Worth knowing
Shapes how you counsel, set expectations, or answer the question a patient brings in from the internet.
J Psychosom Res·Letter — no abstract, no original data·PMID 42962334·paywalled — full text could not be read·MEDLINE record and declarations only
Read this heading before the entry. Two authors at the First Teaching Hospital of Tianjin University of Traditional Chinese Medicine have published a letter whose title names three specific methodological objections to the interpretation of exercise effects in ME/CFS. The letter itself runs to a single page behind Elsevier, has no PubMed Central copy, and no abstract in MEDLINE. Its DOI resolves to ScienceDirect, which served a human-verification challenge; this digest does not complete those, so nothing of the letter's content was read. What follows is the record, the inference about what it targets, and the target's own published numbers — clearly separated.
What the record establishes. Publication type Letter. Received 17 September 2026, revised 30 September, accepted 2 October, published online 16 October, entered PubMed 30 October. The authors declare no competing financial interests or personal relationships. The three concerns are, verbatim from the title: trial independence, safety ascertainment, and exploratory moderator analyses.
What it almost certainly targets, flagged as inference. The record does not name the paper commented on. Searching the same journal for 2026 ME/CFS and chronic-fatigue records returns one candidate onto which all three objections map exactly: Zhao, Gou and Zhang, “The efficacy of exercise in patients with myalgic encephalomyelitis/chronic fatigue syndrome: a systematic review and meta-analysis” (J Psychosom Res 2026 Aug;207:112677, PMID 42000527). That paper searched seven databases to December 2025, pooled 17 RCTs (n = 1944), and reported exercise improving fatigue (16 RCTs, n = 1673; SMD 0.85, 95% CI 0.65 to 1.06, I² = 85.3%), sleep quality (5 RCTs, n = 806; SMD 0.42, 0.31 to 0.53, I² = 41.2%), quality of life (11 RCTs, n = 1171; SMD 0.99, 0.73 to 1.26, I² = 89.7%), depression (9 RCTs, n = 1279; SMD 0.56, 0.43 to 0.69, I² = 45.8%) and anxiety (9 RCTs, n = 1279; SMD 0.14, 0.04 to 0.24, I² = 21.2%), with no significant effect on cardiopulmonary function (3 RCTs, n = 150; SMD 0.13, 95% CI −0.11 to 0.43, I² = 0.0%) or functional capacity (4 RCTs, n = 459; SMD 0.09, −0.21 to 0.38, I² = 47.4%). Its own stated conclusions include that results “may be influenced by potential moderators such as intervention length, frequency, session duration, and medication use” and, critically, that “there was no significant difference in acceptability or adverse events between two groups.” Those are, respectively, the exploratory-moderator and safety-ascertainment claims a letter with this title would be written to contest. The alternative candidate is the Bayesian network meta-analysis of 57 studies / 61 RCTs / 4204 participants in the same journal (J Psychosom Res 2026 Jul;206:112618, PMID 41861756), which pools ME/CFS with fibromyalgia on an explicitly “symptom-oriented perspective” and ranks high-intensity interval training highest for quality of life (SUCRA 94.6% on SF-36, 98.9% on the FIQ) — but that paper used meta-regression rather than exploratory moderators and makes no adverse-event claim, so it fits the title less well.
Bottom lineThe useful thing is not the letter, which she cannot read either. It is knowing that these two papers exist in a mainstream psychosomatic journal, published five and four months ago, and that the first of them is now the subject of a formal published objection on safety ascertainment. Both will be cited at her. A rehabilitation service, an insurer or a colleague reading only the abstract will find a 2026 meta-analysis saying exercise improves fatigue in ME/CFS with a large effect and no excess harm, and a 2026 network meta-analysis recommending HIIT. Three things to have ready. First, the harm claim is the weak point and the letter says so by its title: in this literature adverse events are generally not systematically ascertained, drop-outs are the people who got worse, and “no significant difference in adverse events” in trials that did not measure deterioration after exertion is an absence of ascertainment, not an absence of harm. Second, the heterogeneity is disqualifying for the headline figure on its own terms: I² of 85.3% for fatigue and 89.7% for quality of life means the trials are not estimating the same thing, so a single pooled SMD is close to uninterpretable, whichever direction it points. Third, the two null outcomes are the objective ones — cardiopulmonary function and functional capacity both non-significant, with intervals centred near zero — so even on the paper's own showing exercise moved self-reported fatigue and did not move what a body can do — though both intervals still admit a small effect, up to about 0.4. That last pairing is the single most useful sentence to have ready, because it is the paper's own result and requires no appeal to the letter. None of this bears on pacing, and none of it displaces the three guideline systems logged in previous editions that gate exercise-based rehabilitation on PEM.
Weigh it accordingly. The letter was not read. Its argument, its evidence, whether it concerns the paper named above, and whether the authors are critical of that paper or defending it against someone else, are all unknown from what was obtainable. It is a one-page letter with no data, no abstract and no appraisal, and it is being carried on its title and the accessible record of its probable target — a weaker evidential basis than anything else in this edition. The identification of the target is this digest's inference from a journal-and-year search, not a stated fact; a reply by the original authors, or a prior letter to which this is itself a response, would both be consistent with what is visible. Every figure quoted above is from the Zhao et al. structured abstract read in PubMed; that paper's full text, its 17 included trials, its risk-of-bias assessment, its forest plots and its adverse-event extraction were not seen, and no registration identifier appears in its MEDLINE record. Three points of arithmetic in it. The cardiopulmonary confidence interval is asymmetric about its own point estimate: SMD 0.13 with limits −0.11 and 0.43 gives 0.24 below and 0.30 above, which a symmetric interval around a standardised mean difference cannot be, so one limit is misprinted and the result — a null one — cannot be reproduced from the abstract. The functional-capacity interval (0.09; −0.21 to 0.38) is symmetric to rounding and checks out. The subgroup counts are internally consistent: no outcome's trial count or sample exceeds the 17 trials and 1944 participants claimed, and depression and anxiety share an identical 9 trials and n = 1279, which is what reporting both from the same trials would produce. The SMDs are oriented so that positive means benefit, which is not stated in the abstract and means the anxiety result — SMD 0.14, 95% CI 0.04 to 0.24 — is being reported as a significant improvement on an interval whose lower bound is 0.04. No PEM instrument, no ME/CFS case definition and no severity stratification appear anywhere in that abstract, so which case definitions the 17 trials used, and whether any of them enrolled patients meeting a PEM-based definition, is unknown; in a literature of this age a substantial share will be Oxford-criteria fatigue trials. Zhao et al. declare no competing interests and add that results are presented “without fabrication, falsification, or inappropriate data manipulation”; the paper is indexed as Research Support, Non-U.S. Gov't with no funder named in the record. The network meta-analysis pools two diagnoses deliberately and declares no conflict of interest. Neither meta-analysis is carried as an item in this edition — both are outside the window and are reported here only as the context the letter requires.
