-hour cut-off comes from Jason and colleagues' work. PEM frequency, severity and duration were measured with the brief PEM questionnaire of Cotler et al. (2018). That is a reasonable choice, though the guide's preferred instrument is the DePaul Symptom Questionnaire, version 2 (DSQ-2). When a reviewer asked about validation, the authors replied: "The PEM score is not validated and was used descriptively."

The work-up to rule out other causes was thorough by the standard of this field. Everyone had an electrocardiogram, echocardiogram, chest X-ray and lung function tests, and saw a rheumatologist. Some also saw a neurologist or cardiologist, or had a chest computed tomography (CT) scan, depending on symptoms. Anyone with reduced lung diffusion capacity was excluded. This applied work-up raises the share of true ME/CFS cases among those labelled ME/CFS. For a clinically diagnosed CCC group with this work-up, the guide's estimate of that share (f) would be well above the 0.5–0.6 it gives for a diagnosis by a general practitioner alone.

Two weaknesses remain. First, psychiatric illness was screened only with the 9-item Patient Health Questionnaire (PHQ-9), not with a clinical interview (see below). Second, people with "relevant cardiac, respiratory, neurological, or psychiatric comorbidity" were excluded. That makes the group cleaner, but it may also remove some true cases.

The split between groups was mostly about PEM. 18 of the 23 who did not meet the criteria had PEM lasting under 14 hours, and 8 of 23 did not meet the cognitive criteria. So the "non-ME/CFS" group is a group of people with shorter PEM. It is not a group without PEM. All 23 rated PEM at 1 or more on the symptom scale (Table 2), though one scored 0 for PEM severity on the PEM questionnaire (Fig. 2). Fig. 2 shows how sharp the split is: 17 of the 23 reported PEM lasting 2–10 hours, while all 19 in the PCS/ME/CFS group reported 14 hours or more, 15 of them over 24 hours. 5 of the 23 had PEM of 14 hours or more but missed other criteria.

### Comparison group

The non-COVID ME/CFS group is a sensible comparator for "is post-COVID ME/CFS the same illness?". But it differs in ways that matter. Its illness had lasted about twice as long, and it had several different triggers (9 unspecified respiratory infections, 3 Epstein-Barr virus, 1 gastroenteritis, 6 other). It also has fewer measurements. The 9-item Patient Health Questionnaire and the Epworth Sleepiness Scale were not done in this group, and the standing test results and the lab table (Table 4) are only given for the two PCS groups.

### Conditions the guide asks about

- **Postural orthostatic tachycardia syndrome (POTS) and orthostatic intolerance:** measured with a 10-minute active stand test. 4 people in the PCS/ME/CFS group and 1 in the PCS/non-ME/CFS group met POTS criteria. Orthostatic hypotension was found in 6 and 1. This is reported only for the PCS groups, and the figure (Fig. 4) shows women only. In the women, median heart rate rose by roughly 10–15 beats per minute on standing in both groups, so most did not have marked tachycardia.
- **Mast cell activation syndrome (MCAS):** not mentioned.
- **Severity:** measured with the Bell disability scale. The median was 40 for all PCS patients (50 in PCS/non-ME/CFS) and, by the paper's text, 30 for non-COVID ME/CFS. Fig. 1 shows a median nearer 35 for that group (n = 18), so the text and figure do not quite agree. A score of 30–40 means someone can do light work for 2–4 hours a day. These are mostly moderately affected people who could travel to a clinic. Severely affected, housebound people are absent, as one would expect from a clinic-based design.
- **Illness duration:** 6 months in the PCS groups, about 13 months in the comparison group.

