Skip to content

Curation log — uterine adenosarcoma

Newest entries first. Every content-changing session appends an entry: date, what changed, what was searched, follow-ups for next time. See CLAUDE.md.

2026-09-01 — Independent audit (Codex auditor; Grok author)

Codex independently audited all 14 wiki pages and all 13 literature artifacts authored by Grok. The audit re-fetched 66/66 distinct PMIDs through live PubMed E-utilities and 12/12 distinct NCT records through the live ClinicalTrials.gov v2 API; every identifier resolved and matched the paper or trial now attributed to it. It checked 672 in-text identifier-linked claim–citation pairs outside wiki reference lists and the master bibliography (626 PMID-linked occurrences + 46 NCT-linked occurrences). Wiki body/reference sets, bibliography coverage, front matter, reference numbering, and local links were also reconciled. All 14 pages passed after repair and were promoted to status: curated.

Errors found and fixed. Each item below changed the authored material rather than merely recording a concern.

  1. Mitotic denominator: Zaloudek’s proposed threshold had been reported as 4 mitoses/10 HPF. The live record says 4/100 HPF. Corrected in pathology and the Clement landmark note, with an explicit warning not to conflate it with Clement’s separate 2/10-HPF criterion.
  2. Trial absence was stale and too broad: a full record-level scan of all 287 ClinicalTrials.gov uterine sarcoma results found the mixed-histology randomised phase 3 registration NCT00162721, which explicitly included adenosarcoma. It has UNKNOWN status, estimated n=270, no posted results, and no retrieved adenosarcoma-specific publication. The defensible gap is therefore no adenosarcoma-only randomised trial and no reported histology-specific randomised outcome—not “no trial has ever run.” The audit also added eligible records NCT03509207 (vorinostat; terminated, n=3) and NCT05481645 (immunochemotherapy ± anlotinib; terminated, n=71), updated NCT01979393 to its completed actual n=58 record, and preserved FUCHSia and elacestrant with current status.
  3. Registry false positives: NCT02020707 carries a “cervical adenosarcoma” condition tag but its eligibility permits epithelial cervical cancers and not adenosarcoma; NCT07394413 is a likely miscoded cervical-adenocarcinoma observational study. DART (NCT02834013) does not name adenosarcoma or uterine sarcoma in the retrieved eligibility text. All three are now labelled accurately rather than presented as disease trials.
  4. Molecular-fusion absence was obsolete: the live PubMed search found Agaimy 2026 (PMID 41555057), a 12-case series of adenosarcoma-like uterine neoplasms with recurrent ESR1::NCOA3/2 fusions. The molecular page now reports the finding while preserving the authors’ unresolved classification versus conventional adenosarcoma and UTROSCT; it is not presented as diagnostic-assay validation.
  5. Patient-voice absence needed qualification: Ziegler 2026 (PMID 42334466) contains a short perspective from one fertility-sparing patient. It is now included and explicitly graded as n=1 case-report evidence, while the dated gap is no adenosarcoma-specific qualitative cohort.
  6. Recurrence overstatement: “at least 50% recur” had been used as though it were a general cohort rate. It is now identified as a narrative-review synthesis and set beside primary, case-mix-dependent estimates: Clement 23/100 overall and Carroll stage I 77% with SO versus 22% without.
  7. Epidemiology comparison error: the text had causally attributed a higher SEER share to inclusion of carcinosarcoma. The datasets and denominators are now described as non-comparable without claiming that untested explanation. A synthetic “SO ~25–35%” working average was also removed; series estimates remain side by side with their selection frames.
  8. Publication metadata: Qu was dated 2021; the live citation is Frontiers in Oncology 2020 (PMID 33680946). Harada’s citation was updated to its final 2025 volume/pages. Reference numbering gaps created by deleted citations were repaired.
  9. Diagnostic-mechanism overstatement: preoperative biopsy/imaging failure had been described as frequent and as the established mechanism of morcellation. The sources provide an “occasional” qualitative statement plus five recurrent-polyp cases, not a failure rate. The page now states the inference and its unmeasured frequency explicitly.
  10. Molecular and pathology absences: blanket statements that no diagnostic molecular/IHC test exists, that BAP1 is “not tested,” and that PI3K-pathway findings had reached zero trials were converted to dated search results with precise scope: no prospectively validated diagnostic assay, no retrieved BAP1 diagnostic-accuracy validation, and no PI3K-pathway-selected adenosarcoma trial as of 2026-09-01.
  11. Other absence claims re-searched and dated: no prospective imaging diagnostic-accuracy study, no prospective surveillance-schedule comparison, no continuous-volume SO outcome model, no central-review population incidence estimate, no study of initially missed SO rates, no adenosarcoma-specific sexual-function score, and only one adenosarcoma-plus-morcellation PubMed hit (a case report, PMID 27769260). Bare [unverified] markers were removed.
  12. Guideline scope: GCIG’s literal “no trials” sentence was narrowed in light of NCT00162721. The accessible NCCN Insights paper (PMID 40763788) is endometrial-focused; both official parent-guideline PDF endpoints returned sign-in HTML on 2026-09-01. Claims about current NCCN adenosarcoma recommendations were removed rather than inferred. ESGO recommendations were rechecked against the live pocket-guideline PDF.
  13. Patient-resource drift: the old American Cancer Society “questions to ask” URL now redirects to a general detection/diagnosis/staging hub, so the obsolete question-list claim was removed. Sarcoma Foundation of America returned HTTP 403 to automated re-fetch; changing dollar, grant, trial, and service counts were removed. Other directory pages were re-fetched, and closed communities were not accessed.
  14. Evidence grading and reconciliation: case reports, literature-assembled cases, single-institution retrospective cohorts, registries without central review, and mixed-histology trials are now labelled as such at the point of use. The bibliography contains all 66 cited PMIDs and no uncited PMID; every wiki body PMID appears in that page’s reference list and vice versa.

