Skip to content

Curation log — rheumatoid arthritis

2026-08-28 — Seed

  • Created: four seed files, 18 canonical pages, overview with 25 PMIDs, and 7 seed questions.
  • Searches run (live PubMed E-utilities esearch/esummary): 2010 criteria and validation; GBD; genetics and immune pathogenesis; TICORA/FIN-RACo; EULAR/ACR guidance and remission; methotrexate; ATTRACT, upadacitinib/abatacept, ORAL Surveillance; difficult-to-treat RA; R4RA; RA-ILD and mortality; patient-reported outcomes; ARIAA/APIPPRA prevention and prevention review.
  • Verification: every PMID cited was returned live by PubMed. Three recalled treat-to-target/BeSt identifiers returned unrelated non-RA papers and were excluded.
  • Next: full build queued per CONDITIONS-ROADMAP.md.

2026-08-28 — Full build

  • Built: all 18 canonical wiki pages as draft (2,800 lines, 30–31 references per page); 122-record master bibliography; six landmark notes; 15-document guideline registry; 56-row statistics sheet; patient-voice method, organization, theme and source files; 29 tiered open questions and 14 “Dots not yet connected” junctions.
  • PubMed live searches: re-fetched all 25 seed PMIDs; 2010 classification and validation; unclassified-arthritis imaging; GBD 2017/2021 burden and mortality; preclinical RA, ACPA and prevention; HLA/genetics; smoking, oral, lung and gut mucosa; synovial fibroblast/macrophage heterogeneity and single-cell studies; treat-to-target, remission, imaging targets and tapering; methotrexate monotherapy/combination, SWEFOT and glucocorticoids; ATTRACT, PREMIER, TEMPO, ADACTA, AMBITION, REFLEX and comparative biologics; SELECT-COMPARE, SELECT-CHOICE, ORAL Strategy and ORAL Surveillance; JAK malignancy/infection/ILD; difficult-to-treat definitions, phenotypes and rescue therapy; cardiovascular disease, serious infection, fracture, eye disease, vasculitis, pregnancy, cervical spine and septic arthritis; RA-ILD screening, prediction, treatment and guideline appraisal; fatigue, residual symptoms, work and patient-reported outcomes; EULAR, ACR, French, Mexican, Japanese and physical-therapy guidelines; biomarkers, plasma proteomics and R4RA.
  • ClinicalTrials.gov v2 live queries: RA general landscape; RA prevention; synovial biopsy; RA-ILD; direct terms for ARIAA, StopRA, ORAL Surveillance and other named programs. Thirty unique NCT identifiers now appear in the condition tree; each came from a returned API record in this session.
  • Public-source verification: fetched NRAS, Arthritis Foundation, Arthritis Society Canada, Arthritis Australia/MyRA, Arthritis Ireland, and EULAR PARE pages on 2026-08-28; recorded only aggregate themes and public service descriptions.
  • Deliberate choices: retained every page at draft because the playbook requires a separate-session audit before curated; used UNKNOWN exactly as ClinicalTrials.gov returned it and did not infer trial outcome; did not create an NCT identifier for APIPPRA where today’s API search returned none.
  • Could not verify or complete: no clearly authoritative RA-specific South African national patient-organization page was found in the targeted web search; several trial records had stale UNKNOWN statuses; RA-ILD prevalence was not reduced to one number because screening definitions conflict; no molecular treatment selector is ready for routine use.
  • Next session: independent audit of every claim→PMID pair, exact citation metadata, numeric transcription and registered-trial status; promote only fully checked pages to curated and then update roadmap to audited.

2026-08-30 — Page-by-page deepening

  • Deepened all 18 canonical wiki pages in place: preserved the existing verified text and added page-specific quantitative synthesis, including classification sensitivity/specificity and likelihood ratios; global-prevalence heterogeneity; HLA–smoking, silica, periodontal and microbiome estimates; single-cell and spatial synovial biology; biomarker decision thresholds; methotrexate optimization, RACAT, TEAR, CareRA and bridging data; NORD-STAR and biologic safety; FINCH and JAK safety conflicts; CAMERA, BeSt, ARCTIC and TaSER strategy effects; D2T prevalence and inflammatory/non-inflammatory decomposition; cardiovascular, mental-health, pregnancy and malignancy burden; RA-ILD prevalence, prognosis, screening, methotrexate controversy and INBUILD effects; infection, TB, VTE and vaccine-response estimates; prevention successes and null trials; current trial mechanisms; guideline disagreements; and patient preference, sleep, sexual-health and remote-monitoring outcomes.
  • Density after deepening: pages are 170–185 lines and contain 35–41 unique PMIDs each; clinical-trials-landscape.md contains 29 NCT identifiers. The wiki corpus now contains 245 unique PMIDs, all represented in the master bibliography.
  • Live PubMed work: ran more than 150 topic-specific E-utilities esearch queries across the 18 page scopes, followed by live esummary metadata retrieval and targeted efetch abstract retrieval for numerical transcription. Re-resolved the 126 pre-existing unique PMIDs in two live esummary batches and resolved every newly used PMID live. Two recalled imaging PMIDs returned unrelated ovarian-cancer and phytoremediation papers; both were rejected, and ARCTIC/TaSER were re-identified by title search as PMIDs 27530741 and 27026689.
  • Live ClinicalTrials.gov work: re-resolved all 21 pre-existing unique NCT records through API v2 and queried the current RA interventional landscape. Added eight individually re-fetched records: NCT04862117, NCT07137598, NCT07412704, NCT06570798, NCT06687551, NCT06733935, NCT05869955 and NCT07201558. Broad-condition false positives were excluded; multi-disease protocols are labelled as such, and registry status/enrollment are dated 2026-08-30.
  • Literature layer: expanded BIBLIOGRAPHY.md from 126 to 249 unique PMIDs with cited-by mappings (245 used across wiki pages plus four historical/current guideline-registry records); expanded STATISTICS.md with diagnostic, epidemiologic, strategy-trial, comorbidity, safety, D2T and RA-ILD tables; expanded guidelines/REGISTRY.md with EULAR 2025 supersession, vaccination, reproductive, cardiovascular, lifestyle/work and direct ACR/CHEST screening/treatment documents.
  • Open questions: sharpened OQ-10 through OQ-12 with newly quantified uncertainty; added OQ-30 through OQ-34 on outcome-changing RA-ILD screening, combined antifibrotic/immunomodulatory treatment, equitable remote monitoring, durable vagus-nerve modulation and preference-concordant care; added four cross-domain junctions D-15 through D-18.
  • Editorial choices: retained status: draft because the playbook requires a separate-session audit before promotion; reported confidence intervals and denominators where available; treated observational protective associations (especially methotrexate and RA-ILD/mortality) as non-causal; treated pooled null JAK comparisons as unresolved rather than as rebuttals of ORAL Surveillance; and updated the guideline page/registry to recognize EULAR 2025 as superseding EULAR 2022 while preserving the 2022 evidence discussion.
  • Next check: independent claim-to-abstract audit of all 245 PMIDs and exact journal metadata; verify whether multi-disease cellular-therapy trials have opened RA-specific cohorts; monitor EULAR 2025 implementation, RESET-RA durability and long-term APIPPRA/ARIAA outcomes.

2026-08-30 — Independent audit

  • Scope checked: all 18 wiki pages and every literature artifact. The final corpus contains 3,169 wiki lines, 459 substantive claim→PMID pairs outside evidence maps, 789 wiki-body PMID occurrences, 73 numeric/effect-size claim lines, 154 literature-layer PMID occurrences outside the bibliography, 257 unique PMIDs and 29 unique NCT identifiers.
  • Live verification: re-fetched the original 249 PMIDs in four PubMed efetch batches; retrieved eight focused absence/current-guideline searches and the resulting prevention, RA-ILD and guideline records; directly fetched all 29 ClinicalTrials.gov v2 records; and re-fetched all patient-organization URLs. Final unresolved identifier counts: PMID 0; NCT 0; DOI mismatch across six landmark notes 0.
  • Citation/reference repairs: rebuilt all 18 wiki reference sections and the 257-entry bibliography from live PubMed metadata; normalized incorrect secondary-author labels; removed nine unused wiki references; corrected the overview tapering question to cite RETRO/TARA rather than remission criteria. Final reference-title/author mismatches: 0; wiki body/reference-set mismatches: 0; duplicate bibliography PMIDs: 0; cited PMIDs missing from bibliography: 0.
  • Structural repairs: moved 11 evidence-map blocks out of unrelated clinical tables and into their intended sections, then validated every Markdown table and canonical wiki link. Final malformed-table count: 0; misplaced evidence-map rows: 0; noncanonical wiki links: 0.
  • Stale evidence corrected: incorporated live 2026 StopRA (PMID 40884017), APIPPRA/ALTO (PMID 41576971) and five-year TREAT EARLIER (PMID 42392130) results; replaced vague prevention claims with overall and subgroup effect estimates; dated the lifestyle-prevention evidence gap after fresh searches; and added current APLAR, ALAT, BSR and South African guidance found in the refreshed guideline search.
  • Stale absence corrected: the prior claim that no accessible South African national patient organization was found was false. The Arthritis Foundation of South Africa site resolved live and was added with its national helpline, education, support and advocacy services.
  • Final state: all 18 pages promoted to status: curated; [unverified] flags: 0; unresolved substantive issues: 0. ClinicalTrials.gov UNKNOWN labels remain where the live registry itself returns UNKNOWN; they are documented registry state, not unresolved evidence flags.
  • Post-promotion integrity pass: found and fixed one reference-order mismatch on biomarkers-and-tissue-precision.md after the STRAP update; rechecked all 18 body/reference sequences, reference numbering, Markdown tables, internal links, page density, bibliography coverage, six note DOIs and roadmap/index status. Final errors: 0; pages span 169–187 lines with 35–42 verified references each.