J Psychosom Res·Editorial — no abstract, no original data·PMID 42946587·paywalled — full text could not be read·MEDLINE record and declarations only
Read this heading before the entry too. Seven authors across Université Paris Cité, Sorbonne Paris Nord, INSERM, the Paris Brain Institute and four AP-HP hospitals — H√¥tel-Dieu adult psychiatry and occupational-and-environmental pathology, the Avicenne functional neurological disorder unit, and the H√¥pital europ√©en Georges-Pompidou internal medicine and CASPer units — have written an editorial about the rewriting of the French statutory health insurer's public ME/CFS page. The full text is behind Elsevier with no PubMed Central copy and no abstract in MEDLINE; ScienceDirect served the same human-verification challenge as the letter above, so the editorial's content was not read. The ameli.fr page it concerns was also opened directly and served a human-verification checkbox, which was likewise not completed, so the page itself was not read either. This entry rests on the record.
What the record establishes. Publication type Editorial. Received 18 September 2026, revised 2 October, accepted 11 October, published online 12 October, entered PubMed 28 October. The author-supplied keywords are: Biopsychosocial model; Chronic fatigue syndrome; Health communication; Medically unexplained symptoms; Myalgic encephalomyelitis; Persistent physical symptoms; Stigma. The declared competing interest is that the first author, Cédric Lemogne, is an associate editor of the journal and was recused from the peer review of his own piece, with the editorial process delegated to another editor. No funding statement appears in the record.
Bottom lineTwo things are worth taking from this, and neither requires the text. First, the object itself. A national statutory insurer's public ME/CFS page is a document patients read, quote and bring in — and, in a system where the same body administers sick pay and invalidity, a document that shapes what they expect to be offered and believed. France has revised its. It is worth her knowing that national-payer ME/CFS communications are being actively rewritten and formally discussed, because the equivalent question — what does the page my patient read before this appointment actually say — is answerable and rarely asked. Second, the framing, from the title and the keyword set. “Communicating uncertainty without reinforcing mind–body dualism” is a precise statement of the real problem: a page that says the cause is unknown invites the reading that the illness is psychological, and a page that overcorrects promises mechanisms that are not established. That is the same bind in the consulting room, and it is the one piece of this that transfers directly. The authorship and keywords place the piece in the French psychosomatic and functional-disorders tradition — “medically unexplained symptoms” and “persistent physical symptoms” are that literature's terms, and a functional neurological disorder unit is among the affiliations — which is useful to know in advance, because an argument about how to describe aetiological uncertainty reads very differently to a patient depending on which tradition is making it. If a patient arrives having read either the page or coverage of this editorial, she will want to have formed her own view of the original.
Weigh it accordingly. The editorial was not read. What it argues, what it says the webpage used to say, what it says the webpage now says, whether the authors were involved in the revision, and whether they approve of it, are all unknown. Everything above is drawn from the title, the publication type, the affiliations, the keyword list and the declarations — and the characterisation of the authors' tradition is inferred from affiliations and terminology, not from anything they wrote here. This is an editorial: no data, no patients, no method, no appraisal, and in design terms the same class of document this digest has repeatedly excluded narrative reviews for. It is carried on subject rather than design, and the subject is specifically the therapeutic relationship and health communication, which this digest weights heavily; if that judgement is wrong, the error is that a title was carried for its promise rather than its content. A declared associate-editor relationship with the publishing journal, with recusal, is disclosed and is the correct handling; it is recorded because an editorial by an associate editor in his own journal on a contested framing question is a piece whose provenance the reader should be able to see. No funding statement is indexed. The French page it discusses is Un cas particulier : le syndrome de fatigue chronique ou encéphalomyélite myalgique on ameli.fr, under the Asthénie (fatigue) section — linked here because it returned a bot-verification wall to this digest but will almost certainly open normally in an ordinary browser session, which is the practical way to read it. No publication date or revision date for the page itself was obtained, so when the revision happened, and what changed, is not established here at all — only that an editorial says it happened.
J Transl Med·Single-author commentary — no original data·PMID 42934697·open access·full text and all 10 references read
A single author at McLean Hospital / Harvard, the Einstein Center for Youth Mental Health in Berlin and UKE Hamburg comments on the Charit√© hyperbaric-oxygen study, and the commentary is carried because it is the occasion to look properly at a trial patients are already asking about. Hyperbaric oxygen therapy is sold privately, in chambers, by the course. The underlying paper is Kim et al., “Hyperbaric oxygen therapy improves clinical symptoms and functional capacity and modulates thalamic connectivity in ME/CFS: a prospective cohort study” (J Transl Med 2026 Jun 5;24(1):744, PMID 42249466) — published in June and outside this window; its abstract was read for this entry and its full text was not.
What Kim et al. did, from their abstract. Thirty patients with ME/CFS (mean age 42.3 ± 11.7; 7 men, 23 women) received 40 HBOT sessions, with outcomes at baseline, during treatment and four weeks after. The primary outcome was the SF-36 physical functioning subscale. Thirty age- and sex-matched healthy controls (42.3 ± 11.3; 7 men, 23 women) were included for MRI comparison only. Reported: SF-36 PF improved during HBOT versus baseline (g = 0.71, p = 0.006); SF-36 pain (p = 0.002, g = 0.79) and the Chalder Fatigue Scale (p < 0.001, g = −0.87) showed what the authors call clinically meaningful reductions; exercise capacity (g = 0.66), muscle strength (g = 0.40) and information processing speed (g = 0.52) improved (all p < 0.05); adherence was high and “no major adverse events” were reported. At baseline, thalamic functional connectivity was increased in patients versus controls in bilateral sensorimotor (p < 0.001, t = 5.65, FDR-corrected) and visuo-occipital regions (p < 0.001, t = 5.40, FDR-corrected); after HBOT the hyperconnectivity shifted toward control patterns, and responders — defined as a ≥ 10-point increase in SF-36 PF — showed greater reductions than non-responders (p < 0.001, t = −4.34 to −5.18, FDR-corrected). Registered retrospectively: NCT06118138, 1 November 2023.
