ME/CFS & post-viral illness — clinical management

What landed in the literature, and what it changes at the bedside

A screened digest of new ME/CFS and post-exertional-malaise research, weighted toward diagnosis, management, and prognosis rather than bench mechanism.

Edition: 8 Nov 2026
Window: 7 days to 8 Nov
Screened: 7
Flagged: 3

Seven records, three carried, and for the first time in this digest's recent run the week is about measuring post-exertional malaise rather than about failing to. A Vienna group has published a new PEM questionnaire built with patients; a Manchester group funded by the ME Association has preprinted a second one and a companion severity scale; and a Dutch review of 105 drug trials in long COVID and ME finds that only 18% and 5% of them used any PEM measure at all, and that almost none treated a crash as a harm. That last point is the thread back to last week's null low-dose naltrexone trial, which measured generic fatigue in a PEM-defined population. The item she is most likely to be asked about is different: a 19-million-patient US health-record study saying people coded with hypermobile Ehlers–Danlos syndrome are nearly three times as likely to be diagnosed with long COVID. Its published peer-review file shows three reviewers calling its methods “deeply concerning”, and its own tables contradict two of its headline sentences; the entry below sets out what survives, which is a useful prompt to examine for hypermobility rather than a risk figure to quote. Tier 1 is empty for the fourth edition running.

Changes what you'd do

Findings with a direct line to a clinical decision — a workup step, a referral, a conversation before surgery.

Nothing this week, and the low-dose naltrexone trial is still a preprint

Tier 1 empty·7 records screened, 0 promoted·fourth consecutive empty edition

Of the three carried records, one is a retrospective health-record analysis whose authors themselves call every matched comparison “descriptive and hypothesis-generating rather than causal”; one is a review of trial outcome measures; and one is a 13-person content-validation study of a questionnaire that has not yet been tested for reliability. Each shapes how she reads or counsels, and none changes a decision on its own evidence. All three are in Tier 2 or the PEM section.

Last edition committed to promoting the 160-person low-dose naltrexone trial (medRxiv, 26 October; NCT05430152) to this tier on peer review. It has not appeared in PubMed in this window, so the commitment stands and nothing is moved. Its bottom line, stated in full on 2 November, is unchanged: sixteen weeks at up to 4.5 mg daily did no better than placebo on fatigue, pain, symptom score, step count or quality of life, and was well tolerated.

Bottom lineNothing new to act on. The closest thing to a practice point this week is in the PEM section: two instruments that put a number on post-exertional malaise are now in the literature, one peer reviewed and one preprinted, and neither is ready for routine clinical use. Until one is, a named existing PEM instrument plus a dated note of the exertion and what followed remains the best record she can keep.

Worth knowing

Shapes how you counsel, set expectations, or answer the question a patient brings in from the internet.

More common than we thought: Hypermobile Ehlers-Danlos Syndrome, comorbidity burden, and long COVID risk in a national cohort

PLOS Digit Health·Retrospective EHR cohort, propensity-matched, 25,152 vs 25,152·PMID 42986813·open access·full text and published peer-review file read; supplementary tables not opened

Five authors at the University of Colorado Anschutz, Children's Hospital Colorado and UNC Chapel Hill — the last including Melissa Haendel, a founding lead of the US National COVID Cohort Collaborative — searched that collaborative's database of 19.8 million patient records from more than 80 US institutions for anyone coded with hypermobile Ehlers–Danlos syndrome (hEDS) or with Ehlers–Danlos of unspecified type. After removing the classical and vascular subtypes and incomplete records they had 25,152 people, of whom only 7,786 carried the specific hEDS code; the other two-thirds are unspecified-type EDS counted as “suspected hEDS” on the grounds that 80–90% of EDS is hypermobile. Each was matched one-to-one to a control on age, sex, race, number of visits and contributing site.

What they report. A prevalence of 1 in 788 (0.127%; 1 in 2,549 on the specific code alone). COVID-19 recorded in 9,675 (38.4%) versus 10,370 (41.2%) of controls, which they describe as similar. Long COVID recorded in 1,008 (4.0%) versus 359 (1.4%): conditional odds ratio 2.88 (95% CI 2.55–3.26), and 2.98 (2.18–4.08) for attending a long COVID clinic. Among the hEDS patients with long COVID, 42.7% also carried a POTS code, 42.0% ME/CFS and 25.3% mast cell activation syndrome; 11.9% carried all three. People coded with POTS or MCAS were more likely to receive their hEDS diagnosis after COVID-19 (hazard ratios 2.95 and 2.06); ME/CFS was not (HR 1.08, 95% CI 0.97–1.19). The hEDS group also had more autoimmune, rheumatological and chronic lung diagnoses, more depression and more systemic steroid use.