2026-08-31 — Periodic literature sweep

  • Searches run: 17 topic-specific PubMed E-utilities searches covering classification/diagnosis, epidemiology/burden, preclinical prevention, genetics/environment/mucosa, synovium, treat-to-target/remission, conventional DMARDs, biologics, JAK inhibitors, difficult-to-treat disease, comorbidity, RA-ILD, guidelines, biomarkers, trials, patient-reported outcomes and safety, all restricted to publication dates 2026-08-28 through 2026-08-31; one broad RA title/abstract query returned 16 records. A separate live title query recovered a relevant GBD 2023 LMIC analysis published 2026-08-21 but missed by the previous build. Live abstracts and metadata were fetched for every retained PMID.
  • PubMed triage: five records survived. Wang et al. quantified GBD 2023 RA burden across 129 LMICs (PMID 42495527); Fernández-Codina et al. compared oral and parenteral early glucocorticoid exposure in 2,222 CATCH participants (PMID 42665531); Dalix et al. reported a TNFA-methylation methotrexate-response model requiring external validation (PMID 42664067); Ikeda et al. reported two-year observational baricitinib outcomes including 88 D2T patients (PMID 42667647); and Sugiyama et al. described five-year CKD progression in 601 patients (PMID 42667508). Case reports, incidental RA mentions, an animal celastrol study, a cross-sectional miRNA diagnostic candidate, a small immune-checkpoint-inhibitor cohort and elbow arthroplasty outcomes did not survive triage. No paper was landmark-level; no deep note was created.
  • ClinicalTrials.gov: individually re-fetched all 29 cited NCT records through API v2 and compared overall status, enrollment, last update and presence of posted results with the audited 2026-08-30 state. No tracked record changed status or gained/lost posted results. A broader RA update query identified two RA-inclusive early-phase records updated on 2026-08-28: PF-08154225 across four autoimmune diseases (NCT07782450) and MK-1045 across SLE/RA (NCT07363590); both were individually fetched and added with explicit multi-disease/phase-1 caveats.
  • Guidelines, statistics and patient organizations: official EULAR, ACR, BSR, DGRh/AWMF and targeted web searches found the new German DGRh 2026 S3 DMARD guideline, which replaces the 2018 S2e document and adds specific de-escalation and RA-ILD positions. It was added to literature/guidelines/REGISTRY.md and synthesized in wiki/guidelines.md. The LMIC GBD analysis and four observational studies were added to literature/statistics/STATISTICS.md. Searches of NRAS, Arthritis Foundation, Arthritis Society Canada and German patient-organization material found no post-cutoff campaign or patient-experience source substantial enough to change the patient-voice layer.
  • Files changed: updated wiki/clinical-trials-landscape.md, wiki/guidelines.md, literature/BIBLIOGRAPHY.md, literature/guidelines/REGISTRY.md, literature/statistics/STATISTICS.md, OPEN-QUESTIONS.md (OQ-12, OQ-21 and OQ-29), and INDEX.md. The two edited wiki pages retain status: curated because every new claim was verified directly against the live registry or official guideline source; both carry last-curated: 2026-08-31.
  • Next sweep: check whether NCT07782450 or NCT07363590 opens a separately reported RA cohort; monitor the DGRh S3 journal publication and independent appraisal, especially its early antifibrotic/RA-ILD and DMARD de-escalation positions; look for external validation of the TNFA methylation response model; and continue the prior RESET-RA and APIPPRA/ARIAA durability watch.

2026-08-31 — Independent audit of the 2026-08-31 sweep (auditor: Grok; author audited: Codex)

  • Scope: the two wiki pages the sweep edited (clinical-trials-landscape.md, guidelines.md) and every sweep-touched line in BIBLIOGRAPHY.md, STATISTICS.md, guidelines/REGISTRY.md, OPEN-QUESTIONS.md (OQ-12, OQ-21, OQ-29) and INDEX.md. Pages the sweep did not edit were not re-audited; they retain the 2026-08-30 audit.
  • Live identifier verification (this session): PubMed E-utilities esearch plus efetch for PMIDs 42495527, 42665531, 42664067, 42667647 and 42667508 — 5/5 resolved with matching titles, first authors and abstracts. ClinicalTrials.gov API v2 batch plus individual protocol fetches for all 31 unique NCT IDs on the trials page — 31/31 resolved. Direct retrieval of the DGRh press release and AWMF register 060-004. A targeted PubMed search for a DGRh 2026 S3 journal article returned 0 records, matching the sweep's watch-list stance.
  • Claim–citation pairs checked: 42. Twenty numeric/qualitative claims from the five new papers against live abstracts; bibliography metadata for those five records; protocol-level claims for NCT07782450 and NCT07363590 (status, phase, enrollment, conditions, RA eligibility); overall-status match for the 29 previously tracked NCT IDs; six DGRh recommendation claims against the official 2026-08-26 press release and AWMF 060-004; INDEX counts (262 unique bibliography PMIDs, 31 NCT IDs, 27 registry documents, 94 statistics rows).
  • Verified without change: Wang 2026 (PMID 42495527) — 12.6 million prevalent cases in 129 LMICs in 2023 and a 4.1-fold upper-middle- versus low-income prevalence ratio, correctly scoped to LMICs rather than global GBD. Fernández-Codina 2026 (PMID 42665531) — 2,222 CATCH participants; 12-month glucocorticoid use 47% after early oral, 26% after parenteral and 8% after no early glucocorticoid; adjusted OR 9.8 and 4.1 versus no early glucocorticoid; OQ-12 correctly flags confounding by indication. Dalix 2026 (PMID 42664067) — 201 early untreated RA; combined AUC 0.760 versus clinical-only 0.699; correctly not treated as an externally validated selector. Ikeda 2026 (PMID 42667647) — 353 overall / 88 D2T; 24-month retention 58.2% / 42.0%; CDAI remission 36.4% / 22.9%. Sugiyama 2026 (PMID 42667508) — CKD-stage progression in 106/601 over five years. NCT07782450 is NOT_YET_RECRUITING phase 1/2, estimated n=54, RA eligible in parts 1a/1b. NCT07363590 is RECRUITING phase 1, estimated n=21, SLE and RA. All 29 previously listed NCT statuses match the live registry. DGRh S3 replaces the 2018 S2e document (AWMF 060-004); 1–3-month assessment, 3-/6-month adjustment, glucocorticoid stop generally within three months, gradual DMARD reduction without complete withdrawal, and a first dedicated RA-ILD recommendation all match the official release.
  • Absence re-checks: no PubMed-indexed DGRh 2026 S3 journal publication yet. NRAS/Arthritis Foundation/Arthritis Society Canada material retrieved for 2026 includes pre-cutoff items (SMILE-RA modules February, Healthcare Champion Awards April, Work Without Barriers July) and the annual September RA Awareness Week; nothing post-2026-08-28 required a patient-voice change.
  • Precision fix (1): guidelines/REGISTRY.md labelled PMID 39389412 as “Daien et al.” Live ESummary returns Fautrel B as first author (Daien C last author), matching the wiki page and bibliography. Corrected the registry row. Pre-existing secondary-author label on a file the sweep edited; not introduced by the five new papers.
  • Verdict: the sweep was clean. No fabricated identifier, no numeric or directional mismatch, no stale absence, and the two new trial rows are correctly caveated as multi-disease early-phase programs. Promotion: both edited wiki pages remain status: curated. Unresolved substantive issues: 0. [unverified] flags: 0.
  • Next sweep should also chase: whether NCT07782450 or NCT07363590 opens a separately reported RA cohort; the DGRh S3 journal publication and independent appraisal; external validation of the TNFA methylation model; RESET-RA and APIPPRA/ARIAA durability.