What the commentary adds. It reads the result through predictive processing: the thalamus as a gain-control gate whose job is to discount predictable ascending bodily signal, baseline hyperconnectivity as a failure of that attenuation, and its post-treatment reduction as filtering restored — one mechanism accounting for sensory hypersensitivity, pain, orthostatic intolerance and cognitive interference together. The author is careful that “prediction and precision refer here to neurally implemented statistical operations, not to conscious expectations, psychological causes, or voluntary control over symptoms.” He states his own two limits plainly: HBOT is a vascular intervention, so a change in BOLD connectivity “can reflect altered neurovascular coupling or perfusion as readily as altered neuronal gain,” and a pre/post design even with responders separated “cannot show that the connectivity shift causes the clinical improvement rather than simply going along with it.” He proposes measuring across the post-exertional trajectory rather than at rest — portable EEG paired with exertion, tracking the aperiodic (1/f) spectral slope at the onset, peak and resolution of PEM — and makes the point that matters most for her population: portable, home-based readouts could reach the moderately and severely affected, who are “routinely excluded from studies like Kim et al. that depend on attendance at a specialised centre.” He reports from Kim et al. that 28 initially interested patients had to be excluded, in part because they judged themselves too ill to attend regular appointments outside their home.
Bottom lineThis is the HBOT answer, and it has three parts. First, what the study is. Thirty patients, no control arm for any clinical outcome, outcomes four weeks after a course of forty sessions. The thirty healthy controls were scanned, not treated. So every clinical number above — the physical functioning, the pain, the fatigue, the grip, the processing speed — is an uncontrolled before-and-after in people who attended a hospital chamber dozens of times for an expensive treatment they had sought out, measured on self-report instruments four weeks later. That is the design in which expectancy, regression to the mean and the attention of a specialist centre are indistinguishable from the intervention, and effect sizes of g = 0.7 to 0.9 are exactly what such designs produce. The authors' own conclusion asks for controlled trials; that is the right reading of their own data. Second, what to say when asked. That a Berlin group has published a careful uncontrolled study in which thirty patients improved on several measures after forty HBOT sessions and showed a brain-imaging change that tracked who improved; that this is genuinely interesting and is not evidence that HBOT works, because nobody was randomised and nothing was blinded; that the one controlled question — forty sessions of HBOT versus forty sessions of sham pressure — has not been asked; and that a course at private rates is a large sum to spend on a question in that state. Third, the exclusion is the finding she should keep. Twenty-eight people wanted in and could not attend. Forty sessions means attending a hospital chamber several times a week for months, which selects for the mildly affected and for people with transport, time and a carer — so whatever this study measured, it did not measure it in the patients for whom the stakes are highest, and the same is true of nearly every intervention study in this illness. That is worth saying out loud to a housebound patient who has read about HBOT and is wondering why nothing seems to be about her. As for the predictive-processing account: it is a hypothesis, the author says so repeatedly, and it is not something to put to a patient — a framework in which symptoms arise from mis-weighted prediction is one of the easiest things in medicine to mishear as “it is in your head,” and this author has taken care to forestall that reading precisely because others will not.
Weigh it accordingly. This is a single-author commentary with no data, no patients, no search and no method; it is an interpretive essay, and the digest's usual ground for excluding no-data pieces applies to it squarely. It is in because HBOT is a live and costly question in clinic and because this commentary is the most rigorous public reading of the one study behind it. Every figure attributed to Kim et al. above is from their structured abstract as indexed in PubMed; their full text, tables, figures, exclusion flow and adverse-event reporting were not read for this entry, so the design critique above rests on what the abstract states about design, and the 28 excluded patients are reported second-hand from the commentary rather than from Kim et al. themselves. On Kim et al.: it is indexed as an Observational Study, not a trial; there is no control arm, no randomisation, no blinding and no sham; the primary outcome and the responder definition are both the SF-36 physical functioning subscale, so the responder analysis and the primary outcome are the same instrument; fatigue was measured with the Chalder Fatigue Scale, the instrument this patient population most distrusts and the one least able to register post-exertional deterioration; no PEM instrument, no PEM outcome and no case definition appear in the abstract; the post-treatment assessment is at four weeks, so nothing is known about durability; and the registration is explicitly retrospective (NCT06118138, 1 November 2023), which means the primary outcome and the responder threshold cannot be verified as prespecified. “No major adverse events” in an uncontrolled study with no PEM measure is an absence of ascertainment. Interests in the underlying study are substantial and the commentary does not mention them: the senior author declares consulting fees from Celltrend — a company selling autoantibody testing in this illness — honoraria from Bayer, Fresenius, Roche and Celltrend, a leadership position in the German Society for ME/CFS, and advisory-board participation at Berlin Cures and the BfArM; one co-author is an unpaid board assessor of the German Society for Diving and Hyperbaric Medicine, the professional body for the intervention under study; another declares consulting fees and honoraria from Daiichi Sankyo and NovoNordisk and an advisory board at Artcline GmbH. Ethics: Charit√© EA1/129/23, approved 4 July 2023, with written informed consent. On the commentary itself: the author declares no competing interests, and funding from the Deutsche Forschungsgemeinschaft (SFB/TRR 289, project 422744262, and a Walter-Benjamin scholarship, 531626241) and the Einstein Foundation Berlin (EZ-2025-852), with the funders having no role in it. Two things to see in it. The interoceptive-precision account it applies is the author's own, cited as a PsyArXiv preprint that has not been peer reviewed, and a second supporting reference is his own 2023 Neuron paper — self-citation rather than conflict, but the framework being advanced is his. And the declaration states that Claude Opus 4.8 was used to improve the readability and language of the manuscript, with the author taking full responsibility; recorded because it is declared, and because this digest is itself written by a model and should not pretend that is a remarkable fact about someone else. One reference is malformed — the Sherman and Guillery citation carries “1428” in the volume field, duplicating its issue number, with the year stranded after the PMID. Published 29 July 2026 and only entered PubMed on 27 October, so this is a three-month-old paper arriving new to the index; it is in this window by entry date, which is the window this digest uses, and is flagged so the date is not mistaken for novelty.