Bottom lineTwo things to take from it, and one to leave. Take the prompt. In a woman with long COVID or ME/CFS who also has POTS or a mast-cell story, asking about joint hypermobility and examining for it is cheap, and this paper adds to a body of clinical observation that the three travel together; the diagnosis changes physiotherapy, surgical and obstetric conversations even if it changes nothing about the post-viral illness. Take the ME/CFS null. On the paper's own analysis, an ME/CFS code did not make a post-COVID hEDS diagnosis more likely once POTS and MCAS were accounted for — so it offers no support to the idea that ME/CFS is an unrecognised connective-tissue disorder, which some patients will have read. Leave the “three times” figure. It is the ratio of two diagnosis codes in people who use healthcare very differently, and the paper's own data argue that most of it is about who gets coded rather than who gets ill: the hEDS group had significantly fewer confirmed infections, fewer positive PCR tests, less pneumonia and less than half the odds of death of their matched controls, while being three times as likely to be labelled a “possible” COVID case. A patient with hEDS who asks whether she is at triple risk can be told, accurately, that a large records study found more long COVID diagnoses among people like her, that its authors describe the finding as hypothesis-generating, and that nobody checked a single chart.

Weigh it accordingly. A diagnosis-code study with no chart validation of any diagnosis — not hEDS, not long COVID, not ME/CFS — which the authors name as their first limitation, along with misclassification, differences in coding between institutions and residual confounding by healthcare use. Two-thirds of the cohort never received an hEDS code, and the inference from “unspecified EDS” to hEDS is a population proportion applied to individuals. The published peer-review file is worth reading in its own right. The handling editor summarised the first round as finding “some deeply concerning methodological flaws”, with all three reviewers raising mislabelled data, bias, confounding and changed healthcare contact during the pandemic, and the conclusions “stated far too strongly as if the analysis proves causation”; one reviewer said the manuscript did not meet the journal's criteria and the statistics were not appropriate. The editor asked specifically for a chart-review sub-study of 100–200 randomly selected hEDS cases; the published version does not include one, and instead adds a limitations paragraph calling all matched findings “descriptive and hypothesis-generating rather than causal”. The same review flagged 154 hEDS patients coded with vertebral tuberculosis against fewer than 20 controls, which the authors now attribute to miscoding and have removed. The abstract was not brought into line with that paragraph: it still says the findings “establish” hEDS as more prevalent and that patients were “nearly three times more likely to develop Long COVID”. Three internal inconsistencies. First, “similar” infection rates sit beside an odds ratio for confirmed COVID-19 of 0.89 (0.86–0.92), which is a statistically significant deficit, not similarity. Second, the caption of Table 3 says 27% of patients were diagnosed with hEDS after infection “including 10% within six months”, but the table lists 2,547 (27%) diagnosed more than six months after and 933 (10%) within six months as separate rows: the after-infection share is 37%, and 2,547 + 933 + 5,965 = 9,445 confirms it. Third, the abstract says long COVID risk “was concentrated among” those with POTS, MCAS or ME/CFS, but no analysis compares comorbidity-stratified long COVID risk against controls; what was shown is that comorbidities are commoner among hEDS patients who have long COVID than among hEDS patients overall, which is the reverse question. Beyond that, the encounter-adjusted analysis (Fig 4) found no change in the rate of hEDS diagnosis per visit by COVID or long COVID status, which the authors fairly concede leaves “COVID unmasks hEDS” and “sicker people see more doctors” indistinguishable. One figure to treat with suspicion: about 4% of the matched controls carry an ME/CFS code, many times any population estimate, which suggests either a broad concept set (the code list is in a supplementary table this digest did not open) or controls selected, by matching on visit counts, from very heavy users of care. Funding: an NIH training grant to the first author, the Ehlers-Danlos Society — a patient charity with an interest in the condition being recognised as common — and three family foundations, all to the same senior author, plus UNC departmental funds; the authors declare no competing interests. Received in 2025, first decision 28 October 2025, published 5 November 2026. Arithmetic that checks: 7,786 + 24,031 − 5,709 dual-coded = 26,108, less 916 = 25,152; 25,152 of 19.8 million is 0.127%; 1,008 and 359 of 25,152 are 4.0% and 1.4%; 82 of 359 is the 22.8% of control long COVID patients coded with ME/CFS. The 38.4% infection figure is 38.47% truncated rather than rounded, which is immaterial.

Outcome measures in long COVID clinical trials: a review

Front Public Health·Structured review of 105 trials (68 long COVID, 37 ME)·PMID 42973774·open access·full text read; supplementary trial lists not opened

Nine Dutch authors — from UMC Utrecht, Erasmus MC, Radboud, Maastricht, Twente and Groningen, with two listed as an independent researcher and a research physician in Amsterdam — searched PubMed and ClinicalTrials.gov for trials of biomedical interventions in long COVID and ME (excluding psychological, behavioural, exercise and supplement trials), hand-searched the references of systematic reviews, and catalogued every outcome measure used. They found 68 long COVID and 37 ME trials, using 261 and 175 distinct measures respectively. More than 54% of the patient-reported measures predominantly measure fatigue. Measures specific to PEM were used in 18% of long COVID trials and 5% of ME trials, and all of them were variants of the DePaul Symptom Questionnaire PEM subscale, which the authors criticise for ignoring delayed onset, centring on fatigue and being confounded by pacing — a patient who paces well reports few crashes. Only single-day CPET was used, in 4% of trials. No trial measured return to work.