## Outcome measures

Most outcomes are questionnaires: the Chalder Fatigue Questionnaire (CFQ), the Bell scale, the 36-item Short Form Health Survey (SF-36), the PEM questionnaire, the Composite Autonomic Symptom Score (COMPASS-31), a 1–10 rating of each CCC symptom, the 9-item Patient Health Questionnaire and the Epworth Sleepiness Scale. For a descriptive study, that is fine. Nothing here is a treatment effect that expectation could inflate. One reviewer noted that the Bell scale has not been formally validated. The SF-36 role-limitations score sits at the floor: every PCS/ME/CFS patient and most PCS/non-ME/CFS patients scored 0 (Fig. 1). So the significant difference between the two groups on that score comes from a handful of people in the non-ME/CFS group who scored above 0.

**Hand grip strength.** Hand grip strength (HGS) was measured as ten maximal pulls, repeated 60 minutes later. The paper treats it as a measure of muscle fatigue: "HGS is a reliable parameter to assess muscle fatigue". The guide puts grip strength in the high-risk group for performance bias, because it depends on effort. Here the effort concern is less about expectation and more about the fact that a person who feels ill, or fears PEM, may not pull at full force. That does not make the measure useless, but it is not an objective muscle measure. Both tests come after the effort of travel to the clinic, the second one 60 minutes into the clinic visit, and PEM status on the day was not recorded.

Grip results are shown only for women (13 or 14 per group), compared with cut-offs from a published reference study (Jäkel et al., 2021). No healthy people were measured in this study.

**Standing test.** Heart rate and blood pressure sitting and after 0, 2, 5 and 10 minutes standing. This is an objective measure, with the usual caveat that travel to the clinic can raise resting heart rate.

**Lab tests.** Standard clinical blood tests from the hospital laboratory, plus angiotensin-converting enzyme 2 (ACE2) by an enzyme-linked immunoassay (ELISA) and interleukin 8 (IL-8) measured inside red blood cells. The red-cell IL-8 test is unusual. The paper claims it shows "inflammation during the last 3–4 months", based on one cited study. No control group was measured with it in this paper, so the finding that 37% and 48% were above the normal value rests on the reference cut-off alone.

## Bias

**Selection.** Discussed above. Self-referral to an ME/CFS clinic selects for ME/CFS-like illness.

**Author preconceptions.** The senior author co-wrote a "unifying hypothesis" of ME/CFS built on autoantibodies against adrenergic receptors (Wirth and Scheibenbogen, cited here). The paper also cites an endothelial dysfunction study by Haffke et al., whose first author is a co-author here, and states that "there is emerging evidence that postinfectious ME/CFS has an autoimmune mechanism". The paper states its expectation openly: "We hypothesized that COVID-19 can lead to a persistent fatigue syndrome which fulfills diagnostic criteria of ME/CFS". This is not a problem in itself. It matters in one place: the biomarker interpretation, where small correlations are read as support for hypoperfusion and endothelial dysfunction (see below).

**Funding.** A grant from the Weidenhammer-Zoebele Foundation, and the VolkswagenStiftung for the statistician. No competing interests declared.

**Selective reporting.** The peer review file shows that correlations between fatigue scores and platelets and thyroid tests were dropped from the text after they did not survive correction. The authors wrote: "we decided not to mention this negative data in the revision". Standard errors and test statistics for these correlations remain in the supplement (Table S7), so this is partly visible, but the main text only reports the correlations that survived. The text also mentions a non-significant bilirubin correlation in the non-COVID group as "similar", while not mentioning the larger non-significant IL-8 correlation in the same group (ρ = −0.61 to −0.75 in Fig. 5).

## Statistical and methodological issues

### The group comparisons

The group comparisons are done with rank-based tests (Dunn-type multiple contrasts and Brunner-Munzel tests), with p values corrected across endpoints by the Benjamini-Hochberg (BH) method. This was added in revision, after reviewers pointed out that the first version did not correct for multiple testing. The supplement gives effect sizes and confidence intervals for every comparison, which is good practice. With 13–23 people per group, depending on the measure, the study can only detect large differences. "Not significantly different" between the two ME/CFS groups is weak evidence that they are the same.

