Skip to content

Curation log — ischemic heart disease

2026-08-28 — Seed

  • Created: INDEX.md (canonical 16-row Pages table), wiki/overview.md (draft, 28 verified PMIDs), OPEN-QUESTIONS.md (7 seed questions), this log.
  • Searches run (live PubMed, esearch/esummary/efetch + MCP PubMed tools): ISCHEMIA and ISCHEMIA-EXTEND; COURAGE; ORBITA, ORBITA-2, ORBITA-COSMIC, ORBITA-program predictor analyses; REVIVED-BCIS2; 2023 AHA/ACC chronic coronary disease guideline; 2024 ESC chronic coronary syndromes guideline; GBD IHD burden 1990–2021; GBD CVD 2023 (JACC); GBD elevated LDL-C 2023; IHD quality-of-care GBD analysis; CTT 2010 intensive statin meta-analysis; IMPROVE-IT; FOURIER; REDUCE-IT; CANTOS; COLCOT; LoDoCo2; ATT aspirin meta-analysis; COMPASS; SECURE polypill; cardiac rehabilitation meta-analyses (2021 Cochrane, 2023 EHJ); SELECT.
  • Verification: all cited PMIDs retrieved from live PubMed this session; abstracts fetched for all quantitative claims (event rates, hazard ratios, burden figures).
  • Next: full build queued per CONDITIONS-ROADMAP.md — 15 remaining wiki pages, literature layer, ClinicalTrials.gov verification (anti-inflammatory and sham-controlled device pipelines), patient-voice layer.

2026-08-28 — Full build

  • Built all 16 canonical wiki pages: extended overview.md; added epidemiology/global burden, plaque biology, stable CAD, ACS, PCI/CABG, lipid lowering, antithrombotics, inflammation, rehabilitation/lifestyle, INOCA/special phenotypes, guidelines, biomarkers, trials, patient experience, and red-flags/safety pages. All remain draft pending a separate-session audit.
  • Built the literature layer: BIBLIOGRAPHY.md generated from the page citation graph; 6 landmark notes (COURAGE, ISCHEMIA, ORBITA-2, REVIVED, CTT, CANTOS); 19-document guideline/implementation registry with supersession chains, disputes, regional coverage and watch list; quantitative statistics quick reference; and patient-voice method, organizations, themes and annotated sources.
  • Rewrote OPEN-QUESTIONS.md: 15 Tier 1 and 12 Tier 2 questions with stable OQ identifiers, plus 12 “Dots not yet connected” junctions.
  • Live PubMed work: 40 broad domain searches (burden, plaque, stable/acute syndromes, PCI/CABG, physiology, lipids, antithrombotics, inflammation, rehabilitation, INOCA, women, biomarkers, safety and patient outcomes); targeted landmark-title/acronym searches; direct E-utilities resolution/metadata retrieval for all cited identifiers; and a final citation-graph re-resolution. The completed wiki uses 129 unique PubMed records across more than 740 citation occurrences.
  • Live ClinicalTrials.gov work: API v2 searches for active coronary-disease studies, IHD interventions, Lp(a), ziltivekimab, colchicine, sham/device trials, PREVENT, CorMicA and INOCA; 243 unique registry studies retrieved in broad sweeps plus targeted queries. Page tables include status/enrollment as of this date.
  • Official-web verification: NICE, WHO/HEARTS, JCS, CCS, Australia/CSANZ and South African guideline/implementation sources; Global Heart Hub, Mended Hearts, WomenHeart, INOCA International, BHF, Heart & Stroke Canada, Heart Foundation Australia, and Heart & Stroke Foundation South Africa.
  • Integration: updated INDEX.md with portrait, counts, reading paths, page/literature statuses; set CONDITIONS-ROADMAP.md status to built.
  • Could not verify / limits: no comprehensive current African cardiology-society IHD guideline beyond the South African national STG/EML was identified; Latin American and several Asian patient-organization sources remain underrepresented; registry statuses can lag site activity; several wiki pages finish at 144–149 lines, within the approximate 150-line target, because content was kept claim-dense and non-repetitive. No unresolved verification flags remain; unsupported claims were omitted.
  • Next: run the required separate-session audit: re-fetch every PMID/NCT, check claim/number match, repair discrepancies, and promote only fully passing pages to curated/roadmap audited.