Its argument about harm is the part she will use. Because PEM can cause lasting decline, “every trial that requires exertion on the part of the participant can lead to harm. This includes factors such as talking, filling out questionnaires, or traveling to the study location.” A prolonged crash can meet the formal definition of a serious adverse event, yet PEM is “not yet routinely implemented in safety protocols or termination criteria”. The authors recommend recording the severity and duration of every PEM episode possibly related to a trial, continuing to measure after the intervention ends, building functional capacity into stopping rules, and not using fatigue scales as the sole primary outcome or mental-health scales whose items are physical symptoms of the illness (the PHQ-9's tiredness, sleep and appetite items, for example). They name the Chalder Fatigue Scale and the Checklist Individual Strength as instruments built on a psychosomatic model, report ceiling and floor effects for the Chalder scale, the CIS, the Fatigue Severity Scale and the SF-36, and single out FUNCAP as the only instrument validated against the International Consensus Criteria severity scale.

Bottom lineThis is the reference to have when a patient asks whether a trial is safe to join, or why the trial she read about found nothing. On safety: the honest answer is that most trials in this illness do not look for the harm she is most at risk of, so “no serious adverse events” in a published trial usually means a crash was not counted, not that none happened — the same point this digest made about the hyperbaric oxygen cohort and the exercise meta-analysis last week, now made by a review of the whole field. On null results: a trial that measured fatigue with a scale already near its ceiling, in people who pace, can miss a real effect on what patients actually experience — which is exactly the open question left by last week's naltrexone trial, whose primary outcome was the Fatigue Severity Scale. That cuts both ways and should be said so: the same argument cannot be used to rescue a drug a patient likes and dismiss one she does not. For her own records: the recommendations double as a clinic checklist — a symptom inventory plus a construct-specific measure, a functional measure that reaches the severe end (FUNCAP is the one they recommend), and a dated note of any crash after an intervention she starts.

Weigh it accordingly. A structured narrative review, not a systematic one: two searchers and two databases, but no search dates, no search strings, no flow diagram and no appraisal of the trials themselves are given in the text, and the trial lists are in supplementary tables this digest did not open. Its recommendations are expert opinion built on that catalogue. Counts that do not reconcile: the abstract gives 397 outcome measures in total while the body gives 261 + 175 = 436; the difference would be explained by 39 measures used in both fields, but that is not stated. Several statements are asserted rather than evidenced in the text read — that long COVID affects “approximately 10%” of people infected, that it overlaps with ME “in 50% of cases”, and that the false-illness-belief theory “has since been disproven” — and the review takes a clear position in the field's central dispute, which a reader should know before quoting it. One citation looks wrong: the claim that pacing effectiveness was “recently shown” to correlate with higher SF-36 scores cites the 2021 paper that developed the DePaul PEM questionnaire, which on its title is about something else; this digest did not read that paper and records the mismatch rather than a finding. The one outcome figure on hyperbaric oxygen is the senior author's own work, and second-hand here: a Dutch HBOT registry for long COVID (Sci Rep 2025, Hellemons a co-author) in which, as the review reports it, 56% improved, 31% did not respond and 13% were worse on the SF-36 at three months, and 4% dropped out because their symptoms worsened. The percentages sum to 100; the registry itself was not read. No funding was received and the authors declare no conflicts; no generative AI was used. Published 19 October 2026, entered PubMed 3 November.

PEM and pain

A standing section, carried every edition regardless of evidence grade, with what each item actually rests on stated up front. This week PEM is the subject of three instruments — one peer reviewed, two preprinted — and of a review that measures how rarely trials record it. It was administered as an outcome in no clinical study. Pain was measured only as one item of a symptom checklist, in a study this digest excluded.

Development and initial content validation of the Vienna post-exertional malaise questionnaire: an iterative mixed-methods study

Qual Life Res·Instrument development, 13 patients + 11 experts·PMID 42998183·open access but full text could not be read·abstract and declarations only

Read the meta line first. The article is open access at Springer, but the publisher served a JavaScript “client challenge” to this digest's request for both the page and the PDF; that is a bot check, so it was not pursued, and nothing beyond the MEDLINE abstract and declarations was read. There is no PubMed Central copy yet.

Eight authors at the Medical University of Vienna's Center for Public Health, with a neurological rehabilitation centre in Vienna, the University of Vienna's German studies department and an exercise physiologist at the Vrije Universiteit Amsterdam, built the Vienna Post-Exertional Malaise (VPEM) Questionnaire to measure four things: functional decline, recovery, crash-associated symptoms and pacing adaptations. A 13-item draft drawn from existing PEM, fatigue and functional-capacity instruments was rated and discussed by twelve people with ME/CFS and one with post-COVID syndrome meeting ME/CFS criteria, through ratings, written feedback and interviews; they approved its length (92.3%) and terminology (76.9%) but were divided on completeness and clarity, and identified three problems: telling PEM apart from ordinary fatigue, missing triggers, and the way PEM varies within one person. After revision, an expert panel of eleven rated its PEM coverage complete or largely complete (100%) and its structure suitable (90.9%), with item relevance high (S-CVI/Ave 0.96) and clarity more variable (0.82). The authors conclude only that content and face validity are supported and that psychometric validation is next.