Comparing PEM between the ME/CFS and non-ME/CFS groups is partly circular, because PEM length is how they were split. The paper acknowledges this ("As expected due to the diagnostic criteria").

### The grip-strength correlations

This is the part of the paper that gives it its title, and it is the weakest. The paper computes Spearman correlations between four grip measures and nine lab values (eight in the non-COVID group), separately in each of three groups: 104 correlations, plus the dropped fatigue correlations. It then builds a mechanism on the ones that pass BH correction.

The biggest problem is sex. Each PCS group contains men and women: 15 women and 8 men in PCS/non-ME/CFS, and 14 women and 5 men in PCS/ME/CFS. Men have much higher grip strength than women. They also have higher haemoglobin, creatinine, ferritin and bilirubin, and lower N-terminal pro-B-type natriuretic peptide (NT-proBNP). The haemoglobin, ferritin, bilirubin and NT-proBNP correlations the paper highlights all go in the direction that sex alone would produce. The paper does not say whether the correlations were done within sex, and it does not adjust for sex.

The source data allow a partial check. The source data have no sex column, but the structure is clear. Each PCS group starts with a block of records with lab values and no grip values: 8 in PCS/non-ME/CFS and 5 in PCS/ME/CFS, matching the number of men in Table 1. The records with grip data (13 and 14) match the women-only numbers in Fig. 3. Two more records in PCS/non-ME/CFS, apparently women, also have no grip data. I treat the first blocks as the men. Their haemoglobin is visibly higher: 13.7–17.6 g/dl in PCS/non-ME/CFS men, against 12.2–14.4 g/dl in women with grip data. I recalculated the correlations in women only:

| Correlation with grip strength | Published (Fig. 5) | Women only, from the source data |
|---|---|---|
| Haemoglobin, PCS/non-ME/CFS | 0.46 to 0.59 | −0.07 to 0.23 |
| Haemoglobin, PCS/ME/CFS | 0.64 to 0.76 | 0.37 to 0.55 |
| Ferritin, PCS/ME/CFS | 0.37 to 0.55 | −0.06 to 0.35 |
| Bilirubin, PCS/ME/CFS | 0.44 to 0.53 | 0.23 to 0.43 |
| IL-8, PCS/non-ME/CFS | −0.44 to −0.56 | −0.27 to 0.09 |
| C-reactive protein (CRP), PCS/non-ME/CFS | −0.43 to −0.54 | −0.41 to −0.25 |
| NT-proBNP, PCS/ME/CFS | −0.70 to −0.79 | −0.51 to −0.73 |
| ACE2, PCS/non-ME/CFS | 0.32 to 0.58 | 0.12 to 0.55 (n = 11) |

The published values cannot be reproduced from the women alone. Their size is consistent with men being included. In women only, the haemoglobin and IL-8 associations in the PCS/non-ME/CFS group and the ferritin association in the PCS/ME/CFS group largely vanish, The haemoglobin association in PCS/ME/CFS weakens but stays moderate, and the C-reactive protein and bilirubin associations weaken. Two largely survive: NT-proBNP in PCS/ME/CFS, at about −0.5 to −0.7 in 13 women, and ACE2 in PCS/non-ME/CFS, where the one significant measure (0.58) becomes 0.45 in 11 women. These are the findings I would call worth following up, with the caveat that they are two of 104 correlations.

I could not do the full analysis, because the men's grip values are not in the source data. The reanalysis is mine, not the paper's.

The study did not measure PEM status on the day, iron or hydration status, or physical activity. Any of these could link grip strength to blood values without the mechanism the paper proposes.

### Other points

- The lab table (Table 4) reports how many people were out of the normal range, but no comparison group was measured. Statements such as mannose-binding lectin (MBL) deficiency being "more frequent" rest on 3 of 18 and 5 of 22 people, set against a historical figure of 6% from another study. There is no formal test.
- The 9-item Patient Health Questionnaire medians were 11 and 12. By the paper's own cut-offs, that is "moderate depressive symptoms" (10–14). The scale includes fatigue, sleep and concentration items, which ME/CFS raises regardless of mood. So the score is hard to read either way.