2026-10-07 — Uncited-dot grounding (grounder: Codex)

  • Disposition: started with 18 uncited dot rows; rewrote 6 (D-2, D-6, D-7, D-8, D-16, D-18); deleted 12; left no uncited row unresolved. Stable IDs and the five-column table were preserved. No previously cited dot row existed in this table, so none was edited.
  • Live evidence check: fetched the records used in the six rewritten rows and the three deletion findings through PubMed E-utilities, read their abstracts and publication types, and queried each DOI through Crossref works/<doi> for updated-by. None of the cited records had PubMed Retracted Publication/Expression of Concern types or a Crossref updated-by notice. One initial burst of Crossref requests returned HTTP 429; every affected cited DOI was retried successfully with individual requests. The supplied grounding map was a candidate list only and mapped no row to an OQ record; the 18 rows were traced through the condition's wiki and OQ text directly.
  • Deleted D-1: mucosal ACPA and synovial fibroblast findings are separately discussed, but the wiki supplies no quantified mucosal trajectory tied to a later fibroblast state; the proposed seeding claim would manufacture a link.
  • Deleted D-3: ORAL Surveillance quantifies cardiovascular harm, but the cited synovial material does not quantify JAK/STAT dependence as a benefit selector, let alone a benefit–risk interaction.
  • Deleted D-4: the wiki describes MUC5B/UIP and possible lung ACPA activity, but does not give a quantified, cited longitudinal lung-ACPA maturation finding to pair with MUC5B risk.
  • Deleted D-5: the proposed biopsy-during-withdrawal junction is already the published BIO-FLARE design, including optional baseline and flare synovial biopsies after DMARD cessation (PMID 34275488). The protocol is not proof that a predictor works, but it defeats this row's claim to an unattempted design.
  • Deleted D-9: systemic fracture risk is quantified, but the cited local erosion biology is largely a mechanistic review, not a measured synovial signature linked to systemic fracture risk.
  • Deleted D-10: work impairment is documented, but the condition's cited material does not quantify how route or monitoring burden changes job retention under clinically equivalent regimens.
  • Deleted D-11: the cervical-spine literature spans heterogeneous treatment eras and does not supply the contemporary deep-remission incidence comparison asserted by the pair.
  • Deleted D-12: patient priorities and trial endpoint selection are two descriptions of OQ-28's proposed core set, not independent findings forming a new junction.
  • Deleted D-13: the claim that prevention trials are confined to ACPA-positive people is false: TREAT EARLIER included ACPA-negative people, and its five-year increased-risk ACPA-negative subgroup had arthritis in 3/35 treated versus 10/31 placebo participants (PMID 42392130).
  • Deleted D-14: pharmacovigilance and tissue response pathotypes are each real, but no cited tissue-defined adverse-event estimate supports the proposed safety-selector side; the row merely speculates that one might exist.
  • Deleted D-15: screening-to-treatment stage shift is already the substance of OQ-30 and OQ-31; the row gives no independent pair of findings or tested effect modifier.
  • Deleted D-17: prevention-trial participation preferences have already been studied directly: a mixed-methods study compared 39 enrollees and 31 non-enrollees after risk and trial information, finding different perceived risk, benefit and medication concerns (PMID 36112074). The row restated this established research agenda.
  • PubMed absence searches (2026-10-07; counts are PubMed records, not universal absence):
  • D-2, 1 record (a general evidence review, no pre-arthritis biopsy selector trial): "rheumatoid arthritis"[tiab] AND ("at risk"[tiab] OR preclinical[tiab]) AND ("synovial biopsy"[tiab] OR "tissue signature"[tiab]) AND (abatacept[tiab] OR prevention[tiab]).
  • D-6, 3 records (including a secondary fatigue analysis, no trial allocating fatigue care by inflammatory D2T phenotype): "rheumatoid arthritis"[tiab] AND fatigue[tiab] AND (ultrasound[tiab] OR synovitis[tiab]) AND (randomized[tiab] OR randomised[tiab]).
  • D-7, 1 record (a strategy review, no trial of accelerated escalation with hospitalized infection): "rheumatoid arthritis"[tiab] AND ("treat to target"[tiab] OR "methotrexate escalation"[tiab]) AND glucocorticoid*[tiab] AND (hospitalized infection[tiab] OR serious infection[tiab]).
  • D-8, 0 records: "rheumatoid arthritis"[tiab] AND ("at risk"[tiab] OR preclinical[tiab] OR arthralgia[tiab]) AND ("periodontal treatment"[tiab] OR "periodontal therapy"[tiab]) AND (arthritis onset[tiab] OR incident arthritis[tiab] OR prevention[tiab]).
  • D-16, 0 records: "rheumatoid arthritis"[tiab] AND (ePROM[tiab] OR "remote monitoring"[tiab]) AND (discordance[tiab] OR discordant[tiab]) AND (management[tiab] OR intervention[tiab] OR trial[tiab]).
  • D-18, 0 records: "rheumatoid arthritis"[tiab] AND ("vagus nerve stimulation"[tiab] OR "neuroimmune modulation"[tiab]) AND ("synovial biopsy"[tiab] OR "cell state"[tiab] OR fibroblast[tiab] OR macrophage[tiab]) AND (response[tiab] OR responder[tiab]).
  • Limits and next check: these absence statements concern the PubMed-indexed combinations searched, not all trials or all databases. Europe PMC's search endpoint returned HTTP 503 during this pass, so it could not corroborate an absence; CENTRAL and SciELO were not reachable through the repository's documented routes. Re-run the six queries before treating any junction as still open, especially the periodontal and tissue-precision trials.
  • Broader PubMed synonym check (same date): D-2 returned 5 for "rheumatoid arthritis"[tiab] AND (preclinical[tiab] OR "at risk"[tiab]) AND (biopsy[tiab] OR synovial tissue[tiab]) AND (trial[tiab] OR random*[tiab] OR prevention[tiab]); none was a pre-arthritis biopsy-guided prevention trial. D-6 returned 5 for "rheumatoid arthritis"[tiab] AND fatigue[tiab] AND (ultrasound[tiab] OR inflammatory phenotype[tiab]) AND (trial[tiab] OR intervention[tiab]); the hits included observational or secondary fatigue analyses, not phenotype-assigned fatigue care. D-7 returned 8 for "rheumatoid arthritis"[tiab] AND glucocorticoid*[tiab] AND (escalation[tiab] OR "treat to target"[tiab]) AND infection[tiab]; none reported the proposed escalation-to-infection test. D-8 returned 13 for "rheumatoid arthritis"[tiab] AND periodont*[tiab] AND (prevention[tiab] OR "at risk"[tiab]) AND (trial[tiab] OR random*[tiab]); these included a qualitative acceptability study, not a periodontal-treatment trial with first-arthritis onset. D-16 returned 5 for "rheumatoid arthritis"[tiab] AND (remote monitor*[tiab] OR ePROM*[tiab]) AND (discordan*[tiab] OR pain[tiab]) AND trial[tiab]; no study allocated care using detected assessor discordance. D-18 returned 0 for "rheumatoid arthritis"[tiab] AND (vagus[tiab] OR neuromodulation[tiab]) AND (biopsy[tiab] OR "cell state"[tiab] OR pathotype[tiab]).

2026-10-07 — Independent audit of the uncited-dot grounding (auditor: Grok; author audited: Codex)