2026-08-30 — Page-by-page evidence deepening

  • Deepened all 16 canonical wiki pages in place: retained the existing text and added comparative evidence tables, effect sizes and confidence intervals, mechanism-to-outcome links, population boundaries, competing guideline interpretations, and explicit controversies. Pages now contain 25–34 unique PubMed papers each and remain draft pending the required independent audit.
  • Major additions by domain: DANAMI-2/STREAM-2/CULPRIT-SHOCK/RIDDLE-NSTEMI boundaries in ACS; CAPRIE/CURE/TRITON/PLATO/PEGASUS, AF-PCI dual-versus-triple therapy, STOPDAPT-2 and SMART-CHOICE in antithrombotics; MESA CAC and rapid hs-cTn validation in biomarkers; RAMIT, telerehabilitation, participation inequity and smoking cessation in rehabilitation; INTERHEART/PURE and numerator-denominator caveats in epidemiology; 4S/HPS/TNT and absolute-risk framing in lipids; causal triangulation for IL-1β/IL-6/CHIP; endotype-strength grading for ANOCA/INOCA/MINOCA; route-specific CABG/PCI and shock-specific completeness; and patient-important outcomes separated from clinical composites.
  • Live PubMed searches: ran 128 page-specific ESearch queries (eight per page) spanning acute reperfusion/timing, antithrombotic sequencing, hs-troponin/CAC/CCTA/PRS, rehabilitation/lifestyle, burden and care gaps, guidelines, inflammation, INOCA/MINOCA, lipid trials, plaque biology, patient outcomes, diagnostic safety, revascularization and stable-CAD symptom/prognosis trials. Ran additional targeted searches for CURE, PEGASUS, WOEST, PIONEER AF-PCI, RE-DUAL PCI, AUGUSTUS, ENTRUST-AF PCI, STOPDAPT-2 and SMART-CHOICE. Relevance-mismatched top hits were rejected rather than cited.
  • Identifier verification: live EFetch resolved all 130 pre-existing bibliography PMIDs, then all 183 PMIDs cited by the deepened wiki; page reference sections were mechanically rebuilt from those live records. ClinicalTrials.gov API v2 individually resolved all 21 NCT identifiers retained anywhere in the condition; no unresolved PMID or NCT remained.
  • Literature layer: added 54 newly cited papers to literature/BIBLIOGRAPHY.md, bringing it to 184 deduplicated records, grouped the additions by topic, and refreshed tags/cited by fields against all 31 other condition Markdown files. Expanded statistics/STATISTICS.md with ACS/revascularization, lipid, symptom and antithrombotic estimates; expanded guidelines/REGISTRY.md with an evidence-shock maintenance table and visible implementation assumptions.
  • Open questions: preserved OQ-1 through OQ-27, sharpened DAPT, global implementation, sex-sensitive diagnosis and psychological-care questions with explicit estimands, added a decision-ready trial-design table, and expanded “Dots not yet connected” from 12 to 15 junctions.
  • Could not verify / limits: several broad ESearch queries returned irrelevant relevance-ranked records; these were discarded. No claim from those records was retained. Official regional guideline websites were not comprehensively re-searched in this session, so the registry's negative regional findings remain bounded to the documented 2026-08-28 searches. This was a build/deepening session, not the separately required independent claim-to-source audit; statuses therefore remain draft.
  • Next: in a fresh session, audit every claim→identifier pair against the full abstract/article and registry history, with particular attention to composite definitions, subgroup transportability, guideline class/level wording, and ePublication-versus-print metadata; promote pages only after that pass.