What remains flagged. There is no unresolved substantive citation issue and no live [unverified] marker in the audited wiki or literature layer. Two access limitations remain as positively stated, dated gaps: the current NCCN parent algorithm could not be retrieved without sign-in, and SFA’s site blocked automated access. Trial-result gaps also remain factual: NCT00162721 has unknown status/no posted results, and the retrieved basket trials do not report an adenosarcoma-specific efficacy estimate. These limitations do not support any affirmative clinical claim and therefore did not prevent curation.

Status: all 14 pages curated; condition advanced to audited in CONDITIONS-ROADMAP.md.

2026-08-31 — Full build (Grok)

Built all 14 canonical wiki pages from live literature, plus the literature layer, a rewritten OPEN-QUESTIONS.md, and an updated INDEX.md. Every page left status: draft for independent audit by a different engine. CONDITIONS-ROADMAP.md status → built. Density-exempt rule observed: pages are as long as the sources support; case reports are labelled as case reports.

Seed-anchor re-verification (all six PMIDs resolved live). Two attributions in the seed INDEX were wrong and were corrected rather than repeated: - PMID 29441675 is Tate K et al., not Tanaka H. Title and numbers (110 assembled cases; 34% disease-specific death; stage IA 13% recurrence / 3.3% mortality) match the abstract; the first-author error is corrected throughout. - PMID 26646120 is Raine-Bennett T et al., not Multinu F. Occult sarcoma 1/278 (3.60/1,000, 95% CI 2.97–4.23) is this paper. The Multinu occult-sarcoma paper is PMID 30447212 (2019, Olmsted County) and was added as a separate source.

Literature size re-measured. uterine adenosarcoma esearch count 258 (unchanged). Sarcomatous-overgrowth query 142 (broader than the seed’s 46 because of Müllerian/overgrowth expansion). Molecular query 39. No population-incidence paper with central review was found.

Searches run (PubMed E-utilities, this session). Topic queries included: uterine adenosarcoma (count); SO; molecular/MDM2/CDK4/TERT; SEER/population/incidence; hysterectomy/lymphadenectomy/oophorectomy/fertility; chemotherapy/radiotherapy/hormonal/adjuvant; recurrence/surveillance; MRI/imaging; guideline/ESMO/NCCN/FIGO; morcellation; Clement Scully; author-targeted (Seagle, Tanner, Arend, Kaku, Zaloudek, Nathenson, Friedlander, McCluggage, Gallardo, Verschraegen, Machida, Nucci, Pinto, Hodgson, DICER1, BCOR, Prat FIGO, ESGO, tamoxifen, ifosfamide, trabectedin, pazopanib); Brooks SEER; qualitative/HRQoL. Abstracts fetched via efetch for all papers cited. No PMID was written from memory.

ClinicalTrials.gov v2 (this session). query.term=uterine adenosarcoma → 5 studies; query.term=adenosarcoma → 7; query.cond=uterine sarcoma → 287 (almost none adenosarcoma). Named-eligibility trials: NCT03926936 FUCHSia (completed, n=17, no PubMed results paper); NCT07467772 elacestrant (recruiting, estimated n=30); NCT01979393 EORTC 62113 cabozantinib (protocol PMID 32546554). DART NCT02834013 eligibility text does not contain “adenosarcoma”. Finding: no dedicated adenosarcoma randomised trial.