Int J Mol Sci·Assay development + cross-sectional sample, 15 ME / 5 controls·PMID 42941838·open access·abstract, cohort section and discussion headings read; full text not read end to end
Two authors at Cornell — Arnaud Germain and Maureen Hanson, whose laboratory runs one of the NIH ME/CFS collaborative research centres — built a sequencing assay to hunt enterovirus RNA at the limit of detection and then ran it on patient blood. The assay integrates high-throughput sequencing with cDNA capture using 49,847 probes targeting 3410 enterovirus genomes, with spike-in controls throughout, and reached a detection limit in artificial samples as low as one copy per microgram of RNA — in two of three biological replicates. Applied to blood from fifteen people with ME diagnosed by the Canadian Consensus Criteria and five healthy controls, it detected no enterovirus RNA at all, either directly or after attempted in vitro viral amplification. The authors' own framing is careful in both directions: the results “do not support the presence of circulating EV RNA in the blood of individuals with ME,” and “they do not rule out persistent infection in other tissues or organs.” They set out the enterovirus hypothesis fairly — historical overlap between ME outbreaks and poliomyelitis epidemics, the frequency of flu-like onset, seasonal clustering, serological findings — while noting that it “remains debated.”
Bottom lineThis is a clean, citable answer to a specific question patients bring in, and it is worth having precisely because it is negative. Chronic enterovirus infection is one of the standing persistent-pathogen theories, and it circulates with the practical corollary that someone should test for it. What this paper establishes is narrow and useful: with an assay sensitive enough to find a single copy in a microgram of RNA, there was no enterovirus RNA in the blood of fifteen patients. So a blood test for enterovirus is not a test worth having, and a positive serology is not evidence of ongoing infection — a distinction worth making explicitly, because the serological findings that keep the hypothesis alive are antibody results, not virus. Two honest qualifications to offer alongside it. Fifteen patients is small, and absence in twenty samples does not retire a hypothesis. And the authors' own caveat is the substantive one: enterovirus persistence in this illness has always been proposed in tissue — gut, muscle, brain — not in circulation, so a negative in blood is weak evidence against the theory as its proponents state it, and the paper does not pretend otherwise. What has actually changed is that the tool now exists: the next time someone claims to have found or excluded chronic enterovirus, the question to ask is which tissue and with what sensitivity.
Weigh it accordingly. This is a methods paper, and its clinical content is a negative finding in twenty people with no hypothesis, no power calculation and no comparison that could have been positive. It is carried on relevance, not design. Fifteen cases and five controls is a 3:1 imbalance with five controls in total, so no case–control inference of any kind is available — and none is attempted. Blood only. No tissue, no gut biopsy, no muscle, no post-mortem material — which the authors name as the obvious next step and as the reason their negative does not settle the question. Nothing is reported about when in the illness course samples were taken, whether near onset or decades in, or whether any participant had a recent intercurrent infection; participants were diagnosed by the Canadian Consensus Criteria, which is a PEM-requiring definition and a point in the paper's favour, but no PEM instrument, no severity stratification and no symptom scores are reported in what was read, so how ill these fifteen people were is unknown. The sensitivity claim is weaker than the title implies: “ultra-sensitive” rests on a limit of one copy per microgram achieved in two of three biological replicates, which means the headline sensitivity was not reproducible in a third of attempts — stated plainly in the abstract, and the kind of detail that is lost the moment the assay is cited. The negative result in patient samples carries the same caveat: an assay that misses its limit one time in three can miss a very low-copy target. This entry rests on the structured abstract, the cohort-demographics section identifying the 15 ME and 5 control participants, and the headings and summary sentences of the discussion; the full text, Table 1, the sequencing metrics and the complete limitations passage were not read end to end, so no figure outside the abstract and that cohort count is quoted. Published in International Journal of Molecular Sciences (MDPI): received 11 September 2026, revised 30 September, accepted 8 October, published 13 October, entered PubMed 27 October — a 27-day interval from submission to acceptance, recorded neutrally as a fact about the venue. Funding: NIH NIAID U54 AI178855 and 2U54AI178855-09A1, and the WE&ME Foundation — a federal collaborative-centre award plus a patient-facing charitable funder, both declared. The authors declare no conflicts of interest. Open access in PubMed Central (PMC13705469), with the full text free to read if she wants the sequencing detail.
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 mentioned in two of the five PubMed records that have an abstract, cannot be assessed in the two that have none, and is measured with an instrument in nothing at all. Pain, for the first time in this digest's recent run, was measured — with a visual analogue scale, in a randomised trial, and the result was null.
PEM: absent from the week's only randomised trial, and the exertional question the HBOT literature has not asked
Standing check·2 of 5 abstracts mention PEM; 2 records have no abstract; 0 administered a PEM instrument
The counts, stated so the absence is auditable. Of the seven PubMed records, five carry an abstract and two — the exercise letter and the French editorial — carry none, so PEM cannot be assessed in either. Of the five, PEM appears in two: the Stanford glymphatic perspective lists “persistent fatigue, sleep dysfunction, post-exertional malaise and brain fog” among the cardinal symptoms, and the thalamus commentary proposes measuring across “exertion and post-exertional malaise” rather than at baseline. Neither administered anything. Nor did any record in the window.
The gap that matters this week is in the trial. The low-dose naltrexone RCT enrolled 160 adults whose entry criterion was new symptoms of ME/CFS for three months or more after SARS-CoV-2 infection — a population defined by an ME/CFS criterion set, which for every current definition requires PEM. Its primary outcome was the Fatigue Severity Scale, a nine-item generic fatigue instrument with no exertional item, and its secondaries were a pain VAS, the PQSymp-12 (the Patient Phenotyping Questionnaire Short Form, the research group's own ME/CFS symptom score), daily step count and the EQ-5D-5L. The registry lists fifteen further outcomes in its other-outcomes tier, thirteen of them labelled exploratory, including a fatigue VAS, a sleep questionnaire, PHQ-9, GAD-7, handgrip strength and a thirty-second sit-to-stand — and no PEM-specific outcome among them. The PQSymp-12 is the closest thing in the design to a PEM-sensitive measure, and the abstract reports only its total score (adjusted mean difference −0.70, 95% CI −2.57 to 1.17); whether it contains a post-exertional item, and what that item did, cannot be determined from the preprint abstract or the registry. So the best-designed study in this window tested a drug in a PEM-defined population and measured fatigue, not PEM. A null result on the FSS therefore leaves one question genuinely open: whether LDN does anything to post-exertional symptoms specifically, which is what patients who report benefit from it usually describe.
And the same hole runs through the HBOT literature. Forty sessions, thirty patients, improvements on the SF-36 physical functioning subscale and the Chalder Fatigue Scale — and no PEM outcome, in an intervention that requires attending a hospital chamber several times a week for months. The exertional cost of the protocol was not measured, which is the single thing a patient with PEM would most want to know before committing to it; the twenty-eight people excluded because they were too ill to attend are the visible edge of that. The commentary is the one place in this window where the right design is named: within-person measurement across the post-exertional trajectory — after exertion, at the onset and peak of malaise, and through recovery — with home-based readouts that reach people who cannot travel. That is a research agenda, not a finding, and it is recorded as one.