Bottom lineNot for use yet, and worth knowing about for what it is trying to fix. Its four domains — how far function fell, how long recovery took, what the crash felt like, and how the patient has had to pace — are the four questions a careful PEM history already asks, and the last one matters most: it counts pacing as part of the burden instead of letting a well-paced patient score as mildly affected, which is the confounding the Dutch review above identifies in the DePaul questionnaire. Until reliability and responsiveness are published, it is a structure for her own history-taking rather than a score to record.

Weigh it accordingly. The full text was not read, so the revised item count, the response scale, the recall period, how the 13 patients were recruited, how ill they were and which ME/CFS criteria they met are all unknown. Content validation in 13 patients and 11 experts establishes that the items look right to those people; it says nothing about reliability, responsiveness to change or whether the score distinguishes PEM from other fatigue, all of which remain to be tested. The arithmetic of the abstract checks: 92.3% and 76.9% are 12 and 10 of 13, and 100% and 90.9% are 11 and 10 of 11. Patients took part in refining the items but not in generating them, analysing the data or planning dissemination, which the declaration states plainly. Ethics: Medical University of Vienna, 1192/2025, with written consent. ChatGPT was used for language editing, declared. The authors declare no competing interests; no funder appears in the record.

PEM: three new instruments in one week, and still no trial that used one

Standing check·2 of 7 PubMed abstracts mention PEM; 1 develops a PEM instrument; 0 administered one as an outcome

The counts. All seven PubMed records carry an abstract. PEM is named in two: the Vienna questionnaire and the Dutch outcome-measures review. A third, the symptom-network study excluded below, names it only in its full text's clinical-implications section, recommending that PEM be assessed although the study did not measure it; the TRPV1 case report's full text says ME/CFS symptoms “become markedly worse following exercise” without naming PEM. The CDC infection-associated-illness review, the hEDS paper and the TSPO preprint do not mention it. No study in the window administered a PEM instrument to patients as an outcome. On medRxiv, the post-exertional search returned a hit for the first time in six windows — ending a five-window drought — and it is an instrument: the Manchester Post Activity Symptom Scale, carried in Preprint watch with its sister severity scale.

What the three instruments have in common is the problem they are trying to solve. The Vienna questionnaire includes pacing adaptations as a domain. The Manchester PASS, in its Rasch analysis, found that its item on delay in PEM onset did not fit the rest of the scale and dropped it — worth pausing on, because delayed onset is part of how the illness is defined. And the Dutch review faults the existing DePaul subscale for ignoring delay and for being suppressed by successful pacing. So the field now has three groups converging on the same two hard features of PEM — that it arrives late, and that patients hide it by avoiding its triggers — and handling them in different ways, one by measuring them and one by finding that a single question on delay does not behave like the others.

Bottom lineA better week than any in this digest's recent run, and still not one that changes practice. None of the three instruments has published responsiveness data, one is unreviewed, and the reviewed one has only been content-validated. The practice point stays: a named PEM instrument in the record (the DePaul subscale remains the one with a literature behind it, with its known blind spots), a dated note of the specific exertion and what followed, and — newly supported by both the Vienna domains and the Dutch review — a line on how much she is restricting activity to avoid crashes, because a low symptom score in a patient who has given up most of her life is not mild illness. When a patient says a questionnaire does not capture her PEM because it arrives two days later, the Manchester finding is a reason to take that seriously rather than a reason to doubt her.

Pain: one checklist item, in an excluded study

Standing check·0 of 7 PubMed records administered a pain instrument; 1 scored pain as a symptom-checklist item

No record in the window measured pain with a pain instrument. The nearest is the Hangzhou symptom-network study excluded below, which scored pain as one item of an adapted Memorial Symptom Assessment Scale in 214 outpatients and found it the second-strongest “bridge” symptom linking clusters (bridge expected influence 0.363, after fatigue at 0.430); that is a statistical property of one cross-sectional correlation network in a Fukuda-defined sample, and it says nothing about what to do. The TRPV1 case report, also excluded, describes one woman with ME/CFS, sensory polyneuropathy and marked hypersensitivity to heat, chilli and alcohol carrying a rare TRPV1 variant the authors judge likely pathogenic; it is a hypothesis about one channel in one patient, without functional testing.

Bottom lineNothing new on pain. The controlled result remains last week's: in the low-dose naltrexone trial, pain on a 0–10 scale moved by 0.03 (95% CI −0.50 to 0.57) against placebo, in a sample recruited for fatigue. The Leeds and Manchester musculoskeletal-pain preprint, the Brussels finding on pain distribution and next-day exertional response, and the endometriosis prompt all still stand from previous editions and are not repeated. One small observation from the TRPV1 report is worth having in reserve: a patient who reports that heat, spicy food and alcohol all reliably worsen her pain is describing three agonists of one receptor, and while that proves nothing, it is a more specific history than “sensitive to everything”.

In the pipeline

Registered trials and protocols in view this window — nothing to act on, but these are the results that will arrive next. For the fourth week running no PubMed record was itself a protocol or registration; the new entries are next steps stated by this week's authors, none of them registered.

