Skip to content

Curation log — thoracic aortic aneurysm

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


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

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

Searches run. PubMed esummary + efetch for all 44 unique PMIDs cited on the page (existence, journal, year, title vs. the reference list); efetch abstracts for all of them; full text of the 2022 ACC/AHA aortic guideline via PMC (PMC9876736 ← PMID 36322642) to settle the two [unverified] flags.

Result: 44/44 records exist and match their reference-list entries; 0 fabricated or wrong-paper citations. Every numeric claim on the page was checked token-by-token against the source abstract, then the high-leverage ones (Hagan sign frequencies, Lovatt misdiagnosis rate, Rogers ADD-RS sensitivity, Nazerian failure/efficiency, Tsai false-lumen mortality, Chaddha surveillance HR, Mariscalco relative yield, Kamel IRR, Hibino meta-analysis, Smith diameter failures, Hatzaras case series) re-read in context for direction as well as value. All correct.

Changes made. - Both [unverified] flags resolved against the guideline's own text, not a secondary source. (1) Symptomatic aneurysm: §"Aneurysms of the Aortic Root and Ascending Aorta" recommendation 1 is "In patients with aneurysms of the aortic root and ascending aorta who have symptoms attributable to the aneurysm, surgery is indicated" (COR 1, LOE C-LD), with the ≥5.5 cm / ≥5.0 cm / growth-rate thresholds all scoped to asymptomatic patients — the page now states this with grades instead of flagging it. (2) Pre-pregnancy thresholds: §8.3 transcribed into a condition-by-condition table (Marfan >4.5 cm COR 1 C-LD, 4.0–4.5 cm COR 2b; LDS by gene; nsHTAD ≥4.5 cm; Turner ASI ≥2.5 cm/m²; BAV ≥5.0 cm; sporadic ≥5.0 cm), with the observation that every threshold is C-LD or C-EO. - Exercise section strengthened with verified guideline text: the "no consensus on intermediate intensities" claim is now quoted exactly ("No uniform consensus exists about the safety of intermediate-level static and aerobic exercise"), paired with the two graded recommendations (avoidance COR 1 C-EO; 30–60 min mild-to-moderate aerobic 3–4×/week COR 2a C-LD) and the >300 mm Hg Valsalva figure the guideline gives as rationale. - Over-precision corrected (Anagnostopoulos, PMID 34140225): the 86% thoracic-graft sensitivity holds only under the analysis that assigns MAGIC-"suspected" cases to not diseased; the page stated it unconditionally. Now carries the condition, the CI (73–95), and the cohort denominators. - Greve cocaine series age SD corrected 8.7 → 8.67 (as published). - Reference 2 (PMID 36334952) was an orphan — never cited in the body. Annotated as the simultaneous JACC co-publication of reference 1, with a note that guideline text is cited to the Circulation version. - Frontmatter: status: draftcurated, last-curated: 2026-08-31.

Follow-ups. The §8.3 table duplicates material that belongs on guidelines — check that page carries the same thresholds with the same grades, or cross-link rather than restate. PMC9876736 is the full guideline text and is now known to be retrievable; other TAA pages citing PMID 36322642 for specific COR/LOE claims could be re-verified against it cheaply.


2026-08-28 (later) — Red-flags page; trials stage/safety enrichment

  • Added wiki/red-flags-and-safety-concerns.md (draft, 44 verified PMIDs): dissection symptom frequencies, misdiagnosis quantification (33.8%), screening triggers, graded cautions (hypertension RR 3.07 dose-dependent; pregnancy IRR 4.0; weightlifting case-series-only; cocaine; fluoroquinolone cross-link), surveillance and post-repair warning signs. Needs its own audit pass; two [unverified] flags (symptomatic-aneurysm COR text; pre-pregnancy numeric thresholds) deferred to a guidelines full-text pass.
  • Enriched wiki/clinical-trials-landscape.md (stays curated): trial-stage primer; per-trial phases; published efficacy + safety readouts (PERSEVERE 30-day/1-year, Thoraflex IDE + FDA approval, PHN atenolol/losartan AE profile, Marfan Sartan, N-TA³CT null, INSTEAD-XL 5-year added as SUNDAY precedent); 16 ongoing trials re-verified (no status changes; ARISE III acronym landed; several completion dates extended).
  • BIBLIOGRAPHY.md grew to 344 unique PMIDs (red-flags + trials-enrichment merges).

2026-08-28 — Full citation audit; ask capability; weekly sweep armed

Audit pass (5 parallel auditors, all 19 pages): every inline citation re-verified against live PubMed metadata/abstracts (full texts via PMC/publisher where claims demanded it); all 49 NCT registrations re-verified against ClinicalTrials.gov (no status changes since 2026-08-27 — first trial sweep on record). Results: 0 fabricated PMIDs, 0 wrong-paper citations across ~600 claim-citation pairs; ~15 substantive fixes applied, the biggest: Paruchuri relative-risk conflation (346.8 vs-distribution ≠ vs-<3.5 cm, which is ~6,305×; fixed on 2 pages + statistics sheet), 2 wrong LOE cells + 1 missed LOE-A item in the 2022 ACC/AHA tables (58 COR/LOE rows re-verified against the guideline full text), Lacro 2014 blinding mischaracterized, FDG-aortitis claim cited to an unrelated retroperitoneal-fibrosis study (replaced), Quint reproducibility mis-attributed inter- vs intra-observer, Yacoub 4.6% denominator. All 19 pages promoted to curated.