Bottom lineUnchanged in substance and sharper in one respect. Nothing in this window measured PEM, and the week's one randomised trial recruited on an ME/CFS definition and then measured generic fatigue — which is how a drug can be declared ineffective for a PEM-defined illness without anyone having asked whether it changed PEM. The practice point stays what it has been: a named PEM instrument in the record alongside whatever generic scale a form demands, plus a dated note of the specific exertional exposure and what followed. Add one new question to the counselling repertoire: when a patient is considering any course-based treatment — hyperbaric oxygen, a rehabilitation programme, repeated hospital attendance of any kind — the exertional cost of attending is part of the intervention and is nowhere in the published evidence for it. Asking what forty trips to a chamber would cost her, in her own currency, is a more useful conversation than the trial data can supply.
Pain: measured at last, in a randomised trial, and it did not move
Standing check·0 of 7 PubMed records administered a pain instrument; 1 preprint measured pain by VAS in an RCT — null
Reported plainly: no PubMed record in this window measured pain with any instrument, and pain is barely mentioned in any of them. But the drought in the preprint shelf has broken, and it broke in the best possible design. The low-dose naltrexone trial included a pain visual analogue scale (0–10) as a prespecified secondary outcome, registered as such since 2022, in 160 randomised participants. At sixteen weeks the adjusted mean difference was 0.03, 95% CI −0.50 to 0.57 — a null result with a tight interval, ruling out anything approaching a clinically meaningful analgesic effect in this population at doses up to 4.5 mg daily.
That is worth stating carefully, because it is the first controlled pain result in this digest's recent run. Low-dose naltrexone is prescribed for pain as much as for fatigue — in fibromyalgia it is one of the better-supported off-label options, on small trials — and the usual argument for trying it in post-viral illness borrows from that pain literature. This trial tested it directly in a post-COVID ME/CFS-pattern population and found nothing. One qualification, which is real: the participants were recruited for fatigue, not pain, so the sample is unselected for pain severity and a mean VAS at baseline is not reported in the abstract. A null in a population with mild pain is weaker evidence than a null in a population with severe pain; that distinction cannot be made from what was read, and the full preprint would settle it.
Bottom lineOne new thing, and it is useful. In the only randomised trial in this window, low-dose naltrexone at up to 4.5 mg for sixteen weeks did not reduce pain in post-COVID fatigue syndrome — adjusted mean difference 0.03 on a 0–10 scale, with a confidence interval too narrow to leave room for a meaningful effect. For a patient asking about LDN for pain specifically, that is now a real answer rather than an extrapolation from the fibromyalgia literature, with the caveat that the sample was recruited for fatigue and the baseline pain severity is unknown from the abstract. The items still worth having are both from previous editions and neither is re-carried: the Leeds and Manchester preprint on Long COVID musculoskeletal pain barely moving over twelve months at a mean of 2.61 years, and the Brussels finding that what distinguished ME/CFS from fibromyalgia was pain distribution and what happened the day after exertion rather than a symptom tally. Last edition's endometriosis prompt — asking a woman with ME/CFS-pattern illness and cyclical pelvic pain about endometriosis directly — still stands and is not repeated here.
Preprint watch
medRxiv, 26 Oct – 2 Nov. Four searches returned three hits comprising two distinct postings; both are relevant and both are covered. Neither is peer reviewed; they are listed so nothing arrives as a surprise. The first is the most consequential item in this edition and is a null drug trial, which is reported here at the length a positive result would have been given. Industry funding: none in either.
medRxiv preprint·Not peer reviewed·Double-blind placebo-controlled RCT, n = 160·posted 26 Oct 2026·public and charity funded, no industry support·abstract, declarations and trial registry read; full text not read
Sixteen authors, led from Vancouver with Luis Nacul as senior author — the group behind British Columbia's complex chronic diseases service — report a decentralised, randomised, parallel-group, double-blind, placebo-controlled trial conducted in British Columbia and Ontario between January 2024 and February 2026. Participants were adults aged 19 to 69 meeting criteria for post-COVID fatigue syndrome, defined as new symptoms of ME/CFS persisting three months or more and starting after SARS-CoV-2 infection. Randomisation was 1:1 to oral low-dose naltrexone, starting at 1.0 mg daily and increasing to 4.5 mg as tolerated, or placebo, for sixteen weeks. The primary outcome was fatigue severity on the Fatigue Severity Scale.
The result is null across every outcome reported. Of 160 enrolled, baseline characteristics were similar between LDN (n = 81) and placebo (n = 79). At sixteen weeks there was no difference in mean FSS score (LDN 4.2, placebo 4.4; adjusted mean difference −0.76, 95% CI −4.18 to 2.67). Nor in pain VAS (AMD 0.03, 95% CI −0.50 to 0.57), the PQSymp-12 ME/CFS symptom score (AMD −0.70, −2.57 to 1.17), daily step count (AMD −5, −720 to 729) or EQ-5D-5L utility (AMD 0.02, −0.03 to 0.06). Adverse events of grade 3 or higher occurred in 5 of 81 (6.2%) on LDN and 3 of 79 (3.8%) on placebo. The authors' conclusion: LDN at up to 4.5 mg/day for sixteen weeks “was well tolerated but did not improve fatigue severity in people with PCFS.”
Bottom lineThis is the best evidence that exists on a drug she is asked for by name, and it says the drug does not work for fatigue. Four things follow. First, what to say to someone considering it. There is now a 160-person randomised, placebo-controlled trial in post-COVID ME/CFS-pattern illness, and after two years of recruitment and sixteen weeks of treatment the fatigue scores were 4.2 against 4.4 — both groups still fatigued, no separation on any measure, including an objective one. Step count moved by five steps a day. That is as close to a flat result as a trial produces, and it is not a borderline miss that a bigger trial would rescue. Second, what to say to someone already on it who says it helps. Not that she is wrong. A null average effect is compatible with individual benefit, the trial measured generic fatigue rather than PEM or the symptom clusters patients usually credit LDN with, and it ran sixteen weeks. But it does mean the honest framing has changed: continuing is now a decision based on her own experience against a negative trial, not a treatment with evidence behind it, and that is worth saying plainly before she spends more on it. Third, the practical rule the trial supports best. Sixteen weeks at up to 4.5 mg is an adequate trial of this drug. If nothing has changed by then, the evidence does not support escalating the dose or waiting longer. Fourth, and the most useful part: it was well tolerated. Grade 3 or higher events in 6.2% versus 3.8% — a difference of two patients — in a population notoriously intolerant of medication. So the risk of a careful trial in an individual patient is low, which is the one thing that keeps a shared decision open. What this trial has not settled is whether LDN does anything for post-exertional malaise specifically, because nobody measured it.
