{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T18:35:26Z","timestamp":1786732526251,"version":"build-2736575974"},"reference-count":16,"publisher":"Public Library of Science (PLoS)","issue":"8","license":[{"start":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T00:00:00Z","timestamp":1786665600000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["www.plosone.org"],"crossmark-restriction":false},"short-container-title":["PLoS One"],"abstract":"\n The CDISC Rules Engine (CORE) provides industry-standard validation for Study Data Tabulation Model (SDTM) submissions, yet its imperative, domain-scoped architecture cannot express cross-domain contradictions\u2014such as a Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 overall response that contradicts the combined target lesion, non-target lesion, and new-lesion status\u2014which pass structural validators undetected. We present CAVE-Onc, a two-layer graph-constrained agentic validation engine that augments CORE with declarative Shapes Constraint Language (SHACL) validation (L1) and a LangGraph-based agentic orchestrator (L3). CAVE-Onc operates over an RDF knowledge graph built from XPT datasets. Layer 1 comprises 111 SHACL shapes\u201485 ported from CORE rules, 8 RECIST 1.1 derivation shapes, and 18 archetype-specific SHACL-SPARQL cross-domain constraint shapes\u2014while Layer 3 implements a CaveAgent for RECIST Table 7 verification. We evaluated CAVE-Onc against pre-registered hypotheses (OSF): Track A assessed complementarity with CORE on clean data (52 subjects); Track B evaluated detection of 20 contradiction archetypes injected via a RELREC-preserving generator. On Track A, CAVE L1 and CORE produced disjoint flag sets (Jaccard\u2009=\u20090.004). On Track B, CAVE detected 20\/20 archetypes (100%; 95% CI: 83.2%\u2013100.0%): 19 via L1 SHACL-SPARQL constraints and 1 via L3 agent exclusively. Two industry validators\u2014the CDISC CORE engine (8\/20) and the Pinnacle 21 FDA engine (6\/20)\u2014each detected 0\/10 of the cross-domain RECIST contradictions they cannot express (all detections were CORE-seeded structural checks), versus CAVE\u2019s 10\/10 (McNemar\n p<\/jats:italic>\n \u2009=\u20090.002). Three domain experts validated all 20 archetypes (Fleiss\u2019\n \n \n \n \n \n \u03ba<\/mml:mi>\n =<\/mml:mo>\n 0.705<\/mml:mn>\n <\/mml:mrow>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n ; 0 rated invalid). Because the archetypes were authored knowing the injected patterns, we frame this as construction validation of\n expressiveness<\/jats:italic>\n , not a real-world detection estimate; on five held-out archetypes, existing shapes detected 3\/5. On two real oncology trials mapped to SDTM, the engine stayed specific on unmutated data and detected 10\/11 and 16\/18 of applicable archetypes. Graph-based validation thus augments rather than replaces industry-standard tools.\n <\/jats:p>","DOI":"10.1371\/journal.pone.0350376","type":"journal-article","created":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T17:58:44Z","timestamp":1786730324000},"page":"e0350376","update-policy":"https:\/\/doi.org\/10.1371\/journal.pone.corrections_policy","source":"Crossref","is-referenced-by-count":0,"title":["CAVE-Onc: Graph-constrained agentic validation for cross-domain contradictions in CDISC oncology submissions"],"prefix":"10.1371","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-1786-7259","authenticated-orcid":true,"given":"Jaime","family":"Yan","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"340","published-online":{"date-parts":[[2026,8,14]]},"reference":[{"key":"pone.0350376.ref001","volume-title":"Study Data Tabulation Model Implementation Guide: Human Clinical Trials","author":"CDISC","year":"2022"},{"key":"pone.0350376.ref002","unstructured":"CDISC. CDISC Rules Engine (CORE). 