2026-08-30 — Independent audit

  • Scope completed: audited all 16 canonical wiki pages and every condition literature artifact: the 184-record bibliography, 6 landmark notes, guideline registry, statistics ledger, all 4 patient-voice files, OPEN-QUESTIONS.md, and INDEX.md.
  • Claims and identifiers checked: reviewed 879 claim→PMID occurrences in wiki bodies and the literature-layer uses; live PubMed E-utilities EFetch returned all 184 unique cited PMIDs. Live ClinicalTrials.gov API v2 calls returned all 22 unique NCT records across 86 occurrences. Page-body and reference-section graphs match exactly; the condition-wide bibliography contains all 184 non-bibliography PMIDs with no missing or orphaned record.
  • Record/metadata repairs: corrected 53 incorrect inline author/year attribution occurrences spanning 31 PubMed records. Replaced an indirect Russian DAPT summary with the live-verified primary DAPT Study report (Mauri 2014, PMID 25399658). Final live metadata checks found no unresolved author, year, journal, title, or identifier mismatch; ePublication/print-year differences were tolerated where applicable.
  • Substantive repairs: reversed the incorrect claim that long-term iFR and FFR outcomes were similar; the cited 5-year analysis instead reported higher MACE with iFR (HR 1.18, 95% CI 1.04–1.34) and higher all-cause mortality (HR 1.34, 95% CI 1.08–1.67). Corrected cardiac-rehabilitation hospitalization NNT from 27 to 37. Corrected two COURAGE descriptions so 19.0% versus 18.5% is identified as the primary death/nonfatal-MI outcome. Qualified stable-PCI, ischemic-LV-dysfunction, very-low-LDL safety, plaque-prediction, biomarker-actionability, and implementation claims where the cited evidence did not support the stronger wording.
  • Registry/status repairs: updated Lp(a)HORIZON to completed on 2026-07-16 with 8,323 actual enrollment, OCEAN(a)-Outcomes to active/not recruiting with 7,297 actual enrollment, and current enrollment/status fields throughout. Added RIGHT (NCT06025071) after a live strategy search and described it positively as a registered hsCRP-enriched randomized strategy whose record is stale (UNKNOWN; last known not yet recruiting; no posted results). Recorded the ORBITA-COSMIC discordance: ClinicalTrials.gov returns UNKNOWN and 50 estimated, while the paper reports 61 enrolled, 51 randomized, and 50 analyzed.
  • Negative-evidence searches rerun: repeated live PubMed and ClinicalTrials.gov searches for hsCRP-targeted/titrated anti-inflammatory strategies, selective Lp(a) outcome publications, inclisiran outcomes, and integrated plaque-guided strategies. Dated evidence-gap sentences now replace timeless absence claims. Lp(a)HORIZON searches found the design paper, reviews, and health-technology assessments but no primary outcome report as of 2026-08-30; the hsCRP strategy search found RIGHT but no completed target-titration outcome trial.
  • Literature/web verification: rechecked official guideline and patient-organization URLs on 2026-08-30. All direct requests resolved except the CCS library request, which timed out; its relevant entries were independently confirmed in a live indexed result on the official CCS domain, and that limitation is explicit in the registry. Negative regional findings are dated rather than presented as permanent absences.
  • Final validation: all 16 pages are 165–189 lines and contain 25–34 unique live-verified PubMed records; all local Markdown links resolve; every page has matched body/reference PMID sets; no unresolved verification or placeholder marker remains. All 16 pages passed and were promoted to status: curated; INDEX.md and CONDITIONS-ROADMAP.md now report audited.
  • Remaining flags: no unresolved substantive citation, numeric, metadata, registry, or page-structure issue. Dated research gaps remain intentionally visible; registry lag and the CCS direct-fetch timeout remain explicit source limitations rather than unresolved claims.