Psychometric validation of the Vienna PEM questionnaire, Medical University of Vienna. Stated as the next step in the paper's conclusion; no design, sample, registration or timeline in the record. What will make it easy to apply: a questionnaire that scores pacing restriction alongside crashes would finally let a trial show benefit as “able to do more for the same cost”, which is how patients describe improvement. What will make it hard: the development sample was thirteen people, so the validation sample has to reach the severe end, and a questionnaire long enough to cover four domains is itself an exertion for the people it most needs to reach.
The Manchester PASS, SoMES and MEOSS scales, University of Manchester and the ME Association. Two preprints this week, from the same three authors; the ME Association both funds the work and employs one author. What will make them easy to apply: a severity scale and an occupational-status scale filled in by 995 people online, retested by 579, are cheap enough for routine clinic use, and the PASS was co-produced with patient and clinician advisory groups. What will make them hard: an online, self-selected sample cannot represent the housebound who do not complete surveys, the severity scale's reliability is moderate by the authors' own description, and none has yet been used to detect change in a treatment trial.
Chart validation of hEDS coding in the US national COVID cohort. Requested by the editor of the hEDS paper as a 100–200-case sub-study and not done; the authors list chart-level validation and a comparator cohort without connective-tissue disease as the work required to separate “COVID unmasks hEDS” from ascertainment. Not registered or scheduled. What will make it easy to apply: if even a modest share of the coded cases meet the 2017 criteria, the prompt to examine for hypermobility in post-viral patients with POTS and MCAS will rest on more than codes. What will make it hard: the database is date-shifted and de-identified, so chart review requires going back through each contributing site.
Still standing from previous editions. NCT05430152, the low-dose naltrexone trial: preprinted 26 October, not yet published, with its registered responder analysis and fifteen further outcomes unreported (2 Nov). A sham-controlled trial of hyperbaric oxygen in ME/CFS: called for, and still not registered anywhere in view (2 Nov). The exertion-paired EEG programme for the post-exertional trajectory, unregistered (2 Nov). Enterovirus persistence in tissue rather than blood, Cornell (2 Nov). VITA Monitoring, the 212-person FIOCRUZ post-COVID cohort (26 Oct). Reservoir-directed and biologically stratified long COVID trials as the ERJ guideline revision frames them, with the reply letter at doi 10.1183/13993003.01810-2026 still unread here (26 Oct). The Alberta vaccination-as-immune-perturbation proposal, and the rs6971 genotyping question for TSPO imaging in Birmingham, whose preprint entered PubMed this week and is not re-carried (26 Oct). The rTMS RCT in PASC (ZonMw 11080022420018), still without a registration identifier (19 Oct). The COMET core outcome set for CFS/ME (study 2103), registered since 2022 and unreported (19 Oct). “ME/CFS Kids Bavaria”, and Germany's National Decade against Post-Infectious Diseases (19 Oct). NCT05498493, Mount Sinai's cognitive rehabilitation trial in long COVID (12 Oct). APAV-ME/CFS parts 6 and 7. The COVID-CNS connectivity secondary analysis (1 Oct). The McMaster qualitative protocol on mind–body approaches (25 Sep). ChiCTR2400082268, the Shanghai Tai Chi trial's unreported outcomes (28 Sep). The Munich adolescent stand-test replication (21 Sep). NCT04740736, Mount Sinai's two-day-CPET saline-versus-sham infusion study, completed July 2025 and unreported (14 Sep). RECOVER-VITAL's viral-persistence analyses (7 Sep). The two Tuina protocols for CFS (31 Aug; 7 Sep). REenergizeME, ADDRESS-LC, the ME/CFS rehabilitation-subgroup study and the auricular vagal neuromodulation trial in POTS (16 Aug). The decentralised tirzepatide phase 2 in long COVID fatigue (24 Aug).

Preprint watch

medRxiv, 2 – 8 Nov. Four searches returned four hits comprising three distinct postings; two are relevant and both are covered, and one was off-topic. Neither relevant posting is peer reviewed; they are listed so nothing arrives as a surprise. Industry funding: none. Both are funded by a patient charity that employs one of their authors.

The development and psychometric evaluation of the Post Activity Symptom Scale (PASS) for Myalgic Encephalomyelitis/Chronic Fatigue Syndrome

medRxiv preprint·Not peer reviewed·Online survey with two-week retest, Rasch analysis·posted 8 Nov 2026·funded by the ME Association·abstract and declarations only

Mike Horton (Leeds), Sarah Tyson (Manchester) and Russell Fleming (ME Association) co-produced a PEM scale with patient and clinician advisory groups, piloted it, and then had adults with ME/CFS complete it online twice, two weeks apart. Rasch analysis tested each domain. The item on delay in PEM onset misfitted — under-discriminating and inconsistent with the other items — and was removed, after which the scale was unidimensional. Reported: criterion validity against measures of physical health and ME/CFS severity r = 0.58; discriminant validity p < 0.001; test–retest ICC 0.81; minimal detectable difference 1.93 on a 0–100 scale. Participants found it easy, comprehensive and relevant.

Bottom lineThe first scale in this digest's run built specifically to measure PEM and tested for reliability, and worth watching for that reason. Its most interesting finding is a negative one: a question about how long PEM takes to arrive did not behave like the other questions, which most likely means delay is a separate dimension of PEM rather than a component of its severity — not that delay does not matter. Until it is peer reviewed and its sample is described, it is not for clinical use.