  • Scope: the six rows Codex rewrote (D-2, D-6, D-7, D-8, D-16, D-18) and the twelve rows it deleted (D-1, D-3, D-4, D-5, D-9, D-10, D-11, D-12, D-13, D-14, D-15, D-17). Rows outside this pass were not re-audited. Every PMID below was retrieved with PubMed E-utilities in this session. Retraction checks used PubMed publication types and the Crossref updated-by field. None of the records cited in the repaired table is a retracted publication or an expression of concern, and none has a Crossref retraction or concern notice.
  • Rows checked: 18. Citations that failed verification: 0. The six rewritten rows cite the papers they imply, in the right populations and the right direction. Absences that were stale: 0. Motivated anchoring: 0. Two rows were repaired in place, two deleted rows were restored because the deletion was difficulty rather than an unsupported claim, and no row was deleted by this audit.
  • Numbers re-read against the abstracts: ARIAA arthritis by 6 months was 4/49 (8%) versus 17/49 (35%), HR 0.14 (95% CI 0.04–0.47), in ACPA-positive arthralgia with MRI inflammation (PMID 38364841). R4RA nested cross-validation gave AUCs of 0.74 for rituximab and 0.68 for tocilizumab in 164 people; the full text states they were TNF-inhibitor-inadequate responders, and a stromal signature predicted multidrug resistance at AUC 0.69 (PMID 35589854). Pope's review states that at least one in six people with RA have severe fatigue (PMID 32385141). Xie's 23-study synthesis gives persistent inflammatory refractory disease in 47.1% (95% CI 33.3–61.4) of difficult-to-treat cases (PMID 41188120); the full text is not open access, so that proportion was checked against the abstract, which places it in the same synthesis. Medicare one-year hospitalized infection was 8.6% without glucocorticoids versus 11.0% (95% CI 10.6–11.5) at ≤5 mg/day (PMID 32956604). CAMERA, n=299, had at least one remission period in 50% versus 37% (PMID 17519278). The periodontitis meta-analysis of 13 studies reported OR 1.69 (95% CI 1.31–2.17) (PMID 32593704). The ePROM review was eight controlled studies, 4,473 people, seven including RA, SMD −0.15 (95% CI −0.27 to −0.03), most at high risk of bias (PMID 37204273). PRESERVE positive discordance was 194/762 (25.5%) at baseline and 189/763 (24.8%) at week 36 (PMID 27209012). RESET-RA ACR20 at three months was 35.2% versus 24.2% in 242 people (PMID 41429981). The synovial atlas used more than 314,000 cells from 79 donors and six cell-type abundance phenotypes (PMID 37938773).
  • Repair, D-7: the 11.0% figure was quoted without its interval. The interval is now in the row. The 8.6% figure has no interval in the abstract.
  • Repair, D-6: the unnamed "secondary fatigue analysis" is the ARCTIC analysis, in which clinically relevant fatigue fell from 69% to 38% at 24 months (PMID 34389605). It does not assign fatigue care by a difficult-to-treat inflammatory phenotype. The row now names it.
  • Absences re-run: the six PubMed queries were repeated, then broadened, and the same questions were put to Europe PMC, which answered in this session (Codex had recorded HTTP 503). ClinicalTrials.gov was queried for periodontal treatment, vagus-nerve stimulation plus synovial biopsy, pre-arthritis synovial-biopsy prevention, remote-monitoring discordance, and fatigue phenotype trials.
  • D-2: PubMed exact count 1, the EULAR 2025 efficacy review (PMID 41997804), which includes seven prevention studies and biopsy-guided treatment in established disease, not a pre-arthritis biopsy selector. Europe PMC returned the same single record. ClinicalTrials.gov synovial biopsy prevention rheumatoid arthritis at risk returned no studies.
  • D-6: PubMed exact count 3. The relevant hit is ARCTIC/Holten, above. A phenotype-assignment query returned 0. Europe PMC returned 0 for a difficult-to-treat fatigue trial.
  • D-7: PubMed exact count 1, a 2015 narrative review (PMID 26697765). A trial-filtered query of tight control, glucocorticoids and infection returned 0. Europe PMC returned 0. CAMERA-II bone data (PMID 25228126) do not test hospitalized infection.
  • D-8: PubMed exact count 0. The near-miss is a qualitative acceptability study of periodontal treatment in CCP-positive at-risk people (PMID 37156573; Crossref records a correction, not a retraction, so its text was not used for a number). ClinicalTrials.gov returned no at-risk arthralgia study. Records that did return are in established RA: NCT05502601 enrols people who already meet RA criteria and treats periodontitis; NCT03087240 requires active RA (DAS28 > 3.2); NCT05198271 enrols established RA with modifiable risk factors. Europe PMC hits were the qualitative study, established-RA treatment reviews, and a microbiome review, not an onset trial.
  • D-16: PubMed exact count 0. Europe PMC returned 0. A CareRA analysis links in-clinic discordance to less consistent treat-to-target escalation (PMID 38127964). It is not a remote-monitoring intervention.
  • D-18: PubMed exact count 0. Broader hits are animal experiments, a receptor-expression study, the RESET-RA trial, its interim safety report (PMID 38475923) and a commentary (PMID 38566215). None pairs stimulation response with synovial cell-state profiling. ClinicalTrials.gov vagus nerve rheumatoid arthritis synovial biopsy returned no studies. Europe PMC returned 0 on the tight query.
  • Deletions that stand:
  • D-3. A synovial JAK/STAT query did not yield a human study that quantifies tissue pathway dependence as a benefit selector. The top hits were treatment-effect or non-human papers. Deleting the row rather than inventing that selector was right.
  • D-4. The MUC5B side is real: among people with RA, rs35705950 had adjusted OR 3.1 (95% CI 1.8–5.4) for interstitial lung disease and 6.1 (2.9–13.1) for usual interstitial pneumonia (PMID 30345907). Lung-resident mutated ACPA clones were found in 3 at-risk people and 9 with early RA (PMID 37192126), which is a small descriptive finding, not a quantified maturation trajectory. An inception cohort of 841 people already relates ACPA fine specificities and MUC5B to later pulmonary fibrosis, with 50 fibrosis events (PMID 35532073). That is not pre-divergence airway sampling, but it is enough that the original pair was not an unconnected junction waiting on a missing citation. Deletion stands.
  • D-5. BIO-FLARE is a published DMARD-cessation protocol with an optional baseline synovial biopsy and a biopsy at flare (PMID 34275488; ISRCTN 16371380). Preliminary results now report clinical predictors of flare and drug-free remission in 121 people, with flare by week 24 in 52.3% (95% CI 43.0–61.7) (PMID 40204305), and patient-reported scores as flare warnings (PMID 40169281). Neither results paper reports a synovial predictor. Codex cited the protocol and not these results. The deletion still stands: the design is no longer unattempted. The tissue half remains unpublished.
  • D-9, D-10 and D-12. Fracture risk, work loss and patient priorities are real topics, but the paired sides Codex was asked to cite were a mechanistic erosion review, an unquantified route-versus-job-retention contrast, and a restatement of OQ-28. Those deletions are acceptable.
  • D-11. A contemporary deep-remission incidence comparison is not what the cohorts supply. FIN-RACo found anterior atlantoaxial subluxation in 13/149 (9%) at five years, with 11 in the single-DMARD arm and 2 in the combination arm (PMID 19132793). A later ten-year cohort was enrolled in 2001–2002 (PMID 39501065). Deletion of the modern-remission claim stands.
  • D-13. The premise that prevention trials are confined to ACPA-positive people is false. In TREAT EARLIER's five-year increased-risk ACPA-negative subgroup, rheumatoid arthritis developed in 3/35 treated versus 10/31 placebo participants (HR 0.24, 95% CI 0.07–0.87) (PMID 42392130). Deletion stands. The remaining enrichment question is already OQ-1.
  • D-15. Screening yield and antifibrotic effect are independent findings, so "no independent pair" is too strong. The junction itself is already OQ-30 and OQ-31. Deletion as a duplicate open question stands.
  • D-17. A mixed-methods study already compared 39 enrollees and 31 non-enrollees for a hydroxychloroquine prevention trial and found differences in perceived risk, benefit and medication concern (PMID 36112074). The dots are not unconnected. A co-design trial is still undone, but that is a narrower question than the row Codex deleted.
  • Deletions reversed, because they were difficulty:
  • D-1 was deleted because the wiki had no quantified mucosal trajectory already tied to a fibroblast state. That missing tie is the junction. Sputum IgA ACPA tracks citrullinated-histone NETs in 49 at-risk people (PMID 34369110), and fibroblast-activation PET predicted onset in 18 ACPA-positive people with arthralgia (5/18; HR 4.58, 95% CI 1.38–15.22) (PMID 41314671). PubMed and Europe PMC queries joining mucosal ACPA with a fibroblast trajectory returned 0. The row is restored with those two findings.
  • D-14 was deleted because no tissue-defined adverse-event estimate exists. That estimate is the junction, not a prerequisite for the two dots. ORAL Surveillance did not show noninferiority of combined tofacitinib doses versus a TNF inhibitor for MACE (HR 1.33, 95% CI 0.91–1.94) or cancer (HR 1.48, 95% CI 1.04–2.09) against a margin of 1.8 (PMID 35081280). R4RA's nested model predicted multidrug resistance at AUC 0.69 (PMID 35589854). A PubMed query for pathotype or synovial biopsy plus serious infection, cancer or MACE returned seven records; the titles are methylation, checkpoint-inhibitor synovitis, case reports and erosion markers, not a pathotype safety selector. The row is restored.
  • What D-4 would have said if the lung side had been a trajectory: MUC5B rs35705950, adjusted OR 3.1 (95% CI 1.8–5.4) for RA-ILD and 6.1 (2.9–13.1) for usual interstitial pneumonia (PMID 30345907), set against lung ACPA clones before and at diagnosis (PMID 37192126), with PMID 35532073 named as the study that already links blood ACPA breadth and MUC5B to fibrosis after onset. It was not restored, for the reason in the deletion note above.
  • Judgement: the dot table can now be trusted by the novelty gate. The six rows Codex kept were the right papers and the right numbers, with one omitted confidence interval and one unnamed near-miss, both repaired. Two further junctions were real and had been dropped because the connecting study does not yet exist; they are back in the table. The other ten deletions should stay deleted. Trust is limited to what was searched: PubMed, Europe PMC and ClinicalTrials.gov. CENTRAL and SciELO were not reachable, so these are not universal absences. The Xie proportion is abstract-checked because the full text is closed. The restored fibroblast-PET row is an 18-person cohort and should be read at that size.

2026-10-11 — Periodic literature sweep (author: Grok, grok-4.6)

Window. Publications since 2026-08-28 (prompt date_from; last RA literature sweep). Identifier retrieval was live NCBI E-utilities (esearch/esummary/efetch) and ClinicalTrials.gov v2 /studies/{NCT}. PubMed MCP tools were not available; E-utilities were the live equivalent. Unpaywall skipped (LIFEBANK_CONTACT_EMAIL unset). Crossref queried without mailto. Query manifest: literature/sweeps/2026-10-11.json.