2026-09-02 — Topic-specific evidence-base expansion

  • Outcome: extended all 16 canonical wiki pages in place and preserved the prior verified text. Added 65 PubMed records absent from the previous condition-wide pool, raising wiki distinct records from 183 to 248 (15.5 per page condition-wide), per-page unique PMID counts from 25–34 to 29–38, and total wiki length from 2,878 to 3,073 lines. Every touched page is now status: draft pending independent re-audit.
  • What was deepened: quantified neutral or harmful routine interventions in ACS (oxygen, ECLS, frailty-limited invasive care); antiplatelet agent/duration trade-offs; diagnostic-yield versus treatment-delivery failures for hs-troponin; CAC/plaque burden; rehabilitation access, smoking, diet and exercise dose; trial endpoint/representation/core-outcome problems; modeled burden, socioeconomic access, air pollution and young-adult trends; exact left-main Class/Level grades and DAPT/CR guideline lineage; CHIP and residual inflammatory-versus-cholesterol risk; MINOCA/OCT and endotype-limited therapy; first versus total lipid events and surrogate inclisiran evidence; lifetime/cumulative LDL exposure; retention, efferocytosis, smooth-muscle lineage and trained immunity; clinician–patient angina discordance, return to work and sexual recovery; missed ACS, bleeding, SCAD and type 2 MI safety; and anatomy/time-horizon/symptom-specific revascularization evidence.
  • Live identifier work: ran 128 page-specific PubMed ESearch queries (eight per page) and fetched full EFetch records for the selected additions. NCBI rate limiting required conservative retries; no record was accepted without a completed live retrieval. One additional targeted SCAD ESearch/EFetch verified the AHA scientific statement (PMID 29472380). A final batched EFetch resolved all 249 PMIDs retained anywhere in the condition with zero missing records. All 22 retained NCT identifiers were individually re-resolved through ClinicalTrials.gov API v2 with no ID mismatch. PubMed search terms and returned IDs are retained in the session workspace under /tmp/ihd_pubmed/search-log.json.
  • Literature layer: added all 65 papers to literature/BIBLIOGRAPHY.md; added 14 quantitative rows to literature/statistics/STATISTICS.md; expanded literature/guidelines/REGISTRY.md with the ESC/EACTS focused left-main reassessment and exact Class I/A versus IIa/A grades, plus DAPT-lineage and rehabilitation-implementation disagreements.
  • Open questions: added OQ-28 through OQ-32 on shock-device selection, diagnosis-to-treatment equity after sex-specific troponin, recurrent-event estimands, SCAD reperfusion, and cumulative LDL exposure; expanded “Dots not yet connected” from 15 to 20.
  • Deliberately left alone: no existing verified claim was deleted or rewritten wholesale; the six landmark paper notes and patient-voice source/organization files were not expanded because this pass added no new landmark requiring a deep note and no newly verified public patient-organization source. No NCT identifier or registry-status claim was added or changed. CONVENTIONS.md, tools/, other conditions, and publication files were untouched.
  • Next: an independent model should re-fetch the 65 additions, audit every new number and author/year/title mapping, normalize bibliography journal abbreviations if desired, and only then restore curated status.

2026-09-02 — Independent audit (auditor: claude; author of the audited text: codex)

  • Scope. Audited all 16 canonical wiki pages and every condition artifact: BIBLIOGRAPHY.md, the 6 landmark notes, guidelines/REGISTRY.md, statistics/STATISTICS.md, all four patient-voice/ files, OPEN-QUESTIONS.md and INDEX.md. WHATS-NEW.md is generated by tools/build_recency.py and was left for regeneration at publish.
  • Identifiers re-fetched live. All 249 pre-existing PMIDs were re-resolved through PubMed E-utilities ESummary and full abstracts were fetched by EFetch for claim checking; after the audit's three additions all 252 PMIDs resolve with no errors. All 22 NCT identifiers were individually re-resolved through ClinicalTrials.gov API v2 and every status/enrollment figure written in the pages matched the live record.
  • Claim-citation pairs checked. 1,026 inline (Author Year, PMID) attributions were machine-checked against live author/year metadata (all 62 flags were group-author or particle-name artifacts — CTT, ATT, GBD collaborators, "van Oosterhout", "Van de Werf" — plus tolerated ePub/print year offsets, and none was an error). Every quantitative claim on every page (event rates, hazard/risk/odds ratios, confidence intervals, sample sizes, NNTs, prevalence and burden figures) was read against the fetched abstract.

