What changed: Completed all 18 canonical wiki pages. Preserved and extended the 8 seeded pages; added alk-ros1-and-fusion-drivers.md, kras-braf-met-her2.md, immunotherapy.md, early-stage-and-perioperative-therapy.md, systemic-therapy.md, biomarkers.md, guidelines.md, clinical-trials-landscape.md, patient-experience-and-advocacy.md, and red-flags-and-safety-concerns.md. Every page is 150–400 lines, has current frontmatter, TL;DR, dense cited sections/tables, open questions, canonical related links, and references. Status remains draft pending the separate audit session required by the playbook.
Literature layer: Built BIBLIOGRAPHY.md with 262 unique PubMed records and every citing page; six landmark notes (IPASS, TCGA, FLAURA, ADAURA, CheckMate 816, Sherlock-Lung); a worldwide/current guideline registry with supersession chains, disagreements and watch list; a quantitative statistics/effect-size ledger; and a four-file patient-voice layer with method/ethics, 12 live-fetched organizations, aggregate themes, and annotated sources.
Research agenda: Replaced the 8-question seed with 30 stable questions: 18 Tier 1 practice-changing designs and 12 Tier 2 enabling questions. Added 12 “Dots not yet connected” junctions linking screening, pollution, pathology, MRD, CNS therapy, evolution, immunobiology, ADC safety, perioperative therapy, plasma/tissue, survivorship, and trial access.
PubMed searches: Live E-utilities searches covered existing condition citations plus shared NSCLC evidence and targeted queries for ALK inhibitors/ALEX/CROWN/ALINA; ROS1 crizotinib/entrectinib/repotrectinib; RET selpercatinib/pralsetinib/LIBRETTO-431; NTRK/larotrectinib; KRAS G12C/CodeBreaK/KRYSTAL; BRAF V600E doublets; MET exon-14 capmatinib/tepotinib and resistance; HER2/T-DXd/zongertinib; driver-positive immunotherapy; STK11/KEAP1; perioperative chemo-immunotherapy; ctDNA/MRD; stigma, caregivers, financial toxicity, trial access and supportive care. Final whole-condition validation resolved all 265 unique PMIDs present anywhere in the condition in two ESummary batches (150 + 115); the wiki bibliography contains the 262 unique PMIDs cited by canonical pages.
ClinicalTrials.gov searches: Live API v2 queries covered Non-Small Cell Lung Cancer, Lung Adenocarcinoma, osimertinib, amivantamab, lorlatinib/ALK, repotrectinib/ROS1, KRAS G12C, HER2-mutant, MET exon-14, RET fusion, ctDNA/MRD, and perioperative NSCLC. Snapshot counts: 490 active and 578 completed records indexed to lung adenocarcinoma; 2,429 active and 3,247 completed indexed to NSCLC. Final record-by-record validation resolved all 42 unique NCT identifiers used in this condition; no failed NCT remained.
Public-source verification: Live HTTP fetches returned 200 for NICE NG122/NG12, the ESMO early/local living guideline, the IASLC molecular-testing resource, CAP cancer protocols, USPSTF screening, and 12 patient/advocacy organization sites. The ASCO living-guideline URL returned HTTP 403 to automated retrieval but its URL and archived PMID were verified; the registry labels this access limitation. NCCN professional guidance remains authenticated-access-only and is explicitly flagged for point-of-use version checking. A direct CAP PD-L1/TMB page URL tested returned 404, so the registry cites the live-resolved PMID rather than that stale URL.
Validation performed: 18 wiki pages; 2,939 lines; all page frontmatter dates/statuses and required sections present; all local Markdown links resolve; no unresolved citation flags remain; every wiki PMID appears in the bibliography; 265/265 unique condition PMIDs resolve live; 42/42 unique NCT IDs resolve live. No files under the excluded thoracic-aortic-aneurysm, fibromyalgia, tools, or frontend paths were modified, and no Git command was used.