[unverified] flags 15 → 2: resolved with real sources — Renard 2018 per-gene ClinGen tiers (from PMC full text; gene table + paper note updated), 2024 EACTS/STS numeric thresholds (full EJCTS PDF retrieved; new verified section in guidelines.md incl. the 52.5-mm hinge revision and gene-specific tables), acute/chronic dissection boundary (2022 guideline verbatim + SVS/STS 2020), large-animal models (3 porcine + 1 ovine primary papers; key negative: no root/ascending large-animal model exists), sST2 (a documented replication failure: AUC 0.97 → 0.63 → uninformative), desmosine, lumican, MR blood-pressure and lipids (remnant cholesterol causal for AAA, null for TAA — sharp negative test of the atherosclerotic hypothesis). Converted to explicit evidence-gap statements: FDG threshold, prospective proteomics, gene-editing rescue. Remaining 2 notes on guidelines.md document an extraction limitation (ACC/AHA per-gene table: PMC strips gene symbols; publisher sites 403/CAPTCHA — all routes recorded in-page).

Missing SMAD3 discovery paper found (van de Laar 2011, Nat Genet, PMID 21217753) and added to the gene table.

Bibliography: +27 audit-addition entries (all re-verified this session), 1 orphaned entry removed (Thuermel), 2 print-year fixes → 312 unique PMIDs. Statistics sheet: 36-source spot-check, 1 error fixed (Bickerstaff "presented with" vs "occurred"). Guidelines registry: 24/24 documents verified clean.

System changes: /ask project skill added (.claude/skills/ask/); OPEN-QUESTIONS.md gained OQ-35–40 (patient-centered) and the "Dots not yet connected" table (12 cross-domain junctions); frontend gained the Ask panel (client-side passage retrieval + comment-to-Claude path for synthesized answers) and an Unconnected-dots nav entry; weekly literature sweep scheduled (Mondays ~09:00 local, task taa-weekly-literature-sweep — includes artifact-comment answering and auto-republish).

Follow-ups: ACC/AHA per-gene thresholds still need a full-text route (print copy or authenticated access); ePub-vs-print years could be normalized bibliography-wide in one pass (list in this entry's audit reports); statistics §4 surgical-outcome block not yet sampled (Krüger, Rylski, Farag, Benedetto, Dobaria, Catalano, Cefarelli, Harky, Bottle, Brown, Grant); Davies 2006 vs Zafar ASI band cross-check; duplicate-PMID topic lines (by design, but confirm cited-by lists stay consistent); blocked web sources from 08-27 list still pending.


2026-08-27 — Initial build

  • Created repository scaffold (README, CONVENTIONS, CLAUDE.md protocol, templates) and the TAA condition tree.
  • Built all 18 canonical wiki pages via five domain research passes (foundations/natural history, genetics, mechanism, clinical management, research frontier), each grounded in live PubMed searches; trials page grounded in live ClinicalTrials.gov queries.
  • Created literature layer: per-domain bibliography fragments, master BIBLIOGRAPHY.md, landmark paper notes.
  • Synthesized OPEN-QUESTIONS.md from cross-domain gaps surfaced during the build.
  • All pages start at status: draft — no page is curated until it passes a verification audit (see CLAUDE.md "Audit" workflow).

Frontend: published as a private artifact at https://claude.ai/code/artifact/571412b5-22f6-436b-b6cc-03f7e51b45a7 (build: python3 tools/build_frontend.pyfrontend/taa-wiki.html; republish to the same URL after every sweep).

Build inventory: 19 wiki pages (all draft); BIBLIOGRAPHY.md 286 unique PMIDs; 16 landmark notes; guidelines registry 24 documents (5 superseded chains); statistics sheet ~116 figures + 9 documented conflicts; patient-voice layer (17 orgs verified by direct fetch, 10 themes, 4 files); OPEN-QUESTIONS.md 40 questions (8 Tier-1). Frontend build tooling in tools/.

Follow-ups for next session(s): - Audit pass: spot-check citations on 2–3 pages (claims vs abstracts), promote clean pages to curated. Overview and guidelines pages already passed an informal read-through. - [unverified] items to resolve (15 flags across 6 pages): per-gene ClinGen tier assignments (needs Renard 2018 full-text table); 2024 EACTS/STS numeric thresholds (no retrievable full text yet); gene-level HTAD thresholds in 2022 ACC/AHA Table (PMC rendering dropped gene symbols); 14-day acute/chronic dissection boundary source; desmosine/sST2/lumican primary TAA cohorts (may not exist — if confirmed absent, convert to open-question evidence); FDG threshold for degenerative TAA; porcine/ovine model protocols. - Full-text pulls needed: Hirst 1958 (PMID 13577293) and Anagnostopoulos 1972 (PMID 4557973) internals behind the "1–2%/hour" claim (OQ-31). - Blocked sources to retry: vascern.eu (403), healthdata.org GBD factsheet (403), ctdna.org.au (refused), Marfan Network Japan; patient-org coverage gaps: Spain, Italy, Nordics, Latin America. - Literature sweep with date_from: 2026/08; trial status re-check for TITAN:SvS (NCT03536312), ARISE II (NCT05800743), and the 50 catalogued NCTs; guideline watch list in literature/guidelines/REGISTRY.md (ESVS 2026 just published — watch for ACC/AHA focused updates). - Missing SMAD3 discovery paper (van de Laar 2011 Nat Genet) — did not surface in searches; find and add. - After any sweep: rebuild + republish the artifact frontend (CLAUDE.md workflow step 6).