Errors found and fixed

  1. PREVENT was described only through its design paper and its result was mis-attributed. pathophysiology-and-plaque-biology.md said "the trial's reported diabetic subgroup supports possible benefit"; the diabetes analysis (Kim 2025, PMID 40439104) in fact reports the same direction in both strata with no significant interaction. Retrieved and added the primary report (Park 2024, PMID 38604213) and replaced the passage with the actual 2-year result (0.4% vs 3.4%, absolute difference −3.0 points, 95% CI −4.4 to −1.8; death 0.5% vs 1.3%; MI 1.1% vs 1.7%). Corrected the same design-paper-only framing in clinical-trials-landscape.md (two places) and OPEN-QUESTIONS.md (OQ-9).
  2. COLOCT was described backwards. The plaque page said COLOCT and the LoDoCo2 CT subanalysis "illustrate the gap between event reduction and measurable local plaque change". COLOCT was positive (minimal fibrous-cap thickness +34.2 µm, 95% CI 9.7–58.6, p=0.006). Replaced with both trials' actual findings and the accurate conclusion that local change is measurable but modality- and population-dependent.
  3. ERICA numbers were framed as a within-group change. stable-cad-management.md read "reduced weekly angina episodes from 3.31 to 2.88"; 2.88 vs 3.31 is ranolazine vs placebo during treatment (baseline was 5.63/week). Rewritten with the correct comparison, baseline, n and p values.
  4. MULTISTARS AMI's result was omitted. The ACS page said only that the trial showed "immediate and staged strategies could be compared". Replaced with the finding (8.5% vs 16.3%; RR 0.52, 95% CI 0.38–0.72; noninferior and superior) and the shock exclusion.
  5. A meta-analysis was mischaracterized. The ACS page cited Manfrini 2016 (PMID 27273697) as supporting early invasive care "in appropriately selected non-ST-elevation disease"; the paper found benefit across UA/NSTEMI (RR 0.79) but not within NSTEMI (RR 1.19) and an early hazard to discharge. Rewritten with both estimates.
  6. EXCEL's five-year mortality difference was glossed. revascularization-pci-and-cabg.md described "broadly similar mortality"; all-cause death was 13.0% vs 9.9% (difference 3.1 points, 95% CI 0.2–6.1). Replaced with the full result and added the same detail where the page mentions "the five-year controversy" without naming it.
  7. Four cited PMIDs were missing from a reference list. revascularization-pci-and-cabg.md cited Serruys 2009 (19228612), Xaplanteris 2018 (29785878), Akintoye 2021 (32921597) and Spertus 2020 (32227753) in the body but listed none of them; all four added with live-verified metadata.
  8. A malformed reference block. stable-cad-management.md carried four unnumbered bullet references above its numbered list; merged and renumbered (29 entries).
  9. A residual NNT error the previous pass claimed to have fixed. cardiac-rehabilitation-and-lifestyle.md still said "approximately 27 treated to prevent one cardiovascular hospitalization"; Dibben 2023 reports NNT 37 for hospitalization. Corrected and the other two pooled estimates added with their confidence intervals.
  10. A registry trial was named wrongly. NCT05192941 is V-DIFFERENCE (placebo-controlled phase 4 inclisiran on top of optimized lipid-lowering therapy in hypercholesterolaemia), not "VICTORION-INITIATE", and it is not an "upfront implementation/LDL strategy" trial. Corrected in clinical-trials-landscape.md.
  11. Two wrong first authors in the patient-voice source list. PMID 37033594 is Borah N (listed as "Sharma R") and PMID 36447215 is Xie X (listed as "Zhang Y"). Both corrected; the wiki pages already cited them correctly.
  12. A citation's journal did not match its PMID. guidelines/REGISTRY.md cited PMID 34895950 as Circulation 2022; that identifier indexes the J Am Coll Cardiol co-publication (2022;79:e21–e129). Corrected, with the co-publication noted; the chest-pain guideline citation was completed with volume and pages.
  13. Three claims stated more than their source. Rehman 2020 (PMID 33189191) does not document "large gaps between indicated and received therapy" — it reviews Indian registries and quality-improvement programmes; Sun 2017 (PMID 27676464) reports age, sex, education, insurance and employment gradients but not race in its abstract; Rashid 2022 (PMID 36102261) shows no practice change in one health system rather than "health-system dependent" change. All three rewritten to what the sources report, with the actual figures.
  14. Two safety caveats were softened out of their sources. HiSTORIC (PMID 33752439) did not formally meet noninferiority for its 30-day safety outcome (upper 1-sided 95% CI 0.70% against a 0.50% margin; P=0.068), and Chiang 2020 (PMID 32245882) concluded the ESC 0/1-hour algorithm "may not be sufficiently safe if the 1% miss-rate is desired" with 8.1% one-year mortality in the observation zone. Both now stated.
  15. An omitted harm signal. The DAPT Study's higher all-cause death with extended thienopyridine (2.0% vs 1.5%; HR 1.36, 95% CI 1.00–1.85; P=0.05) was missing from the extension paragraph on a page whose own rule is to display benefit and harm separately. Added.