## Biological plausibility

The proposed chain is: lower haemoglobin, bilirubin, ferritin and ACE2 and higher NT-proBNP mean worse oxygen delivery or more vasoconstriction, which means hypoperfusion, which causes muscle fatigue. Each link is a hypothesis. For haemoglobin, the simplest explanation for the correlation is sex. Bilirubin and ferritin also differ by sex. The idea that NT-proBNP comes from ischaemic skeletal muscle cites one study on satellite cells. All NT-proBNP values were close to normal (median 47 and 39 ng/l), and the authors report no evidence of heart failure.

Hypoperfusion and endothelial dysfunction in ME/CFS and long COVID have support from other, more direct measures, including the authors' own endothelial function study (Haffke et al.). But this paper does not measure perfusion or endothelial function. It measures blood values and grip strength. Against the guide's "theory" questions, the paper does not try to explain how hypoperfusion would produce the delay and duration of PEM, or PEM after cognitive or emotional exertion. In fairness, it does not claim to.

The descriptive findings fit other work. A subset of people with long COVID meet strict ME/CFS criteria. The Italian study the authors cite found 10 of 37.

## Interpretation and spin

The limitations paragraph is candid. The spin is mostly in the abstract, the title and parts of the discussion.

- **Title:** "biomarkers associated with symptom severity". The biomarkers were associated with grip strength, not with symptom severity. The correlations with fatigue questionnaires did not survive correction and were dropped.
- **Abstract:** "Hand grip strength is diminished in most patients compared to normal values in healthy." No healthy people were measured. In the PCS/non-ME/CFS women, grip strength sits close to the reference cut-offs, and several are above them (Fig. 3).
- **Abstract:** the correlations "may indicate low level inflammation and hypoperfusion as potential pathomechanisms". The hedging is appropriate. The discussion is firmer: the correlations "point to endothelial dysfunction and hypoperfusion as cause of muscle fatigue", and the NT-proBNP finding "is consistent with our hypothesis of hypoperfusion". Given the likely sex confounding, the first of these is not supported.
- **Introduction:** "Our findings confirm initial concerns that COVID-19 leads to persistent fatigue syndromes in a subset of young individuals following mild to moderate infectious disease." A selected case series without uninfected or recovered controls shows that such people exist after COVID-19. It cannot show how often this happens, and only weakly supports the causal reading on its own. The timing (first wave, confirmed infection, previously well) does make the causal reading plausible.
- **Discussion:** "we have, however, no evidence of ... major depressive or anxiety diseases", and later, "mental health is not relevantly impaired in most patients with PCS". No anxiety measure is reported. Depression was screened only with the 9-item Patient Health Questionnaire, on which the medians fall in the moderate range by the paper's own cut-offs. "No evidence" here means "not looked for in depth", and the stronger sentence about mental health is not supported by these data.
- **Conclusion:** early "pacing and coping" offering "a better therapeutic prognosis" is an aspiration, not something this study tested.

## What the study actually demonstrates

- Some people who had mostly mild, confirmed COVID-19 in spring 2020 met strict ME/CFS criteria (CCC with a 14-hour PEM threshold) six months later, after a substantial work-up for other causes.
- Their symptom burden on questionnaires was broadly similar to people with ME/CFS after other infections. Flu-like symptoms were less severe.
- People with post-COVID fatigue who did not meet the criteria still had PEM, mostly shorter, and were still substantially disabled (median Bell score 50).
- Orthostatic problems (POTS or orthostatic hypotension) showed up in a minority, more often in the ME/CFS group.

