Skip to content

Curation log — fibromyalgia

Newest entries first. Every content-changing session appends an entry. See CLAUDE.md.

2026-08-31 — Audit pass: red-flags-and-safety-concerns.md → curated

Scope: the one remaining draft page in this condition. No other FM page touched.

Searches run. PubMed esummary + efetch for all 41 unique PMIDs on the page; full text via PMC for the 2012 Canadian guideline (PMC3673928) and Häuser & Fitzcharles 2018 (PMC6016048); targeted PubMed searches for a triptan/SSRI-SNRI serotonin-syndrome source, for the German AWMF diagnostic guideline, and for any guaifenesin FM trial; direct retrieval of NICE NG193 recommendations.

Result: 41/41 records exist and match their reference-list entries; 0 fabricated or wrong-paper citations. Numeric claims checked against source abstracts and re-read in context where direction mattered (Wolfe κ values, Treister-Goltzman SMRs, Adawi ideation/attempt estimates, Duffield score inflation, Derry NNH, Fitzcharles opioid prevalence, Gore claims figures, Wall CAM percentages, Shaver polypharmacy). All correct. Two apparent misses were source-side typos, not page errors: Bennett's title says 2,596 where its abstract says 2,569 respondents, and the abstract spells alprazolam "aprazolam".

Errors found and fixed. - Canadian guideline overstated. The page said it "permits only tramadol among opioids". Its actual recommendations, read from the full text: rec 29 — a trial of opioids beginning with a weak opioid such as tramadol, reserved for moderate-to-severe pain unresponsive to other modalities (level 2, grade D); rec 30 — strong opioid use is discouraged, with continued users required to show improved pain and function. Rewritten verbatim-accurate. This also softened the TL;DR's "near-universal guideline recommendations against", which flattened a real spread between Canada 2012, EULAR 2017 and NICE. - NICE claim was unverifiable from its cited record (Carville 2021 has no PubMed abstract) and slightly wrong in verb. NG193 rec 1.2.10 says do not initiate opioids — alongside gabapentinoids/antiepileptics, antipsychotics, benzodiazepines, NSAIDs and paracetamol — for chronic primary pain in people 16+, and governs initiation rather than existing users. Now cited to NICE directly per CONVENTIONS §1 non-journal format. (Supersedes the standing follow-up "NICE NG193 full text 403/CAPTCHA-blocked" — it is retrievable.) - Mis-attribution: "pregabalin is preferred by misusers over gabapentin" was credited to Evoy 2021; that finding is Bonnet & Scherbaum's (users cite faster, stronger euphoria than oral gabapentin). Moved, and Bonnet's more informative numbers added (14 studies + 38 case reports; only 4 individuals met behavioural dependence criteria without another SUD, all on pregabalin). - Over-precision: Derry's NNH 3.7/7.4 are "for all doses combined", not "at 300–600 mg". Corrected, and the 70–90% adverse-event figure clarified as applying to all arms including placebo — which is the point of quoting it. - Orphan reference used: Macfarlane 2017 (EULAR, PMID 27377815) was in the reference list but cited nowhere. Now carries its own verified claim — every pharmacological option 'weak for', exercise the only 'strong for'.

Three of the five [unverified] markers resolved; two rewritten as explicit evidence gaps. - SNRI/triptan serotonin syndrome — resolved and now the strongest passage in the section. The FDA's July 2006 alert rested on 29 case reports; obtained under FOIA and scored, 7 met Sternbach criteria, none met Hunter criteria, and Evans concluded the data do not support prohibiting the combination (PMID 18092054). The 2010 AHS position paper classed it Level U; co-prescribing rose 90.1% between 2003–04 and 2007–08 to 1.8% of the relevant population with the resulting case count still undocumented (PMID 22289074). This converges with the page's existing Ryan & Isbister finding — signal strong in spontaneous reports, thin in criteria-confirmed events. - German AWMF checklist — the guideline itself is now cited (Eich 2017, PMID 28421273: AWMF 145/004, 13 societies + 2 patient organisations, diagnosis by ACR 1990 or modified ACR 2010/2011); its item-level red-flag list is not in the indexed record, stated as such. - Guaifenesin trial — searched; PubMed returns three guaifenesin+FM records, none a trial. Named as an absence rather than flagged. - Benzodiazepine harm data in FM and re-evaluation protocols — genuinely absent, not unverified. Rewritten as evidence-gap statements ("this is an evidence gap, not a safety clearance") and the [unverified] markers removed, since the markers implied doubt about a cited claim when the claim is the absence itself.

Frontmatter: status: draft → curated, last-curated: 2026-08-31. Reference list grew 41 → 45 (Eich, Evans, Sclar, NICE NG193).