Sources actually queried. PubMed topic searches + title-restricted RA + publication-type filters (RCT, systematic review, guideline) + CDSR-per-topic + dated follow-up queries (DGRh S3, TNFA methylation, ARIAA/APIPPRA, RESET-RA); Europe PMC REST (FIRST_PDATE:[2026-08-28 TO 2026-10-11]; first burst HTTP 503, later retry hitCount 1747 MEDLINE/PMC records, 0 records without PMID, 46 preprints, 0 CDSR); ClinicalTrials.gov v2 (all 31 previously cited NCT IDs plus first-posted/updated/results-posted RA queries); ISRCTN API (APIPPRA → ISRCTN12680338, lastUpdated 2026-01-23); Crossref updated-by for every added DOI; EULAR, ACR, AWMF 060-004, DGRh press release, NICE search, NRAS, Arthritis Foundation, Arthritis Society Canada, Arthritis Foundation of South Africa. Not queried / blocked: Cochrane Library/CENTRAL (HTTP 403); SciELO (HTTP 403); LILACS not separately reached. WHO ICTRP portal opened (HTTP 200) but is JS-driven so not counted as searched; ChiCTR HTML search page opened, not a programmatic result set. Absences below are absences in PubMed / Europe PMC / ClinicalTrials.gov / ISRCTN, not universal absences. No regional-literature absence claim is made.

Cochrane. "Cochrane Database Syst Rev"[jour] × RA topic groups in the window: 0 PubMed records and 0 Europe PMC records. No CDSR review in this window superseded or contradicted anything this condition asserts. Latest undated CDSR RA hits (outside window, not ingested): Thomas 2026 Jul 13 network meta-analysis after b/tsDMARD failure (PMID 42440279); Li 2026 May 20 protocol on non-TNF tapering (PMID 42159078 — Crossref DOI/title mismatch, not used).

Follow-ups from 2026-08-31. NCT07782450 remains NOT_YET_RECRUITING with RA eligible only in multi-disease parts 1a/1b; NCT07363590 remains a 21-person SLE/RA phase 1 with no separately reported RA cohort. DGRh S3: still no PubMed-indexed journal article; AWMF 060-004 and the DGRh press release remain the live documents. TNFA-methylation methotrexate-response external validation: PubMed 0 in window. RESET-RA durability: PubMed 0; NCT04862117 still ACTIVE_NOT_RECRUITING (n=14), last update 2026-09-25. APIPPRA durability: no new outcome paper; Jasenecova 2026 used the APIPPRA cohort to externally validate risk-prediction rules.

What survived triage. Thirteen papers, one posted-results module, and seven new NCT IDs. No landmark note: Park is a 24-week open-label LDA tapering trial; Bucci 2026 is a 15-person series; ATI-450 is a negative phase 2 without a journal report in the window.

Record Why it survived
Park 2026, PMID 42839723 348-person RCT; b/tsDMARD tapering non-inferior for 24-week DAS28-ESR LDA (RD −1.43%, 95% CI −8.09 to 5.23)
Zhang 2026, PMID 42836514 27 publications, 5,262 people; tapering RR flare 1.56, withdrawal RR 2.23
Tandi 2026, PMID 42843067 19-study NMA; stopping bDMARDs while continuing csDMARDs RR 3.06 vs maintenance
Russell 2026, PMID 42755040 29 trials, 13,387 MTX-naive; first-line b/ts+MTX DAS28 remission RR 1.62 at 6 months
Solomon 2026, PMID 42248165 Lancet Respir Med RA-ILD expert consensus; does not replace ACR/CHEST
Bucci 2026, PMID 42595657 15-person low-dose blinatumomab series; 9/15 CDAI LDA, 14/15 later flared
Jasenecova 2026, PMID 42676706 14 risk rules externally validated in APIPPRA; 10/14 C<0.6
Bechman 2026, PMID 42546722 IPD MA 11,883 people; class 3 obesity ACR20 RR 0.78 on JAK inhibitors
Ibis 2026, PMID 42823079 1,304 matched ≥65 pairs; observational null JAK vs TNF for MACE/VTE/cancer/death
Smolen 2026, PMID 42720931 JAMA review; adult prevalence 0.53% worldwide / 0.74% US (different frame from GBD)
Rojas-Serrano 2026, PMID 42722091 39-person tofacitinib RA-ILD phase II (NCT05246293); 3 deaths, 15/39 PPF/death
Webers 2026, PMID 42485861 Residual disease SR; 5–25% of RA in remission still have swollen joints
van der Helm-van Mil 2026, PMID 42751359 Prevention trial-design review
NCT05279417 Results posted 2026-09-28; week-12 ACR20 42.2%/34.7% vs placebo 46.5% (negative)

Dropped (honest). Herbal/nanoparticle/animal series; hand-surgery and arthroplasty outcomes; qualitative significant-other and US physical-therapy studies; sacral-nerve-stimulation n=21 pilot (Crossref issued 2026-07-09, outside window); SELECT-EARLY/filgotinib Japanese 5-year subgroups (31 Aug print-date edge, not practice-changing); ultra-low-dose SC rituximab n=35 PK study (AUC non-inferiority not definitive); MTX vaccine-hold narrative SR of five studies (does not replace Park 2018 numbers); Yap MTX transcriptomic ML (not externally validated); Imai SORAIRO RCT (PMID 41763966, issued April 2026, outside window; only correspondence hit the window); five PubMed records whose attached DOIs resolved on Crossref to unrelated older titles (PMIDs 42832028, 42684557, 41733109, 42159078, 42839246) — quarantined, not cited.

ClinicalTrials.gov. All 31 previously cited NCT IDs resolved. No overall-status changes. Several last-update dates moved (including NCT04862117 2026-09-25, NCT06570798 2026-10-05, NCT07782450 2026-10-01) without status change. Results newly posted: NCT05279417 (ATI-450, negative); NCT04504136 is IBD and was excluded. Added: NCT07276581 (AZD1163 phase 2, n=320), NCT07856511 (Proximod phase 3, n=590), NCT07796516 (telitacicept refractory RA, n=420), NCT07849673 (ILD screening CDS, n=800), NCT07857993 (DETECTRA, n=320), NCT07802678 (csDMARD discontinuation on TNFi, n=202), NCT05279417. NCT05246293 remains UNKNOWN on the registry while Rojas-Serrano 2026 reports the 39-person phase II.

Guidelines / statistics / patient voice. EULAR, ACR, AWMF 060-004 and DGRh pages retrieved; no new society pharmacologic RA guideline in the window. Solomon 2026 added as expert consensus. Italian SIR RA-ILD (PMID 41247141) still lacks item-level recommendations in the PubMed abstract. JAMA adult-prevalence figures added beside GBD, not averaged. NRAS, Arthritis Foundation, Arthritis Society Canada and AFSA pages retrieved; no post-cutoff campaign required a patient-voice change.

Retraction checks. All 13 added PMIDs: PubMed publication types show neither retraction nor expression of concern; Crossref updated-by empty for each DOI; Crossref titles matched PubMed titles. Window-wide retraction/expression-of-concern search: 0. A retraction notice for a plumbagin paper (PMID 42821489) was seen and not used as evidence.

Pages edited and demoted to draft: treat-to-target-and-remission.md, clinical-trials-landscape.md, biologic-dmards.md, jak-inhibitors-and-targeted-therapy.md, guidelines.md, ra-associated-interstitial-lung-disease.md, preclinical-autoimmunity-and-prevention.md, difficult-to-treat-and-refractory-ra.md. Bibliography 262→275. OPEN-QUESTIONS OQ-1, 4, 5, 8, 11, 25, 30, 32. INDEX counts/statuses updated. No new deep note.

Decision noted. Park's non-inferiority is 24-week LDA, not two-year radiographic non-inferiority; Zhang/Tandi remain the longer-horizon caution. Ibis is not written as a rebuttal of ORAL Surveillance. ATI-450 numbers are from the ClinicalTrials.gov results module, not a journal article. Rojas-Serrano is single-arm. Claims are abstract-checked; journal full texts were not open on Europe PMC for the added records.

Next sweep should chase. Park full text and any radiographic/extension data; Zhang/Tandi full-text subgroup tables; DGRh S3 journal publication; Italian SIR item-level recommendations; NCT05279417 journal report; NCT07276581 / NCT07856511 / NCT07796516 / NCT07802678 status; whether NCT07782450 or NCT07363590 opens a separately reported RA cohort; RESET-RA (NCT04862117) durability; APIPPRA/ARIAA further follow-up; Thomas 2026 Cochrane NMA (PMID 42440279, July, outside this window); TNFA-methylation external validation; D-8 periodontal onset trial (NCT07844564 is established RA). Auditor must be a different engine.