Asserted absences re-run

  • hsCRP-targeted anti-inflammatory strategy. Repeated live PubMed and ClinicalTrials.gov searches on 2026-09-02: still no completed strategy trial titrating colchicine or another agent to an hsCRP target; RIGHT (NCT06025071) remains UNKNOWN, last updated 2023-09-06, no posted results. Absence confirmed and re-dated in guidelines.md, biomarkers.md, inflammation-and-residual-risk.md and OQ-5.
  • Lp(a)HORIZON primary outcome. Re-run pelacarsen/olpasiran/Lp(a)HORIZON searches on 2026-09-02 found only reviews and a trial unit-concordance analysis (Cho 2026, PMID 42250319, added); no primary outcome report. Absence confirmed, re-dated and given its positive supporting citation.
  • ZEUS. The trial's registry record now reads COMPLETED (2026-06-09, 6,385 actual). A live PubMed search found no primary outcome report — the most recent ZEUS output is its rationale/design/baseline paper (Ridker 2026, PMID 41369941, added). A dated evidence-gap sentence now states this positively rather than leaving the reader to infer it.
  • Inclisiran outcomes. ORION-4/HPS-4/TIMI 65 remains a design publication (searched 2026-09-02); the pages' "awaits definitive outcomes evidence" framing stands unchanged.
  • All other verification dates carried forward from 2026-08-30 in the wiki were re-run in this session and re-dated to 2026-09-02; official-web entries in the registry and patient-voice layer were not re-fetched and keep their documented 2026-08-30 access dates.

Checked and confirmed correct

Every trial figure on the condition's anchor evidence matched its source, including COURAGE, ISCHEMIA and ISCHEMIA-EXTEND, ORBITA and ORBITA-2, ORBITA-COSMIC, REVIVED and its viability analysis, STICHES and the STICH age analysis, SYNTAX at 1 and 10 years, EXCEL, FAME 2, COMPLETE, DANAMI-3–PRIMULTI, CULPRIT-SHOCK, RIDDLE-NSTEMI, MATRIX, DETO2X-AMI, ECLS-SHOCK, MOSCA-FRAIL, STREAM-2, DANAMI-2, CTT, 4S, HPS, TNT, IMPROVE-IT, FOURIER, ODYSSEY OUTCOMES and its mortality analysis, CLEAR Outcomes and its total-event analysis, ORION-10/11 and the pooled inclisiran analysis, REDUCE-IT, SAMSON, Lp(a) apheresis and Mendelian-randomization work, ATT, CURE, CAPRIE, TRITON, PLATO, PEGASUS, TWILIGHT, MASTER DAPT, DAPT Study, GLOBAL LEADERS, HOST-EXAM and its extended follow-up, TICO, POPular AGE, STOPDAPT-2, SMART-CHOICE, WOEST, PIONEER AF-PCI, RE-DUAL PCI, AUGUSTUS, ENTRUST-AF PCI, COMPASS, CANTOS, CIRT, COLCOT, LoDoCo2, CLEAR SYNERGY, the colchicine meta-analysis, RESCUE/RESCUE-2, the IL6R Mendelian randomization, CHIP cohorts, CorMicA, EDIT-CMD, WISE/WISE-CVD, the MINOCA OCT/CMR study, SWEDEHEART MINOCA, PROSPECT, SCOT-HEART's two analyses, MESA CAC work, the hs-troponin algorithm literature, High-STEACS, MASS II, FREEDOM health status, ISCHEMIA quality of life, PolyIran, SECURE, SELECT, the cardiac-rehabilitation meta-analyses and Cochrane reviews, RAMIT, PREDIMED, the Lyon Diet Heart Study, the smoking-cessation syntheses, and the GBD burden, LDL-C, air-pollution and quality-of-care analyses.

Promotion and remaining flags

  • All 16 pages passed and were promoted to status: curated; INDEX.md and CONDITIONS-ROADMAP.md updated. Final validation: every page has matching body and reference PMID sets, sequentially numbered reference lists, 175–201 lines, 29–38 unique records per page, 252 condition-wide distinct records (15.8 per page — the condition has cleared the NARROW breadth flag recorded in the roadmap), and all local Markdown links resolve.
  • Remaining limitations, stated rather than hidden: a small number of citations are to documents PubMed indexes without an abstract (Falk 1995, Braunwald 2013, Samuel 2021, Conti 2010) or to guideline/statement documents whose recommendation text is not machine-verifiable from the record; for these the audit could confirm identity, authorship and topic but not each paraphrase against full text. Two exact recommendation-grade claims that were checkable — the ESC/EACTS left-main Class I/A versus IIa/A grades and the DAPT class-lineage argument — matched their sources. The joint ESC/EACTS left-main review is cited under both of its co-publication PMIDs (37632756, 37632766); both resolve and both are in the bibliography. Official-organization web sources were not re-fetched in this pass.
  • Next: normal weekly sweep. The first sweep should re-check ZEUS, Lp(a)HORIZON, OCEAN(a)-Outcomes and ORION-4 for primary outcome publications, and re-fetch the patient-voice and guideline web sources whose access dates are now older than a week.