What it does not show:

- How common post-COVID ME/CFS is.
- That grip strength is lower than in comparable healthy people (no one was measured).
- That haemoglobin, ferritin, bilirubin, IL-8 or C-reactive protein relate to muscle fatigue independent of sex.
- Any mechanism.

## Critical missing elements

- Sex-stratified or sex-adjusted correlations, or a statement of which sexes were included in Fig. 5.
- The men's grip-strength data.
- A breakdown of infection confirmation (PCR, serology, neither).
- A healthy or recovered-COVID control group measured with the same tests.
- Lab values, standing test results and depression and sleepiness scores for the non-COVID ME/CFS group.
- PEM status at the time of grip testing and blood draw.
- A structured psychiatric assessment and any anxiety measure, if mental health claims are to be made.
- A breakdown of the 24 screened out by questionnaire.

## Recommendations

**For readers and clinicians:** Use this as early, careful clinical evidence that ME/CFS by strict criteria occurs after mild COVID-19. Do not use it to estimate how common that is. Do not cite the grip-strength biomarker correlations as evidence for hypoperfusion or inflammation.

**For researchers:** The NT-proBNP and grip-strength association in post-COVID ME/CFS is the clearest correlation to survive a rough sex check, with ACE2 in the non-ME/CFS group a weaker second. Test it with sex-stratified analysis in a larger sample and a pre-specified hypothesis. Measure perfusion directly if hypoperfusion is the claim.

**For the authors:** Publish the men's grip data and a sex-adjusted reanalysis of Fig. 5. Retitle or reframe the biomarker claims accordingly.

## Final assessment

The descriptive half of the paper is solid for what it is. The case definition is strict and stated, PEM was measured with a published questionnaire, other causes were ruled out with a real clinical work-up, infection was mostly confirmed, the cohort is mostly non-hospitalised, and the time point is clear. Statistics were corrected in revision, effect sizes are given, and source data are published. That is enough to make it a useful early document that post-COVID ME/CFS exists and resembles ME/CFS after other infections.

The analytic half, which the title leads with, is weak. 104 correlations in small groups, with no sex adjustment, in groups that very likely mix men and women, produce mostly the correlations that sex differences would predict. In women only, most weaken and several vanish. The mechanistic reading built on them goes beyond the data. Grip strength is effort-dependent, and its "diminished" status rests on external cut-offs. The mental health statements are stronger than the measures allow.

What prevents a higher score: no controls measured with the same tests, a self-referred sample, and a biomarker analysis that is very likely confounded by sex and still drives the title and part of the abstract.

What prevents a lower score: the central descriptive claim, that a subset meets strict ME/CFS criteria after mild COVID-19 and looks like other post-infectious ME/CFS, is supported. The authors call the study exploratory and list the missing controls themselves. The source data are good enough that a reader can check and partly correct the main analysis.

## Executive summary

This early study from Berlin looked at 42 people who still had fatigue and exertion intolerance six months after COVID-19, mostly mild illness in spring 2020. 19 met strict criteria for ME/CFS after a careful medical work-up to rule out other causes. Their symptoms were broadly similar to people who got ME/CFS after other infections. The other 23 were also quite disabled and also had a worsening of symptoms after exertion, but in most it lasted less long.

That descriptive part is useful and fairly careful. But the people were self-referred to a specialist ME/CFS clinic, so the study cannot say how often COVID-19 leads to ME/CFS.

The title highlights "biomarkers associated with symptom severity". In fact these are about 100 correlations between grip strength and blood values in small groups. Men and women appear to have been analysed together, and the correlations the paper stresses are the ones you would expect from sex differences alone: men have stronger grip and higher haemoglobin. Using the published source data, most of them weaken or disappear when only women are analysed. One link, between a heart-strain marker in the blood (NT-proBNP) and weaker grip, largely holds up, and one other partly does. The paper's suggestion that its blood results point to poor blood flow or inflammation is not supported. No healthy people were tested for comparison, and grip strength depends on effort.

Grade: 5 / 10.