Follow-ups. Reference 45 is this repository's first non-journal citation on an FM page — check guidelines uses the same NG193 wording ("do not initiate", 16+, initiation vs. continuation), since that page also summarises NICE. The Canadian-guideline correction may propagate: any other page saying FM guidelines uniformly recommend against opioids should be re-read against rec 29/30.

2026-08-28 — Initial build + audit pass

  • Scaffolded the condition per the add-a-new-condition workflow; canonical 18-page list added to CONVENTIONS §5 (designed for FM's shape: criteria genealogy, contested nosology, central/peripheral/autoimmune mechanism split, outcomes-and-measurement as first-class page).
  • Built all 18 wiki pages via eight domain research passes (foundations, mechanism, risk/sleep/overlap, clinical, frontier, patient voice, guidelines registry, statistics), each grounded in live PubMed/ClinicalTrials.gov/web queries.
  • Literature layer: BIBLIOGRAPHY.md (355 unique PMIDs, six sections, fragments merged and removed after parity check); 15 landmark notes; guidelines registry (23 documents + 5 regulatory decisions, 14 verified discordances); statistics sheet (~85 criteria-stratified figures, 9 documented conflicts); patient-voice layer (14 orgs verified by fetch, 10 themes, 73 annotated sources).
  • OPEN-QUESTIONS.md: 29 questions (9 Tier-1) + 12-junction "Dots not yet connected" table.
  • Audit pass (complete): first launch of five auditors was terminated mid-run by an account session limit; relaunched same-day (protocol is idempotent). Results: ~290 unique PMIDs and all 68-70 NCTs re-verified; ~700 claim-citation pairs checked; 0 fabricated/wrong-paper citations; ~15 substantive errors fixed — biggest: IVIG-in-SFN trial (NCT02637700) listed as pending had reported negative in Neurology 2021 (PMID 33766992, propagated through the autoimmunity argument); TNX-102 SL dose narrative corrected (RELIEF/RESILIENT used identical titration; RALLY was the 5.6mg-throughout trial); an unsupported FM/a conflict-of-interest claim reversed (paper declares no competing interests); AAPT criteria conjunction corrected (sleep OR fatigue); NICE-vs-FDA contradiction narrowed to pregabalin only; Kerrebijn GWAS cohort characterization corrected from full text (uniform ICD-code definition). Soft spots resolved: Tonmya approval date (Aug 2025, PMID 42479282), Stanza FDA clearance (PMID 40469887), NICE acupuncture (PMID 33824137), sodium oxybate rationale (PMIDs 21679091, 23373093), Bennett 2003 tramadol RCT (PMID 12753877), ICD-11 code MG30.01 (PMID 36427885), miRNA absence claim overturned with 4 sourced studies. All 18 built pages promoted to curated. Two false "no such guideline" negatives overturned → registry now 25 documents (added Yokota 2013 juvenile Japan, PMID 23758613; Zhang 2025 China patient-version, PMID 41410191). Statistics sheet sampled ~50 sources/120+ figures: 0 numeric errors.
  • Red-flags page added (wiki/red-flags-and-safety-concerns.md, draft, 41 refs — mimic red flags, medication safety, suicide-risk signal, exploitation risks); bibliography grew to 392 unique PMIDs (audit additions + red-flags merge); needs its own audit pass next session.
  • Trials safety enrichment: in progress (stage primer + per-trial safety numbers, mirroring the TAA page).

Notable build-time finds: Kerrebijn 2026 Nat Med GWAS (2.5M individuals, 26 loci, brain-exclusive heritability — PMID 42521817); the 2025 FDA approval of sublingual cyclobenzaprine verified via PubMed-indexed sources; the microbiome causal-chain-vs-null contest (PMID 40280127 vs 42550534); Brazilian SBR Jan-2026 guideline (newest in the field); the diagnosed-vs-criteria-positive κ≈0.3 divide as the field's measurement crisis.

Follow-ups for next session(s): - Append audit results; promote clean pages; update INDEX statuses and curation-state paragraph. - Unresolved build-time items to keep chasing: AAPT core-criteria exact wording (no PMC for PMID 30453109); ICD-11 FM code; EMA CHMP rationale documents (not PubMed-indexed); Stanza FDA-authorization status; NICE NG193 full text (403/CAPTCHA-blocked); AWMF post-2017 revision existence; sodium oxybate FDA advisory record; JCFI Japanese guideline (web-verified only); Queiroz 2013 "2.7% global prevalence" (not in abstract — full-text pull); FM-in-SLE meta-analysis (none found); FM-migraine meta-analysis (none found); population FIQ/FIQR norms (none found); primary claims source for opioid-prescription rates. - Patient-voice coverage gaps: Africa/South Asia orgs (Facebook-only), several org sites blocked (NFMCPA 403, FCSA JS-only, Dutch F.E.S. TLS) — retry. - Add FM to the weekly sweep task and republish frontends after audit integration.