Weigh it accordingly. Preprint, not peer reviewed; medRxiv's notice that preprints should not be used to guide clinical practice applies, and the discrepancy named below is the reason this digest is not promoting it out of this section yet. This entry rests on the abstract, the author declarations and funder information on the posting, and the ClinicalTrials.gov registry record read in full; no full text, PDF, table, figure or supplement was opened. The headline confidence interval does not reconcile with the means printed beside it. The Fatigue Severity Scale is nine items scored 1–7, reported either as an item mean (1–7) or a total (9–63). The reported means, 4.2 and 4.4, can only be item means. An adjusted mean difference of −0.76 with limits of −4.18 and 2.67 on that same 1–7 scale would span most of the usable range of the instrument, which 160 randomised participants cannot produce alongside an observed difference of 0.2; the interval is consistent with an analysis on the 9–63 total score, which the registry's wording (“FSS total score”) supports. So the abstract most likely reports its means on one scaling and its treatment effect on another, without saying so — which must be checked in the full text before either figure is quoted. It does not change the direction: on either scaling the result is null. The registered primary outcome is not the one reported. The registry specifies “Fatigue Intensity — change in the Fatigue Severity Scale (FSS) total score by 4.7 points or over” at sixteen weeks: a dichotomous responder endpoint. The preprint reports a continuous between-group mean difference instead. The proportion of responders in each arm — the registered primary analysis — is not in the abstract at all, and should be, because a null mean difference and a null responder analysis are not the same claim. Everything else matches the registry exactly: phase 2, quadruple masking, 160 actual enrolment, the four secondary outcomes in the same order, start 15 January 2024 and completion 28 February 2026 as stated, and first posting on 24 June 2022, eighteen months before the first participant. No PEM outcome appears in the reported results, the registered secondaries or the fifteen further registered outcomes. The trial was decentralised, which is what made a 160-person trial feasible in this population and is a genuine methodological strength, but it also means outcomes were collected remotely with no in-person assessment reported, and step count is an actigraphic measure whose device, wear-time rule and handling of housebound participants are not described in the abstract. Nothing is reported about retention, withdrawal, adherence to titration, how many participants actually reached 4.5 mg, or how many of the 160 contributed to the sixteen-week primary analysis — and in a sixteen-week trial in this illness, differential drop-out is the first thing to check. Baseline severity is described only as “similar between groups”; no baseline FSS, pain VAS or step count is given, so whether the sample was mildly or severely affected, and whether there was room to improve, cannot be judged — and a final FSS item mean of 4.2 to 4.4 is moderate, not severe. The trial enrolled a post-COVID population only; it says nothing directly about LDN in ME/CFS of other or unknown onset, which is a distinct and common clinical question. Adverse-event reporting is a single line: no breakdown, no listing by system, no withdrawals-for-adverse-event figure, and the arithmetic checks (5/81 = 6.2%, 3/79 = 3.8%, 81 + 79 = 160). Ethics: University of British Columbia / Children's and Women's Health Centre of BC Research Ethics Board, H21-02254, with ethics, consent, registration and reporting-guideline declarations all confirmed on the posting. CC-BY 4.0. Interests are declared fully and none is commercial: no support from any organisation for the submitted work; funding from the Canadian Institutes of Health Research (177749) and the ME Association (QGGC GR035116 MEA 2025) — a patient charity, recorded because this digest records aligned-interest funders in both directions, and noting that in this instance a charity funded a trial that returned a negative result on a treatment its members use. Two role disclosures that belong in view: the senior author was a member of the NICE ME/CFS guideline committee and of the EUROMENE consensus, uncompensated; the first author was a member of the Canadian Guidelines for Post-COVID-19 Condition pharmacologic and non-pharmacologic interventions team, for which he received honoraria, holds a CAN-TAP-TALENT post-doctoral award and fellowship funding from the Post-COVID-19 Interdisciplinary Clinical Care Network and CIHR, and is an unpaid editorial fellow at the Journal of General Internal Medicine. Trial registration NCT05430152.
medRxiv preprint·Not peer reviewed·Prospective cohort with survival analysis, n = 303·posted 30 Oct 2026·no competing interests, no funder declared·abstract and declarations only
Five authors in Lusaka followed 303 people randomly sampled from seven healthcare facilities in Lusaka and Southern provinces who were infected during Zambia's third wave (June–August 2021), with data collected at baseline and monthly from clinical records and interview over 24 months (September 2021 to August 2023). They analysed time to symptom resolution by Kaplan–Meier, fitted a Poisson model for the crude recovery rate, and used Cox proportional-hazards regression for covariate associations. 236 resolution events occurred, giving a crude incidence rate of 1575.6 events per 1000 person-years (95% CI 1363.2–1784.4). The median time from diagnosis of Long COVID symptoms to resolution was 6 months (IQR 5–8). Of the covariates tested, only vaccination status was significant: adjusted hazard ratio 1.71 (95% CI 1.02–2.86, p = 0.04) for faster resolution, after adjustment for age, number of children, occupation, pre-existing conditions, number of vaccine doses and hospital admission. The authors are explicit that “causation can not be ascertained due to the observational design” and that an experimental design would be needed.
Bottom lineCarry this one for the harm it can do in the wrong hands, not for what it teaches. A median time to resolution of six months, with three-quarters resolved by eight, is a far more optimistic prognosis than anything in the ME/CFS or severe-long-COVID literature — and it will be quoted. The reason is definitional, not geographic: this cohort is people who had symptoms after an acute infection in a general facility-based sample, followed monthly, with resolution recorded from notes and interview. That is a population in which most post-acute symptoms do resolve. It is not the population in her clinic, which is selected precisely because resolution did not happen, and it contains no ME/CFS case definition, no PEM criterion and no severity stratification at all. So the useful framing when a patient brings this in is that it is a true finding about a different question: most people with symptoms after COVID-19 get better within months, which is why the people who do not are a distinct group needing a distinct answer — and a study that stops at whether symptoms resolved cannot tell you anything about the minority in whom they did not. The vaccination finding is the part worth keeping, with its limits attached. An adjusted hazard ratio of 1.71 for faster resolution in vaccinated participants, from a lower confidence limit of 1.02 and p = 0.04, is a weak signal pointing the same way as the stronger prevention evidence already logged in previous editions. It is not a reason to promise anyone a faster recovery, and it is one more small reason to keep vaccination current. There is also a quiet positive: this is a prospective post-COVID cohort from a middle-income African public health system, a setting almost entirely absent from this literature, and that it exists at all is worth something.