Pages written (all draft). overview, pathology-and-diagnosis, sarcomatous-overgrowth, molecular-and-genomic-features, epidemiology-and-burden, clinical-presentation-and-imaging, staging-and-prognostic-factors, surgical-management, adjuvant-and-systemic-therapy, recurrence-and-surveillance, guidelines, clinical-trials-landscape, red-flags-and-safety-concerns, patient-experience-and-advocacy.

Literature layer. BIBLIOGRAPHY.md; 6 landmark notes (Clement 1989 SO, Clement 1990, Carroll 2014, Howitt 2015, Nathenson 2018 LVSI, Seagle 2016 NCDB); guidelines/REGISTRY.md (10 documents); statistics/STATISTICS.md; patient-voice/ (README, organizations, themes, sources) with ethics rules observed. No adenosarcoma-specific qualitative study exists; den Hollander 2022 (PMID 35443673) is mixed uterine sarcoma n=13.

Patient organisations fetched. Foundation for Women’s Cancer, Sarcoma UK, Sarcoma Foundation of America, Sarcoma Alliance, NCI PDQ, ACS question-prompt page. Closed groups signposted only.

Could not verify / left explicit. - Full NCCN Uterine Neoplasms sarcoma algorithm (Insights 2025 PMID 40763788 is endometrial-focused); parent-guideline sarcoma wording marked [unverified] beyond that paper. - FUCHSia results (completed 2022, not PubMed-indexed this session). - EORTC 62113 primary results as an adenosarcoma subset analysis. - Population incidence of uterine adenosarcoma per 100,000 with central review — not in the retrieved set. - Adenosarcoma-specific morcellation outcomes — one case report (PMID 27769260). - FIGO 2009 stage table body: Prat PMID 19135669 is a short communication without a usable abstract; IA/IB/IC invasion substaging was confirmed against NCI PDQ (fetched) and secondary tables, and is cited as Prat 2009 plus NCI PDQ access date.

Follow-ups for the auditor / next sweep. 1. Re-fetch every cited PMID; check Carroll HRs, Raine-Bennett 1/278, Tate vs Tanaka, Raine-Bennett vs Multinu. 2. Re-query ClinicalTrials.gov for FUCHSia results and NCT07467772 accrual. 3. If an adenosarcoma-specific qualitative study appears, it should replace the mixed-sarcoma default in patient-voice/themes.md. 4. Do not inflate citation counts to clear a density threshold.

Next: independent audit by a different engine; then ROADMAP → audited if the audit passes.

2026-08-31 — Seeded

  • Created the condition scaffold (INDEX.md, OPEN-QUESTIONS.md, LOG.md, wiki/, literature/) per CLAUDE.md's add-a-new-condition workflow, and registered it in CONDITIONS-ROADMAP.md under Cancer.
  • Ran live PubMed scoping searches and recorded six resolved anchor records. Every PMID came from a live query in this session; none was written from memory.
  • Measured the size of the literature before designing the page list, which turned out to be the decision that shaped everything else: 258 total records for the condition, 46 on sarcomatous overgrowth, 20 on molecular features, 3 population-based. Most of the corpus is case reports and single-institution retrospective series; the largest cohort found while scoping was 74 patients with multivariable analysis, and the largest dataset a 797-patient SEER extract without central pathology review.

Two scoping decisions follow from that.

  1. Fourteen pages, not the usual 19–20. Mechanism, prevention and risk-factor pages that earn their place in a common condition would be empty here. Sarcomatous overgrowth gets a page of its own because it is the one consistently supported prognostic variable and the single most decision-relevant fact about the disease.
  2. Density-exempt. .density-exempt records why: fourteen pages at 24 citations each would require 336 citation-slots from a 258-record literature, reachable only by citing case reports as evidence. The build and any future deepen pass must not inflate this condition to clear a threshold designed for common diseases.

A framing instruction for the build. In most conditions here, "open question" means the evidence leaves a decision unresolved. In this one it usually means no adequately powered study exists, because the disease is too rare for one. Those are different claims and must not be written as though they were the same; a question answerable only by a rare-tumour consortium or registry should say so rather than proposing a trial nobody can run.

Next: full build from live literature, then an independent audit run by a different model than the one that wrote the pages.