2026. https:\/\/github.com\/cdisc-org\/cdisc-rules-engine"},{"key":"pone.0350376.ref003","unstructured":"Pinnacle 21. Pinnacle 21 community. 2024. https:\/\/www.pinnacle21.com\/community"},{"issue":"2","key":"pone.0350376.ref004","doi-asserted-by":"crossref","first-page":"228","DOI":"10.1016\/j.ejca.2008.10.026","article-title":"New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)","volume":"45","author":"EA Eisenhauer","year":"2009","journal-title":"Eur J Cancer"},{"key":"pone.0350376.ref005","volume-title":"Guidance for Industry: Providing Regulatory Submissions in Electronic Format \u2014 Standardized Study Data","author":"US FDA","year":"2021"},{"key":"pone.0350376.ref006","unstructured":"LangChain A. LangGraph: Framework for building stateful, multi-actor applications with LLMs. 2024. https:\/\/github.com\/langchain-ai\/langgraph"},{"key":"pone.0350376.ref007","volume-title":"Pharmaversesdtm: SDTM example datasets for the pharmaverse","author":"pharmaverse","year":"2024"},{"key":"pone.0350376.ref008","unstructured":"PHUSE. SENDConform: RDF\/OWL ontology and SHACL shapes for SEND conformance. 2023. https:\/\/github.com\/phuse-org\/SENDConform"},{"key":"pone.0350376.ref009","unstructured":"Novo Nordisk. Neo4j-based clinical metadata repository for CDISC standards. 2023. https:\/\/gitlab.com\/Novo-Nordisk\/nn-public\/openstudybuilder"},{"key":"pone.0350376.ref010","volume-title":"Shapes Constraint Language (SHACL)","author":"W3C","year":"2017"},{"key":"pone.0350376.ref011","unstructured":"RDFLib. pySHACL: A Python SHACL validator. 2024. https:\/\/github.com\/RDFLib\/pySHACL"},{"key":"pone.0350376.ref012","unstructured":"US FDA. Title 21 CFR Part 11: Electronic Records; Electronic Signatures. 2024. https:\/\/www.ecfr.gov\/current\/title-21\/part-11"},{"issue":"4","key":"pone.0350376.ref013","doi-asserted-by":"crossref","first-page":"404","DOI":"10.1093\/biomet\/26.4.404","article-title":"The use of confidence or fiducial limits illustrated in the case of the binomial","volume":"26","author":"CJ Clopper","year":"1934","journal-title":"Biometrika"},{"issue":"2","key":"pone.0350376.ref014","doi-asserted-by":"crossref","first-page":"153","DOI":"10.1007\/BF02295996","article-title":"Note on the sampling error of the difference between correlated proportions or percentages","volume":"12","author":"Q McNemar","year":"1947","journal-title":"Psychometrika"},{"issue":"5","key":"pone.0350376.ref015","doi-asserted-by":"crossref","first-page":"378","DOI":"10.1037\/h0031619","article-title":"Measuring nominal scale agreement among many raters","volume":"76","author":"JL Fleiss","year":"1971","journal-title":"Psychological Bulletin"},{"issue":"1","key":"pone.0350376.ref016","doi-asserted-by":"crossref","first-page":"159","DOI":"10.2307\/2529310","article-title":"The measurement of observer agreement for categorical data","volume":"33","author":"JR Landis","year":"1977","journal-title":"Biometrics"}],"container-title":["PLOS One"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0350376","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,14]],"date-time":"2026-08-14T17:58:48Z","timestamp":1786730328000},"score":1,"resource":{"primary":{"URL":"https:\/\/dx.plos.org\/10.1371\/journal.pone.0350376"}},"subtitle":[],"editor":[{"given":"Le","family":"Zhang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"editor"}]}],"short-title":[],"issued":{"date-parts":[[2026,8,14]]},"references-count":16,"journal-issue":{"issue":"8","published-online":{"date-parts":[[2026,8,14]]}},"URL":"https:\/\/doi.org\/10.1371\/journal.pone.0350376","relation":{},"ISSN":["1932-6203"],"issn-type":[{"value":"1932-6203","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,14]]}}}