Weigh it accordingly. Preprint, not peer reviewed; the posted notice applies, and this entry rests on the abstract and the declarations only — no full text, table or figure was opened. Observational throughout, with no control group and no comparator for recovery. The exposure of interest, vaccination status, is not randomised and is strongly confounded in 2021–22 Zambia by age, occupation, urban residence, health-seeking behaviour and access — and people who sought vaccination are people who engaged with health services, which is also how resolution was ascertained. The adjustment set is internally odd: vaccination status is the exposure and “number of vaccine doses” is among the covariates adjusted for, which are near-collinear, and adjusting an exposure for a near-identical variable can shrink or distort the estimate in unpredictable directions. The result sits at the edge of significance on both measures (lower limit 1.02; p = 0.04) and is the only significant covariate among many tested, with no multiplicity adjustment mentioned — which in a Cox model with at least six covariates is the ordinary explanation for one borderline finding. “Long COVID” is never defined in the abstract: no case definition, no symptom list, no threshold, no PEM criterion, no severity measure, and no instrument. Resolution was recorded from clinical records and interview, so the outcome is a clinician's or participant's statement that symptoms stopped, which in a monthly-interview design will capture fluctuating illness as resolved whenever it is in a good phase — exactly the failure mode that matters in relapsing post-viral illness and the one that would most inflate an apparent recovery rate. The arithmetic is internally consistent: 236 events at 1575.6 per 1000 person-years implies about 150 person-years of follow-up, which across 303 participants averages roughly six months each and matches the stated median. But 236 of 303 is 78% resolving, leaving 67 participants censored or unresolved at 24 months, and the abstract says nothing whatever about them — which is the only subgroup in this paper relevant to her practice. The cohort is pre-Delta and pre-Omicron, infected in mid-2021 and almost entirely before vaccination in many cases, so it speaks to no current variant. Ethics approval is numbered UNZA BREC 2565-2022 for a study the abstract says began in September 2021, so approval appears to postdate the start of data collection; the declarations confirm ethics, consent, registration and reporting-guideline compliance, and no trial registration identifier is given for what is described as a prospective study. The posting carries several uncorrected errors — “Agust” for August, “Netherless” for nevertheless, “willngness,” “consnet,” “particpate,” and the crude rate typeset as “1, 575.6” — noted only as evidence of its uncertified state. No funder is declared and the authors declare no competing interest; the senior author list includes the former national director of public health, which is context rather than conflict. CC-BY 4.0.
-
Four searches, three hits, two distinct postings, both carried in full. The myalgic encephalomyelitis phrase search returned one posting — the LDN trial — its second consecutive non-empty window after a three-window drought. The post-COVID search returned the same posting, the only duplicate across the four searches, and the first duplicate in this digest's recent run. The long COVID phrase search returned one, the Zambia cohort. The post-exertional search returned zero for the fifth consecutive window, extending the longest such run in this digest's record: five weeks in which no preprint has put post-exertional malaise in its title or abstract at all. Nothing was judged off-topic and nothing was left unopened; both abstract pages were read in full, including their competing-interest, funding, ethics, registration and reporting-guideline declarations. Dates reconcile: the LDN posting carries DOI date 2026.10.22 and posted date 26 October, the Zambia posting 2026.10.28 and 30 October — both a two-day interval, both inside the window on the posted date, which is the date this digest uses. One structural note worth recording. Two postings in a week, one of them a null randomised drug trial, is the smallest preprint shelf and the highest average quality in this digest's recent run; a 160-person placebo-controlled trial is a different order of evidence from the genotype study, the EHR-coding analysis and the observational cohorts this section has carried in recent weeks. The lesson for next week is not that the shelf is thinning but that volume and usefulness are unrelated, which is the premise of the whole exercise.
How this was assembled
PubMed was queried through the browser for records added between 26 October and 2 November. The count was confirmed separately at seven by reading the results-count heading element directly, and that check earned its place for the fourth week running: a plain text extraction of the results page again returned only the first of the seven records, as it has in every previous edition. The seven PMIDs in the MEDLINE listing match the confirmed count of seven, so the page was not truncated. The format=pubmed request was served a reCAPTCHA interstitial on first load — “checking your browser before accessing pubmed.ncbi.nlm.nih.gov” — which resolved itself on its own after a short wait, with no challenge completed; the full MEDLINE listing was then retrieved normally. Entry dates ran from 27 October at 05:29 to 30 October at 23:29; nothing entered PubMed on 26 October, 31 October, 1 November or 2 November. Last edition's records ran from 20 to 25 October, so the shared boundary date of 26 October produced no overlap and nothing was double-counted or lost off either end. None of the seven appears in last edition's listing, whose records ran from 20 to 25 October, so nothing is re-carried. Full MEDLINE records — abstracts where they exist, affiliations, publication types, MeSH headings, grant numbers and competing-interest statements — were read for all seven.
One paper was read end to end: the thalamus commentary, open access in PubMed Central, including the figure legend, the convergent-evidence and limitations sections, all ten references, the author-contribution and funding statements, the data-availability statement, the AI-usage declaration and the competing-interests statement — which is where the Kim et al. exclusion of 28 patients, the author's reliance on his own unreviewed PsyArXiv preprint, the declared use of a language model, and the malformed Sherman and Guillery reference were found. For the enterovirus paper the PubMed Central full text was opened and queried for its cohort description, which established fifteen ME participants diagnosed by the Canadian Consensus Criteria and five healthy controls, and for its discussion, whose limitation and summary sentences were read; the full text was not read end to end and its tables and sequencing metrics were not seen, so that entry quotes nothing beyond the abstract and the cohort count, and says so.