Weigh it accordingly. Preprint, not peer reviewed; medRxiv's notice applies. This entry rests on the abstract and the posted declarations; no full text, table or supplement was opened. The sample size is not given in the abstract, nor the case definition, severity mix or recruitment route, so who completed it is unknown; an online survey with retest selects for people well enough to do both. The minimal detectable difference is hard to reconcile with the reliability figure. On the usual formula, MDD = 1.96 × √2 × SD × √(1 − ICC), an ICC of 0.81 gives MDD ≈ 1.21 × SD, so an MDD of 1.93 implies a between-person standard deviation of about 1.6 points on a 0–100 scale — implausibly narrow for a severity measure. Either the scale range, the units (Rasch logits rescaled) or the formula differs from what the abstract implies, and the full text must be checked before the figure is quoted. Interests: funded by the ME Association; the authors declare no competing interests “but note that Mr Fleming is employed by the ME Association”. A patient charity has an interest in an instrument that shows PEM clearly, recorded here as this digest records aligned interests in both directions. Ethics: University of Manchester Faculty of Biology, Medicine and Health research ethics committee. Data on reasonable request. CC-BY 4.0.

Development and psychometric evaluation of the Severity of Myalgic Encephalomyelitis Scale (SoMES) and the Myalgic Encephalomyelitis Occupational Status Scale (MEOSS)

medRxiv preprint·Not peer reviewed·Online survey, n = 995, retest n = 579·posted 5 Nov 2026·funded by the ME Association·abstract and declarations only

Tyson and Fleming co-produced two brief scales from the literature and lived experience — one for disease severity, one for occupational status — and tested them in 995 adults with ME/CFS online, 579 of whom repeated the survey two weeks later. Against SF-36 physical functioning, correlation was r = 0.663 for severity and r = 0.472 for occupational status; physical functioning differed significantly across levels of both (p = 0.003 to 0.0001). Test–retest agreement was moderate for the severity scale (weighted κ 0.53), with a minimal detectable difference of 1.6, and excellent for the occupational scale (κ 0.76; MDD 1.4).

Bottom lineA short severity scale would be genuinely useful in clinic and in benefit and insurance paperwork, where “mild / moderate / severe” is usually asserted rather than measured. This one is not there yet: a weighted kappa of 0.53 means the same patient, two weeks apart, often lands on a different level, which matters for a scale meant to track change. The occupational scale is the more reliable of the two and is worth a look when the paper is published.

Weigh it accordingly. Preprint, not peer reviewed; abstract and declarations only. The conclusion calls both scales “valid, reliable, sensitive”, but the abstract's own word for the severity scale's reliability is moderate, and responsiveness to change — what “sensitive” would require — was not tested in a two-week retest of stable patients. The scale ranges are not given, so the MDDs of 1.6 and 1.4 cannot be interpreted. 579 of 995 completing the retest is 58.2%, and nothing is said about whether the 416 who did not differed. “Construct validity was confirmed from feedback” is a statement about acceptability, not construct validity. Case definition, severity distribution and recruitment route are not stated. Interests: the funder is the ME Association, which employs the second author; the authors declare no competing interest. medRxiv's structured funder field reads “NA” while the posting's own metadata names the ME Association, so a reader relying on the listing would miss it. Ethics: University of Manchester. CC-BY 4.0.

Four searches, four hits, three distinct postings, two carried. The myalgic encephalomyelitis phrase returned two, the PASS and the SoMES/MEOSS postings. The post-exertional search returned one, the PASS again — its first non-empty window after five consecutive empty ones, ending the longest run in this digest's record. The long COVID phrase returned zero, for the first time in this digest's recent run. The post-COVID search returned one, a German respiratory-disease forecasting paper whose abstract mentions “post-COVID-19 times”; off-topic and not carried. Both relevant postings were read at abstract level with their competing-interest, funding, ethics, consent, registration and reporting-guideline declarations. Dates: SoMES DOI date 4 November, posted 5 November; PASS DOI date 6 November, posted 8 November — both inside the window. One structural note: both carried preprints are from the same group and the same funder in the same week, and together with the Vienna paper that makes three PEM-measurement papers from three groups; that is a cluster worth noticing, and it is also a reminder that a week's shelf can be shaped by one research programme's publication schedule.

What was screened out, and why

Four of the seven PubMed records didn't make the cut — a 57% exclusion rate, the highest in this digest's run, though one of the four is a preprint already covered on 26 October that has now been indexed, so on substance three of six were excluded. The grounds, so the filter stays auditable:

Already covered — not re-carried — Fibromyalgia, Chronic Fatigue Syndrome, and Multiple Sclerosis are associated with lower translocator protein binding affinity genotypes (medRxiv preprint, indexed through PubMed Central). Jones and Younger genotyped rs6971 by Sanger sequencing in 86 people (30 fibromyalgia, 21 ME/CFS, 11 MS, 24 controls) and report more mixed- and low-affinity genotypes in all three illnesses, and associations with fatigue and depression scores. The posting is dated 21 October and so fell in the 26 October edition's window; the “rs6971 genotyping question for TSPO imaging in Birmingham” on the standing pipeline list, and the “genotype study” that edition carried in Preprint watch, are this paper. That edition is not in the local archive, so the identification is inferred from the date, the genotype, the authors' institution and the standing-list wording rather than checked against the old text. It is counted here so the PubMed total reconciles, and not carried again. The practical point it raised then still stands: TSPO-PET studies that do not genotype for rs6971 cannot be compared across groups if the groups differ in genotype.
Narrative review, no data — A New Outlook on Infection-Associated Chronic Conditions and Illnesses (Open Forum Infect Dis 2026 Nov). Eleven CDC authors — two of them, Jeanne Bertolli and Elizabeth Unger, listed as retired — argue that ME/CFS, post-treatment Lyme disease syndrome, dysautonomia and long COVID belong together as infection-associated chronic conditions and illnesses, that the conditions lacking objective tests are “underrecognized and often stigmatized”, and that the multi-agency long COVID response holds lessons for future outbreaks. Out on design: a narrative review with no search, no appraisal and no data. PubMed Central holds only the abstract and front matter as text (the body is PDF-only) and the publisher's page was not tried, so only the abstract was read. Recorded because it is the US public-health agency formally adopting the umbrella term — which patients and advocacy groups use, and which she will increasingly see on referral letters — and because the retirement of the agency's long-standing ME/CFS lead is a fact about who will be writing the next federal guidance. One author declares IDSA travel support; the rest none.
Case definition and measurement unsuited to the question — Symptom clusters and symptom network characteristics in chronic fatigue syndrome patients: a cross-sectional study (Front Med 2026;13:1955021). Six nursing authors at Zhejiang Chinese Medical University surveyed 214 patients recruited by convenience from a traditional Chinese medicine internal-medicine clinic between September 2025 and April 2026, using an adapted Memorial Symptom Assessment Scale (a cancer-symptom instrument), and report six symptom clusters with fatigue the most central symptom (expected influence 1.541) and fatigue, pain and poor sleep the main bridges. Out because participants were defined by the 1994 Fukuda criteria, under which PEM is optional, PEM was not measured, and a sample selected for fatigue that then finds fatigue present in 100% and most central is close to circular; the recommendation to prioritise fatigue does not follow from a correlation network. The paper also contradicts itself on method: the abstract and results describe principal component analysis with varimax rotation, while the statistics section describes principal-axis factoring with promax rotation — different techniques. Arithmetic that checks: 22 retained symptoms give the 231 pairs screened for redundancy. Funded by the Zhejiang provincial health commission; no conflicts declared.
Single case report, genetic hypothesis — A Woman With Multiple Neuropathic Symptoms, Hypersensitivity to Heat, Peppers, and Alcohol, With a Variant in the TRPV1 Capsaicin Receptor Gene: Possible Implications to ME/CFS (Case Rep Genet 2026). Two authors at a private mitochondrial-medicine practice in Pasadena describe a 65-year-old medical professor, healthy until 57, whose orthostatic intolerance began three days after an influenza vaccination and was followed by ME/CFS, migraine, functional gut symptoms and a mild sensory polyneuropathy; tilt testing produced asystole within 35 seconds requiring CPR, and a pacemaker was fitted but has never activated. Reanalysis of a commercial genome that had been reported negative found a very rare heterozygous TRPV1 variant (p.Lys725Met), which the authors class as likely pathogenic and gain-of-function on the strength of her history. Out because it is one patient, the variant's function was not tested, and the authors say causality “cannot be definitively established”; the ME/CFS link rests on a rat forced-swim model. The pain observation it offers is in the PEM and pain section. No funding, no conflicts declared; written consent obtained.

Three borderline calls worth naming. The first is carrying the hEDS paper at all. It is a coding study whose own reviewers called its methods deeply concerning, and on design it is weaker than several things this digest has excluded. It is in because it will be quoted — “three times the risk” is exactly the kind of figure that reaches patients — and because its useful content, the prompt to examine for hypermobility and the ME/CFS null, survives its flaws; the entry is written to arm her against the headline rather than to pass it on. A reader who thought it should have been excluded and named here would not be unreasonable. The second is the Dutch review, which is closer to a narrative review than to a systematic one and takes a side in the field's central dispute. It is carried, where the CDC review and last week's Stanford perspective were not, because it rests on a stated search and a count of 105 trials and because its subject — how trials measure harm — bears directly on what she tells a patient about joining one. The third is the Vienna paper, carried with its full text unread because the publisher served a bot check; it is in on the strength of an abstract that reports its own numbers in full and makes a modest claim, and every statement about it above is drawn from that abstract.

How this was assembled

PubMed was queried directly through the NCBI E-utilities this week, not through a browser — this edition is the first produced on a machine where that interface is reachable. An entry-date search for 2 to 8 November returned a count of seven and seven PMIDs, and the full MEDLINE records fetched for those PMIDs contained seven records, so nothing was truncated. No bot-check interstitial arises on that interface. Entry dates ran from 3 November at 12:05 to 8 November at 00:48; nothing entered on 2 November. The boundary was checked by re-running the query for 26 October to 1 November, which returned exactly the seven PMIDs carried or excluded in the 2 November edition and no others, so nothing fell between the two windows and nothing is double-counted. Full MEDLINE records — abstracts, affiliations, publication types, grant numbers and competing-interest statements — were read for all seven.