2026-10-18 — Periodic literature sweep (author: Claude, claude-opus-5-5)

Window. Publications since 2026-08-28 (prompt date_from), run by both PubMed publication date and Entrez date to 2026-10-18. The 2026-10-11 sweep already covered 2026-08-28 to 2026-10-11, so every record already cited in this condition or listed in the 2026-10-11 manifest was excluded from triage. Query manifest: literature/sweeps/2026-10-18.json.

Sources queried. PubMed E-utilities: 20 topic queries (classification, epidemiology, prevention, genetics/mucosa, synovium, treat-to-target/tapering, csDMARDs, biologics, JAK, D2T/CAR-T, extra-articular, RA-ILD, biomarkers, patient experience, safety), RCT, SR/MA, guideline, CDSR and retraction filters, plus 13 targeted or follow-up queries. Europe PMC REST: 386 RA-titled records in the window; 66 without a PMID, all PMC copies awaiting indexing or preprints (none admitted); 53 preprints; 0 CDSR. ClinicalTrials.gov v2: all 43 NCT IDs already present anywhere in the condition tree (38 cited in wiki/literature files), plus RA studies first posted since 2026-10-11 (5) and results posted since 2026-08-28 (2). ISRCTN API: 187 RA records, sorted by last update. Crossref updated-by was checked for every added DOI, without mailto. Unpaywall was skipped because LIFEBANK_CONTACT_EMAIL is unset. Web pages checked: ACR RA guideline page (still lists 2021 as latest), EULAR recommendations page, AWMF 060-004, NRAS news (including the NEIAA 2026 report PDF), Arthritis Foundation news. Arthritis Society Canada returned HTTP 403.

Not queried / blocked. Cochrane Library/CENTRAL (HTTP 403), SciELO/LILACS (HTTP 403), and WHO ICTRP, whose JS portal could not be searched programmatically. bioRxiv/medRxiv were reached only through Europe PMC SRC:PPR. Every absence below is an absence in PubMed / Europe PMC / ClinicalTrials.gov / ISRCTN only. The Argentine guideline's evidence base includes LILACS, which this repository cannot re-check.

Screening. 1,171 unique PMIDs across all queries; 681 were new to this condition and to the 2026-10-11 manifest; the 384 of those with RA in the title were screened by title and publication type, and about 20 by abstract.

Cochrane. Neither PubMed nor Europe PMC returned a CDSR record in the window. The 2026-10-11 sweep logged the living Cochrane NMA of DMARDs after b/tsDMARD failure (Thomas 2026, PMID 42440279; July 2026) but did not ingest it. It was ingested in this sweep. It does not contradict the wiki. It reports high-certainty ACR50 benefit over placebo for nine therapies, moderate-certainty benefit for two, and no significant pairwise differences between active drugs (low certainty). That supports the D2T page's existing "no universal sequence" statement and adds a quantified anchor to it.

What survived triage (20 PMIDs, 3 NCT IDs, 2 ISRCTN IDs; no landmark note).

Record Disposition
Albach 2026, PMID 42661098 (COMPARE CD19 CAR-T phase 1, n=6; NCT06475495) D2T page, trials landscape, OQ-25, statistics. No deep note: an uncontrolled six-person phase 1 is a first, not a landmark
Thomas 2026, PMID 42440279 (Cochrane NMA) D2T page, OQ-10
Nordberg 2026, PMID 42881069 (ARCTIC 10-year) treat-to-target strategy table, statistics
McClenahan 2026, PMID 42877103 (CARDINAL-RA target-trial emulation, 18,757 pairs) JAK page, OQ-5, statistics. PubMed indexes it as an RCT, but it is observational; this is flagged on the page
Tsai 2026, PMID 42869207 (JAK cycling vs biologic switch) D2T page, OQ-10, statistics
Antoniou 2026, PMID 40907995 / 40912974 (ERS/EULAR CTD-ILD guideline) Registry catch-up: published January 2026 and missed by every earlier sweep. Added to REGISTRY, guidelines page and RA-ILD page
Garcia-Salinas 2026, PMID 42888356 (RA-ILD guideline comparison) guidelines page, RA-ILD page, OQ-7. The monitoring intervals are attributed to this review's abstract, not to the guideline texts
Kouna 2026, PMID 42873714 (lung ultrasound + DLCO rule-out, n=103) RA-ILD page, OQ-7, statistics
Santos Moreno 2026, PMID 41240967 (PANLAR); Schneeberger 2026, PMID 42507767 / 42767235 (Argentine Parts I/II) REGISTRY, guidelines page, OQ-1. PANLAR and Part I are catch-ups
Wu 2026, PMID 42875583; Hocaoğlu 2026, PMID 42378266 preclinical page, OQ-1
Zhao 2026, PMID 41847932 (CHIP and seropositive RA) statistics and bibliography only. The genetics page is curated and was not edited for one association
Abhishek 2026, PMID 42663270 (NIHR HTA monitoring intervals) bibliography only. A candidate for conventional-dmards.md (curated) at its next audit
Hagen 2026, PMID 42616666; Özüpek 2026, PMID 42863421; Raffray 2026, PMID 42342555; Jung 2026, PMID 42557168 bibliography only. The DASH RCT lost 33 of 82 participants, too much attrition to warrant a page edit
NCT07873931 (BMS-986553 in vivo CD19 CAR-T, first posted 2026-10-14); ISRCTN44988547, ISRCTN94259329 (3TR precision trials) trials landscape
NEIAA 2026 State of the Nation report (non-journal) statistics: EIA seen within 3 weeks 42%, DMARD within 6 weeks 53%, 3-month remission 33%. The cohort is all early inflammatory arthritis, not RA alone, and the row says so

Dropped (honest). Mechanistic animal and herbal papers; the miRNA and transcriptomic network papers; Preglej's MTX single-cell study (6+22 patients); the Italian, JAGUAR and Turkish JAK registry retention reports; a tofacitinib sociodemographic-index post hoc analysis; a D2T cross-sectional study; the olokizumab PK/PD simulation; correspondence and editorials; and the PD-1 agonism (PMID 42810368) and EMPEROR-RA (PMID 42867205) records, which have no PubMed abstract to check against. The obinutuzumab erratum (PMID 42891068) corrects a case report that is not cited here. No preprint was admitted.

ClinicalTrials.gov. No overall-status changes among tracked IDs since 2026-10-11. NCT07137598 updated its record on 2026-10-16 without a status change. The only results posted since 2026-08-28 remain NCT05279417 (already recorded) and NCT04504136 (IBD; excluded). NCT06475495 is ACTIVE_NOT_RECRUITING with enrollment 13, and its phase 2 randomizes against rituximab.

Retraction checks. For all 20 added PMIDs, PubMed publication types show no retraction or expression of concern, Crossref updated-by is empty, and Crossref titles match PubMed. The window-wide PubMed retraction/EoC search (RA in title/abstract) returned 0.

Follow-ups from 2026-10-11, re-run. No journal record in PubMed for: DGRh S3 (0), the ATI-450 journal report (0), RESET-RA durability (0 relevant), APIPPRA/ARIAA follow-up (0), TNFA-methylation validation (0), Park radiographic/extension data (0), a periodontal-treatment onset trial in at-risk people (0), or Italian SIR RA-ILD item-level recommendations (0). Thomas 2026 Cochrane NMA: ingested (above).

Pages edited: difficult-to-treat-and-refractory-ra.md, clinical-trials-landscape.md, treat-to-target-and-remission.md, jak-inhibitors-and-targeted-therapy.md, ra-associated-interstitial-lung-disease.md, preclinical-autoimmunity-and-prevention.md, guidelines.md. All seven were already draft and remain draft, with last-curated set to 2026-10-18. Also updated: BIBLIOGRAPHY (275→295), REGISTRY (28→32 documents), STATISTICS (new 2026-10-18 table), OPEN-QUESTIONS (OQ-1, 5, 7, 10, 25 sharpened; none resolved or newly opened), INDEX. All claims were checked against abstracts only, not full texts. CARDINAL-RA, Hocaoğlu, Zhao and the HTA report have PMC full texts that were not opened.

Next sweep should chase. The COMPARE phase 2 (CAR-T vs rituximab) and NCT07873931 status. CARDINAL-RA full text, checking whether the agent-level mortality estimates and the claims-data MACE definition hold up. Item-level RA-ILD recommendations from the ERS/EULAR full text, which would replace the Garcia-Salinas abstract-level intervals. A prospective validation of the LUS+DLCO rule-out. External validation of the activated-naive-B-cell signature. Whether conventional-dmards.md should take the HTA monitoring-interval result at its next audit. Still open from before: DGRh S3, the NCT05279417 journal report, RESET-RA durability, and the periodontal onset trial (D-8). The auditor must be a different engine from Claude.

2026-10-25 — Periodic literature sweep (author: Codex)

Window and method. Requested publication window 2026-08-28 through 2026-10-25, with a parallel broad Entrez-date search for delayed indexing. Read the previous log, index and source rules before editing. Search provenance, returned identifiers, translations, per-record retraction checks and registry snapshots are preserved in the sweep manifest. No publishing or git commands were run.