Two carried records could not be read at all beyond their PubMed entry, and this is the first edition in which that has happened to a carried item. The exercise letter and the French editorial are both one-page Elsevier pieces in the Journal of Psychosomatic Research with no abstract in MEDLINE and no PubMed Central copy. Their DOIs resolve to ScienceDirect, which returned a “Are you a robot?” captcha page; this digest does not complete bot-detection challenges, so no attempt was made and both pages were abandoned. Each entry therefore rests on the MEDLINE record, the publication type, the affiliations, the keyword list where one exists, the dates and the declarations, and each says so in its heading, its meta line and its caveat. For the exercise letter, the substantive content was obtained instead by searching the same journal for its probable target and reading that paper's structured abstract; the identification is this digest's inference and is labelled as such in the entry, with the alternative candidate named. Separately, the ameli.fr page that the French editorial discusses was located by name and URL and opened directly; it served a Cloudflare human-verification checkbox, which was likewise not completed, so the webpage at the centre of that entry was not read either, and nothing is asserted about what it says.
Two further papers outside this window were read at abstract level because carried entries depend on them, and both are flagged as context rather than carried as items: the Zhao et al. exercise meta-analysis (J Psychosom Res 2026;207:112677) and Kim et al.'s hyperbaric oxygen cohort (J Transl Med 2026;24(1):744), the second of which also supplied the competing-interest statements quoted in the HBOT entry. Neither full text was retrieved. The Gao et al. network meta-analysis (J Psychosom Res 2026;206:112618) was read at abstract level and is named only as the alternative target of the letter.
The ClinicalTrials.gov record for the low-dose naltrexone trial was read in full, and it changed the entry. The registry's status, design and outcomes modules were retrieved for NCT05430152, which established the prospective registration date eighteen months before first enrolment, the quadruple masking, the exact match between registered and reported secondary outcomes, the fifteen unreported further outcomes now in the pipeline section, and the discrepancy that matters most: the registered primary outcome is a dichotomous responder endpoint — a change in FSS total score of 4.7 points or over — while the preprint reports a continuous between-group mean difference, and the responder proportions are not in the abstract. The registry record also supplied the evidence for the scaling problem named in that entry. Retrieval note, recorded because it constrains what can be checked: the cloud container's egress policy returned a proxy 403 for clinicaltrials.gov, so the API was read through the browser instead; the human-facing study page rendered only its glossary to text extraction and was abandoned in favour of the JSON.
Where a paper's own numbers or statements disagree, the disagreement is named rather than smoothed over. In the low-dose naltrexone preprint: the reported Fatigue Severity Scale means of 4.2 and 4.4 can only be item means on a 1–7 scale, while the adjusted mean difference's interval of −4.18 to 2.67 would span most of that scale and is consistent only with an analysis on the 9–63 total — so means and effect appear to be reported on different scalings without that being stated; and the registered primary analysis is absent from the abstract. In the Zhao meta-analysis invoked by the exercise letter: the cardiopulmonary confidence interval, SMD 0.13 with limits −0.11 and 0.43, is asymmetric about its own point estimate by 0.06 and cannot be a symmetric interval, so one limit is misprinted. In the Zambia cohort: ethics approval is numbered 2565-2022 for a study said to have begun in September 2021; the exposure, vaccination status, is adjusted for “number of vaccine doses”; 67 of 303 participants are unaccounted for in the abstract; and six spelling and typesetting errors survive. In the enterovirus paper: the title's “ultra-sensitive” rests on a limit of detection reached in two of three biological replicates. In the thalamus commentary: one reference carries its issue number in the volume field with the year stranded after the PMID. Arithmetic that checked out is also recorded: in the LDN preprint, 81 plus 79 is 160, 5 of 81 is 6.2% and 3 of 79 is 3.8%, and every reported secondary matches the registry in name and order; in the Zambia cohort, 236 events at 1575.6 per 1000 person-years implies about 150 person-years, which across 303 participants matches the stated median of six months; in the Zhao abstract, no outcome's trial count or sample exceeds the 17 trials and 1944 participants claimed, and the functional-capacity interval is symmetric to rounding; in Kim et al., the patient and control groups match exactly on size, sex split and mean age. The colorectal-cancer study was checked too, and the observation recorded in its exclusion: three of its four reported confidence intervals have a limit within 0.03 of zero, at −0.03, −0.01 and 0.02, while the fourth runs 0.25 to 3.03.
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 mentioned in two of the five abstracts available, could not be assessed in the two records that have no abstract, and was measured with an instrument nowhere — including in the window's only randomised trial, which recruited on an ME/CFS definition and measured generic fatigue. Pain, by contrast, was measured for the first time in this digest's recent run, by visual analogue scale in that same trial, and the result was null; both facts are reported as they stand rather than padded. Nothing is filtered for a positive result: the longest and most prominent entry in this edition is a drug trial that failed, its bottom line is written at the length a success would have received, and the one encouraging thing in it — that the drug was well tolerated — is placed last rather than first. Two carried entries are explicitly labelled as resting on titles and records, in their headings, where a reader cannot miss it. Commercial, charitable and institutional interests are stated in each entry rather than buried: a patient charity co-funding a negative trial of a treatment its members use; two guideline-committee memberships and a set of honoraria declared by the trial's authors; a hyperbaric-medicine professional-body position and consulting income from an autoantibody-testing company among the authors of the study a carried commentary assesses, none of which that commentary mentions; a commentary advancing its own author's framework by way of his unreviewed preprint; a federal collaborative-centre award paired with a patient-facing charitable funder; and an associate editorship declared with recusal in the journal publishing the declarer's own editorial. A declared use of a language model in one paper is recorded without comment beyond the observation that this digest is written by one. Items carried in a previous edition are referred to, not recycled: last edition's Alberta zoster case, ERJ reinfection letter, endometriosis meta-analysis and Brazilian POTS case, the three guideline systems that gate rehabilitation on PEM, and the earlier Leeds pain and Brussels comparison findings 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 over the same window on title and abstract for myalgic encephalomyelitis (as a phrase), post-exertional, long COVID (as a phrase) and post-COVID, returning one, zero, one and one postings respectively — three hits comprising two distinct postings, both relevant and both carried as full entries. The post-COVID search returned the same low-dose naltrexone posting as the myalgic encephalomyelitis search, the first duplicate across searches in this digest's recent run. The post-exertional search returned zero for the fifth consecutive window. Both abstract pages were opened in full and their competing-interest, funding, ethics, consent, registration and reporting-guideline declarations read; no preprint full text or PDF was opened, and both entries say so. Nothing was judged off-topic and nothing was left unopened. No bot-verification interstitial was encountered on medRxiv this week.
Every title links to its PubMed record or, for preprints, to the medRxiv abstract page; the two trial registry records and the French insurer's webpage are linked where they are discussed. Where a paper is paywalled, the PubMed page carries the free-full-text link when one exists. Previous editions are archived alongside this file. Nothing here is treatment advice; it is a literature digest for a clinician.