Read end to end from PubMed Central: the hEDS paper, including its methods, results, discussion, limitations, funding, data-availability statement and the published peer-review file appended to it, which is where the editor's summary, the requested chart-review sub-study and the vertebral-tuberculosis exchange were found; and the Dutch outcome-measures review, including its methods, its sections on PEM and safety, its recommendations table, its declarations and the specific references checked (the HBOT registry and the DePaul development paper). The supplementary tables of both — the hEDS code lists and the review's trial lists — were not opened. Read in part: the symptom-network study (methods, results, limitations and implications, enough to establish the case definition and the method contradiction) and the TRPV1 case report (case description, declarations). Read at abstract level only: the CDC review, whose PubMed Central record carries no body text, and the TSPO preprint, which is not re-carried.

One carried record could not be read beyond its abstract. The Vienna PEM questionnaire is open access at Springer, but both the article page and the PDF returned a JavaScript “client challenge” to this digest's requests. That is bot detection; it was not pursued, and the entry says so in its meta line, its opening paragraph and its caveat.

Where a paper's own numbers or statements disagree, the disagreement is named. In the hEDS paper: “similar” infection rates beside a significantly reduced odds ratio of 0.89; a Table 3 caption that puts the after-infection share at 27% “including 10%” when the rows give 37%; an abstract claim about comorbidity-concentrated risk that no analysis tests; an abstract still saying “establish” after the limitations were rewritten as hypothesis-generating; and an ME/CFS code rate of about 4% among controls. In the Dutch review: 397 outcome measures in the abstract against 261 + 175 = 436 in the body, and a citation that does not appear to support the sentence it follows. In the symptom-network study: two different factor-analytic methods described for the same analysis. In the PASS preprint: a minimal detectable difference of 1.93 on a 0–100 scale that, with an ICC of 0.81, implies a standard deviation of about 1.6 points. In the SoMES preprint: a reliability called moderate in the results and “reliable” in the conclusion, and a funder recorded as “NA” in medRxiv's listing and named in the posting. Arithmetic that checked out is also recorded: the hEDS cohort counts (7,786 + 24,031 − 5,709 = 26,108; − 916 = 25,152), its prevalence (25,152 of 19.8 million = 0.127%), its long COVID proportions (4.0% and 1.4%), its timing table (2,547 + 933 + 5,965 = 9,445) and its 82-of-359 ME/CFS figure; the Vienna percentages (12/13, 10/13, 11/11, 10/11); the HBOT registry split reported in the review (56 + 31 + 13 = 100); the SoMES retest rate (579/995 = 58.2%); and the network study's 231 symptom pairs from 22 symptoms.

Each finding carries its study design, sample size and the authors' own stated limitations. Post-exertional malaise and pain carry standing extra weight: this week PEM was named in two of seven PubMed abstracts and was the subject of three new instruments, and was administered as an outcome in no study; pain was scored only as a checklist item in an excluded study. Nothing is filtered for a positive result: the hEDS entry spends more of its length on why its headline should not be quoted than on the headline, and the most practically useful result in it is a null. Commercial, charitable and institutional interests are stated in each entry: a connective-tissue patient charity and family foundations funding a paper about that condition's prevalence; a patient charity funding, and employing an author of, two instrument preprints; a review whose only outcome figure on hyperbaric oxygen is its senior author's own registry; a provincial health commission funding a traditional-medicine clinic study; and a private practice reporting its own patient. One declared use of a language model (ChatGPT, for language editing, in the Vienna paper) and one declared non-use (the Dutch review) are recorded without comment beyond the observation that this digest is written by one. Items from previous editions are referred to, not recycled: the low-dose naltrexone trial, the HBOT cohort, the exercise meta-analysis, the Leeds pain and Brussels comparison findings, the endometriosis prompt and the TSPO genotype preprint are all referenced and none is re-carried.

The standing query, with an entry-date range appended:

("Fatigue Syndrome, Chronic"[MeSH] OR "myalgic encephalomyelitis"[tiab]
 OR "chronic fatigue syndrome"[tiab] OR "ME/CFS"[tiab]
 OR "systemic exertion intolerance"[tiab] OR "post-exertional malaise"[tiab])

medRxiv was searched through its public API rather than its search page, which returned an HTTP 403 to a scripted request. Every medRxiv posting dated 2 to 8 November — 416 in all — was retrieved and filtered on title and abstract for myalgic encephalomyelitis, post-exertional, long covid and post-covid, which is the equivalent of the four standing searches; a broader filter (long-COVID variants, post-acute, PASC, chronic fatigue, ME/CFS, post-viral, infection-associated) found nothing further. The four searches returned two, one, zero and one hits — three distinct postings, one off-topic. For the two relevant postings, the source XML was read for the abstract, author affiliations, funder, competing-interest statement and the ethics, consent, registration and reporting-guideline declarations; no preprint full text or PDF was opened, and both entries say so.

No trial registry was consulted this week, because no carried item depends on a registered trial. Every title links to its PubMed record or, for preprints, to the medRxiv abstract page. Previous editions are archived alongside this file. Nothing here is treatment advice; it is a literature digest for a clinician.