Sources actually queried. PubMed E-utilities: 36 main searches, comprising broad publication/Entrez queries, 16 major-topic searches, 16 matching Cochrane topic checks, an all-RA Cochrane check and a retraction/EoC check; 11 additional follow-up/corrected searches, followed by live efetch. Europe PMC REST: 18 searches (broad, all 16 topics and Cochrane), with full-text availability metadata and selected fullTextXML retrievals. ClinicalTrials.gov API v2: exact fetches for every NCT identifier in the condition Markdown, plus first-posted/last-updated searches since 18 October and first-results-posted since 28 August. ISRCTN XML API: all 187 RA-condition records, including the prevention and precision registrations. Crossref works: all six additions and both reused sources supporting new substantive claims, with no mailto. Official/public web checks: ACR RA guideline page; EULAR management recommendations; BSR guideline index; CRA living-guideline page and journal recommendation table; AWMF announcement search result and attempted register access; NRAS home/news listing; Arthritis Foundation news and World Arthritis Day release; Deutsche Rheuma-Liga guideline page and patient version. New burden data were checked against the Springer full-text incidence table.

Coverage limits. Cochrane Library/CENTRAL was not searched (documented HTTP 403); CDSR was searched through PubMed and Europe PMC. SciELO and LILACS could not be searched (documented HTTP 403), so regional transportability and Latin American coverage remain incomplete. A live WHO ICTRP query returned HTTP 403; ISRCTN supplied the accessible non-US registry sweep. AWMF's register denied bot access; its guideline document was not retrieved in this session. Initial guessed ISRCTN /api/studies endpoints returned 404; the documented XML /api/query/format/default endpoint succeeded. Unpaywall was skipped because LIFEBANK_CONTACT_EMAIL is unset; no personal address was obtained or sent. Preprints were encountered through Europe PMC, not direct preprint-server APIs. No broad absence claim is made beyond these searched sources.

Screening and Cochrane. The main PubMed queries returned 1,261 distinct identifiers; efetch returned 1,254 journal-article records and seven book records. Ninety-two identifiers were new to the condition and both preceding manifests (91 journal records and one book record); all were title-screened, with selected abstracts examined. Europe PMC returned 2,615 broad-window records, including 523 without a PMID; 136 no-PMID records with arthritis/DMARD/interstitial terms in the title were screened. These included preprints, PMC duplicates/records awaiting linkage, low-priority reports and unrelated arthritis material; no DOI-only or preprint record was admitted. Neither PubMed's 17 CDSR searches nor the Europe PMC CDSR search returned a review in the requested window. The previously ingested living Cochrane review was not replaced. The window-wide PubMed retraction/EoC query returned zero; this is a PubMed-only finding. An unfielded ALTO follow-up search produced unrelated hits and was replaced by a fielded query; its initial result was not used to claim an absence.

What survived triage: six bibliography additions; no landmark note.

Record Disposition
Nikolic 2026, PMID 41921998 Canadian living guideline update #3, published August: explicitly labelled pre-window catch-up. Full recommendation table supports gradual tapering of either drug group first in combination treatment, conditional on reliable rescue access. Added to guidelines, registry and OQ-4
Corrales 2026, PMID 42907925 Blood functional/proteomic treatment-response signatures. Full-text limitations, rather than an efficacy claim, added to biomarkers and OQ-3: observational allocation, no independent external validation of the classifier, no individual counterfactual test
Martins 2026, PMID 42909770 Two RA-specific incidence rows in STATISTICS, verified in full-text Table 2: separate Optum EHR and MarketScan estimates with full-base-population denominators; no CTD-wide abstract rate relabelled as RA
Curran 2026, PMID 42894571 Lifestyle systematic review, full text read: bibliography only; heterogeneous low/very-low-certainty adjunctive evidence does not warrant a recommendation change
Tariq 2026, PMID 42918701 Imminent-RA interferon/cell-state study: bibliography only, abstract-screened; prioritize full text and independent validation
Fujisawa 2026, PMID 42900496 Reason-stratified ANSWER JAK-discontinuation cohort: bibliography only, abstract-screened; full-text comparison with existing cycling/switching evidence deferred

Prior-sweep follow-ups completed. Opened the ERS/EULAR CTD-ILD full text (Antoniou, PMID 40907995). Replaced the RA-ILD page's secondary-source screening summary with the explicit recommendations against replacing HRCT with PFTs/ultrasound, and added the RA-specific PFT/HRCT follow-up intervals with their conditional/very-low-certainty labels and table locations. Opened CARDINAL-RA in full (McClenahan, PMID 42877103): clarified that its primary MACE outcome excludes cardiovascular death and that the landmark/EHR design includes first follow-up events even with prior events; sharpened OQ-5 and the statistics label. Neither finding resolves the underlying clinical question. OQ-7 now explicitly places the proposed ultrasound rule-out pathway against the guideline's replacement restriction.

Trials and results. All 46 identifiers found anywhere in the condition Markdown resolved live; 42 are in the research corpus, with the remainder retained only in historical logs. The 45 shared with the prior manifest showed no overall-status, enrollment-count or posted-results-availability change. NCT07873931 remains recruiting without posted results. NCT07412704 updated on 2026-10-23 and records an actual study start while remaining recruiting with estimated enrollment 84. All eight result-bearing tracked records have last-update dates before the previous sweep, so no subsequent registered results revision is indicated; results-section hashes are retained for the next comparison. The broad CT.gov searches returned zero first postings since 18 October, 26 updates and two first results postings since 28 August; the latter remain the already-recorded ATI-450 trial (NCT05279417) and the unrelated IBD study (NCT04504136, excluded). ISRCTN's latest update is the mixed-inflammatory-arthritis MAMA pregnancy strategy trial (ISRCTN89126536, 21 October); no outcome result was ingested. The two tracked 3TR registrations and ALTO have no newer update than their preceding recorded dates.

Guidelines, burden and patient material. ACR's official RA page still lists 2021 as its latest pharmacologic guideline. The new registry addition is the Canadian catch-up; no new October replacement for the existing EULAR/DGRh management documents was verified. The DGRh journal query returned two irrelevant hits rather than a relevant guideline report. Added two public patient-source annotations: the German S3 medication-guideline patient version (newly catalogued; release date unestablished) and the Arthritis Foundation's 5 October World Arthritis Day campaign. The latter covers arthritis broadly; its prevalence claim was not imported into RA statistics. NRAS's visible latest substantive audit item was already captured by the previous sweep.

Dropped/deferred. Small uncontrolled methotrexate route-switch cohorts; the very small Kenyan miRNA case-control substudy; cross-sectional food-insecurity findings; case reports; animal/nanoparticle/herbal and in-silico mechanistic reports; and non-RA returns. Newly indexed RA-ILD mortality and long-term remission cohort records lacked PubMed abstracts and publisher full texts returned HTTP 403, so no result was inferred from their titles. Targeted PubMed follow-ups did not identify a new ATI-450 journal report, RESET-RA durability report, APIPPRA/ARIAA follow-up, TNFA-methylation external validation, periodontal intervention preventing RA onset, or Park tapering extension beyond the existing report; several queries returned older or unrelated records, detailed in the manifest. These are bounded PubMed findings, not claims of worldwide absence.

Retraction and citation integrity. All six additions and both reused research sources supporting new claims were live-fetched through E-utilities; none has a PubMed retraction/EoC publication type or a Crossref updated-by notice, and titles match. Existing historical citations retained within edited lines were additionally re-fetched live. New substantive wiki/OQ claims were checked in full text, with brief supporting quotations and exact table locations in the manifest. The two abstract-only additions remain bibliography entries.

Files and status. Edited five wiki pages: guidelines, RA-ILD, biomarkers/tissue precision, JAK therapy and trials landscape. Each is draft with last-curated: 2026-10-25; biomarkers was demoted from curated. Also updated bibliography (295→301), guideline registry (32→33), statistics (two incidence rows plus a clarified endpoint label), patient-source annotations, OPEN-QUESTIONS (OQ-3/4/5/7 sharpened; none resolved or newly opened), INDEX, this log and the exact legal manifest path. Nine wiki pages remain curated and nine are draft. Independent audit remains pending; this author has not promoted its own work.

Validation. Checked all 13 changed files: added-line PMID/NCT provenance, new local links, edited table structure, reference coverage/orphans, bibliography and registry counts, and wiki status/date metadata; zero failures. All 1,365 additional PMID identifiers encountered through Europe PMC also resolved through live E-utilities; this was an identifier cross-check, not another evidence admission. The manifest records the checks. No full-site build was run because publishing is handled separately.