Check in the separate audit session: Re-fetch each claim→PMID pair and compare titles, author/year/journal, numerical values, and abstract support; correct bibliographic abbreviations where needed; re-run absence claims; verify current ASCO/NCCN/ESMO versions with human/authenticated access; inspect ClinicalTrials.gov site-level recruitment rather than sponsor-level status; then promote only pages that pass to curated and change the roadmap from built to audited.
Scope completed: Audited all 18 canonical wiki pages and all 13 literature artifacts: the 270-record bibliography, guideline registry, statistics ledger, six landmark notes, and four patient-voice files. The wiki contains 2,961 lines, 1,086 PMID occurrences representing 268 unique page-cited records, and 42 unique NCT identifiers. The literature layer contains 474 PMID occurrences representing 270 unique records; two records are registry-only context rather than wiki citations.
Live citation verification: Re-fetched all 270/270 unique condition PMIDs through PubMed E-utilities in this audit session and compared identifiers, titles, authorship/year/journal metadata, numerical claims, and abstract support. Re-fetched all 42/42 unique NCT records through ClinicalTrials.gov API v2 and checked study identity, recruitment status, phase, and enrollment fields. No unresolved PMID or NCT identifier remains.
Absence and currency searches: Re-ran live PubMed searches for randomized repotrectinib-versus-entrectinib evidence, direct alectinib-versus-lorlatinib comparisons, ctDNA/MRD-guided adjuvant intervention results, never-smoker LDCT mortality trials, EGFR oligoprogression randomized evidence, and STK11/KEAP1-directed immunotherapy evidence. Re-ran ClinicalTrials.gov disease-level and topic-level searches; the dated 2026-08-29 snapshots remain 490 active and 578 completed records indexed to lung adenocarcinoma and 2,429 active and 3,247 completed records indexed to NSCLC.
Errors found and fixed: 50 discrete defects were corrected: 19 incorrect first-author attributions in the patient-voice source ledger; 2 incorrect or stale ASCO living-guideline records/version fields; 7 ClinicalTrials.gov phase fields; 1 misspelled central-airway registry author; 1 superseded ROS1 organization URL in 2 local locations; 3 stale or overbroad evidence statements (STK11/KEAP1, MRD-guided treatment, and first-line EGFR intensification); and 17 PMID entries cited in page bodies but omitted from 3 page reference lists. Current ASCO records were replaced with live-verified PMIDs 42607282, 42190143, and 42190141; the emerging CTLA-4/STK11-KEAP1 evidence was added with PMID 39385035; and active MRD protocols were added with PMIDs 39659920 and 41628935.
Claims corrected or softened: The STK11/KEAP1 discussion now distinguishes adverse prognosis and PD-(L)1 resistance from the subgroup/translational evidence supporting a CTLA-4-containing strategy, without claiming validated biomarker-directed selection. MRD language now states the dated positive gap—interventional protocols exist, but mature randomized survival evidence for changing therapy from postoperative ctDNA was not identified. EGFR intensification language now separates proven aggregate PFS benefit from the unresolved patient-selection question. Trial tables now use registry-reported phase and actual-versus-estimated enrollment labels.
Structural validation: All 18 pages remain within the required 150–400-line range (150–202 lines), contain the required frontmatter and sections, have every body PMID represented in the page reference list, and resolve all local Markdown links. The bibliography covers every PMID used anywhere in the condition. There are zero unresolved verification markers and zero unresolved substantive audit issues.
External-source limitations retained explicitly: NCCN professional guidance remains authenticated-access-only and requires point-of-use version checking. ASCO living-guideline web pages returned HTTP 403 to automated retrieval, so current recommendations are anchored to their live PubMed records and the access limitation remains recorded. These are access/version-control cautions, not unresolved claim-to-citation failures.
Promotion: All 18 wiki pages passed and were promoted from draft to curated. INDEX.md, the literature status labels, and CONDITIONS-ROADMAP.md were updated to audited.