Next sweep should chase. Full texts and outcome tables for the newly indexed RA-ILD mortality and long-term remission cohorts; independent validation and locked prospective testing of the Corrales classifier; the Tariq imminent-RA study's full text; reason-specific safety/retention confounding in the ANSWER report; COMPARE phase 2 and the two BMS cellular-therapy records; whether MAMA posts substantive results. Carry forward the item-level Italian RA-ILD extraction, prospective LUS/DLCO rule-out validation, activated-naive-B-cell validation, the HTA monitoring-interval review for conventional DMARDs, the DGRh journal report, ATI-450 publication, RESET-RA durability and periodontal-onset prevention. The auditor must use a different engine from Codex.

2026-11-01 — Periodic literature sweep (author: Grok, grok-4.6)

Window and method. Requested publication window 2026-08-28 through 2026-11-01, with a parallel Entrez-date search for delayed indexing. Prior sweeps already covered this window through 2026-10-25; new-to-search identifiers were triaged first. Identifier retrieval was live NCBI E-utilities (esearch/efetch) and ClinicalTrials.gov v2 /studies/{NCT}. Unpaywall skipped (LIFEBANK_CONTACT_EMAIL unset). Crossref queried without mailto. Query manifest: literature/sweeps/2026-11-01.json. No publishing or git commands were run.

Sources actually queried. PubMed E-utilities: 18 topic searches (broad publication date, broad Entrez date, 16 major topics), 18 Cochrane/retraction filters, 15 follow-up or publication-type filters, then live efetch of 423 new-to-prior-search records. Europe PMC REST: 19 searches (broad, 16 topics, Cochrane, preprints); broad hitCount 999 with pageSize 1000, treated as a capped page. ClinicalTrials.gov v2: exact fetches for all 42 research-corpus NCT IDs, plus first-posted/last-updated since 25 October and results-posted since 28 August. ISRCTN XML API: condition "Rheumatoid arthritis" (187-record query) and five tracked identifiers (ISRCTN44988547, 94259329, 12680338, 16371380, 89126536). Crossref updated-by for every added DOI. Official/public web checks: ACR RA guideline page, EULAR management recommendations, BSR guideline index, NICE NG100, AWMF 060-004 (HTTP 200 shell), DGRh home, NRAS news listing, Arthritis Foundation news, Arthritis Foundation of South Africa home, Deutsche Rheuma-Liga home. CRA living-guideline URL used in a prior sweep returned HTTP 404.

Coverage limits. Cochrane Library/CENTRAL was not searched (HTTP 403); CDSR was searched through PubMed and Europe PMC. SciELO and LILACS could not be searched (HTTP 403), so no regional-transportability absence claim is made. WHO ICTRP portal returned HTTP 200 but is JS-driven and was not counted as a programmatic search. Unpaywall skipped. Preprints were reached only through Europe PMC SRC:PPR; none admitted. Every absence below is an absence in PubMed / Europe PMC / ClinicalTrials.gov / ISRCTN.

Screening and Cochrane. PubMed union 1,675 identifiers; 412 were new to the bibliography and the 2026-10-25 search lists; 423 high-PMID or new records were fetched. Neither PubMed's 18 CDSR searches nor Europe PMC's CDSR search returned a review in the window. The previously ingested living Cochrane NMA was not replaced. The window-wide PubMed retraction/EoC query returned 0. Two retraction notices of unrelated papers (infliximab hydrogel PMID 42948410; zinc aspartate PMID 42925452) were seen and not used as evidence. PubMed Practice Guideline[pt] in the window: 0.

What survived triage: eight bibliography additions, two NCT IDs; no landmark note.

Record Disposition
Lee 2026, PMID 42269230 (bDMARD stop vs reduce; 11 studies, 2,861 people) treat-to-target, OQ-4, statistics. Search to 1 Feb 2026
Rotea-Salvo 2026, PMID 42963167 (15 RCTs; class-specific tapering) treat-to-target, JAK page, OQ-4, statistics. Search February 2024, so it does not include Park
Lee 2026, PMID 42034889 (JAK1 vs pan-JAK retention; 21,244 people) JAK page, OQ-6, statistics. June 2026 catch-up; observational
Ota 2026, PMID 41422438 (TNF vs JAK serious infection; n=5,018) JAK page, statistics. Japanese claims catch-up
Lampa 2026, PMID 41881637 (NORD-STAR 48-week PROs) bibliography only. April 2026 catch-up; between-group PRO differences generally small
Lao 2026, PMID 42490391 (NZ GP incidence 26.9/100,000; prevalence 417.5/100,000; SMR 1.07) STATISTICS. Issued 24 July 2026 catch-up
Nagafuchi 2026, PMID 41796486 (NORD-STAR+BeSt joint-pattern IPD) bibliography only
Apaydın 2026, PMID 42927519 (ultrasound-guided synovial biopsy n=87, 47% RA) bibliography only; mixed rheumatology
NCT07904754 (ABBV-2294 platform, n=250); NCT07895524 (MK-8690, n=60) trials landscape

Dropped (honest). Mechanistic animal, herbal, nanoparticle and in-silico papers; bee-venom pharmacopuncture SR; paraffin/cupping and TCM-basket registrations; multi-disease QLS2416 phase 1; not-yet-recruiting dapagliflozin adjunct phase 1/2; n=66 RA-ILD nomogram; gout-in-RA OBRI (null on CDAI); spatial multiomics n=6; Pickles CAT PROM methods paper; letters on tapering/remote DAS. Quarantined, not cited: PubMed records whose attached DOIs resolved on Crossref to unrelated older titles, including PMIDs 42919618, 40975268, 41578952, 41506743, 41929500, 42399075, 42342288, 42963440, 42919621, 42958501 and 42958313.

ClinicalTrials.gov. All 42 previously cited research-corpus NCT IDs resolved. No overall-status, enrollment-count or posted-results-availability changes. Last-update dates moved for NCT06733935 (2026-10-30) and NCT07201558 (2026-10-28) without status change. Six first postings since 25 October; two RA-powered programs ingested. Results first posted since 28 August remain ATI-450, an IBD study (excluded) and NCT04313725 (Stevens–Johnson/Sjögren/GVHD; excluded). NCT07782450 remains not yet recruiting with RA only in parts 1a/1b; NCT07363590 remains a 21-person SLE/RA phase 1; NCT05246293 remains UNKNOWN; NCT07873931 remains not yet recruiting.

ISRCTN. Tracked 3TR (ISRCTN44988547, ISRCTN94259329) last updated 2026-09-07; APIPPRA ISRCTN12680338 2026-01-23; BIO-FLARE ISRCTN16371380 2025-04-23; MAMA ISRCTN89126536 2026-10-21. No new posted results.

Guidelines / statistics / patient voice. ACR's official RA page still lists 2021 as its latest pharmacologic guideline. EULAR's management page still lists the 2025 update already in the registry. NICE NG100 last reviewed 19 November 2024 with no 2026 update. DGRh journal query returned two irrelevant hits; the 26 August 2026 S3 press release remains the live German document. No guideline-registry addition. NRAS news since the last sweep is membership-price, anniversary raffle, 5 October Healthcare Champions and the already-ingested NEIAA report; nothing required a patient-voice change. Arthritis Foundation and AFSA pages retrieved; no new RA-specific campaign.

Follow-ups from 2026-10-25, re-run. No PubMed-indexed journal record for DGRh S3, ATI-450, RESET-RA durability, APIPPRA/ARIAA further follow-up, TNFA-methylation external validation, Park radiographic/extension data, a periodontal-onset prevention trial, or Italian SIR RA-ILD item-level recommendations. COMPARE (NCT06475495) remains ACTIVE_NOT_RECRUITING. MAMA has no new results text.

Retraction and citation integrity. All eight additions: PubMed publication types show neither retraction nor expression of concern; Crossref updated-by empty; Crossref titles match PubMed. Numbers were checked against live abstracts (Lao SMR from the Europe PMC/Crossref abstract, because the PubMed abstract truncated). No full texts were opened for the new records.

Files and status. Edited three already-draft wiki pages: treat-to-target, JAK inhibitors, trials landscape (last-curated: 2026-11-01). Also updated bibliography (301→309), statistics (new 2026-11-01 table), OPEN-QUESTIONS (OQ-4 and OQ-6 sharpened; none resolved or newly opened), INDEX, this log and literature/sweeps/2026-11-01.json. Nine wiki pages remain curated and nine are draft. Independent audit remains pending; this author has not promoted its own work.

Next sweep should chase. ABBV-2294 / MK-8690 status changes; COMPARE phase 2 and NCT07873931; CARDINAL-RA and Ota full texts (infection vs MACE discordance); Rotea-Salvo/Lee tapering full-text subgroup tables and whether they include Park; Lao full-text ethnicity-specific rates; whether conventional-dmards.md should take the HTA monitoring-interval result at its next audit. Carry forward: DGRh journal report, ATI-450 publication, RESET-RA durability, periodontal-onset prevention (D-8), Corrales/Tariq/ANSWER full texts, Italian SIR RA-ILD item-level extraction, prospective LUS/DLCO validation, activated-naive-B-cell validation, MAMA results. The auditor must use a different engine from Grok.