(deep-research round 2, ASME V&V40 + FDA 2023 verified 3-0) Compiled 2026-06-20 to ground the risk-COU axis of the nx_sim_credibility roadmap. Verified vs PRIMARY ASME/FDA sources (15 claims confirmed 3-0). IEC 62304 / DO-178C / GCI claims = SOURCED but UNVERIFIED this run (verifiers abstained at the credit limit) -> marked PROVISIONAL below; verify before grading against them. === ASME V&V 40-2018 "Assessing Credibility of Computational Modeling through V&V: Application to Medical Devices" === RISK-INFORMED principle (verified): required model credibility must be COMMENSURATE with how heavily the model is relied on as evidence of device performance. MODEL RISK (verified) = model INFLUENCE (how much the model affects the decision, vs other evidence) x DECISION CONSEQUENCE (severity of harm from an incorrect decision; consequence is ISO-14971 risk = severity x probability of harm). Combine via a grid (FDA Fig 2: 3x3, or alternatively 5x5 / 5x4) -> model risk. ANCHORS (verified): Question of Interest (QOI) + Context of Use (COU = the specific role/scope of the model in informing a decision). Higher-risk COU => higher rigor => higher credibility GOAL. CREDIBILITY FACTOR CATEGORIES (verified): VERIFICATION (solve the math right), VALIDATION (model represents reality), APPLICABILITY (relevance of the validation evidence to the COU). Each factor has a GRADATION of progressively increasing rigor; a credibility GOAL is chosen from the gradation commensurate with model risk. Output-comparison gradation example: (a) qualitative comparison -> (b) quantitative point/profile matching -> (c) quantitative with uncertainty bands. LINEAGE (verified): developed by the ASME V&V 40 Subcommittee WITH the FDA + device makers; explicitly complements ASME solid-mechanics (V&V 10), CFD/heat (V&V 20), and the NASA credibility-assessment standard (NASA-STD-7009) -- this is the cross-standard hook to extend a NASA-7009-style ordinal census to the medical-device lane. WORKED EXAMPLES (verified): BBCT femur-fracture-risk model -> risk MEDIUM, achieved credibility judged acceptable; orthopaedic-implant FEA + pedicle-screw-cannulation QOI = published end-to-end V&V40 applications. SOURCES: ASME https://www.asme.org/codes-standards/find-codes-standards/assessing-credibility-of-computational-modeling-through-verification-and-validation-application-to-medical-devices ; framework paper https://pmc.ncbi.nlm.nih.gov/articles/PMC6493688/ ; worked example https://www.sciencedirect.com/science/article/pii/S0169260723003930 === FDA 2023 FINAL GUIDANCE "Assessing the Credibility of Computational Modeling and Simulation in Medical Device Submissions" === Issued 2023-11-17 (draft 2021-12-23). 9-STEP risk-informed credibility process (verified verbatim): 1) describe Question(s) of Interest; 2) define Context of Use (COU); 3) determine MODEL RISK (influence x consequence, per ASME V&V40; consequence per ISO 14971); 4) identify+categorize credibility evidence; 5) define credibility factors + GOALS (select from each factor's gradation commensurate with risk; a goal below the risk-commensurate level needs a documented rationale); 6) prospective adequacy assessment; 7) generate the credibility evidence; 8) post-study adequacy assessment (were goals met?); 9) prepare a CM&S credibility assessment report for the submission. 8 CREDIBILITY-EVIDENCE CATEGORIES (verified): (1) code verification, (2) model calibration, (3) bench-test validation, (4) in-vivo validation, (5) population-based validation, (6) emergent model behavior, (7) model plausibility, (8) calculation-verification/UQ using COU simulations. SOURCE (primary): https://www.fda.gov/media/154985/download === PROVISIONAL (sourced, NOT verified this run -- verify before grading) === IEC 62304 (med-device SW life cycle): three software SAFETY CLASSES A/B/C scaling lifecycle+documentation rigor by the harm a software failure can cause. SOURCE https://www.iso.org/standard/64686.html DO-178C / FAA AC 25.1309-1B (avionics, cross-domain): 5 ordinal failure-condition severities (No Safety Effect / Minor / Major / Hazardous / Catastrophic) mapped INVERSELY to 4 probability bands (Probable 1e-3..1e-5 / Remote 1e-5..1e-7 / Extremely Remote 1e-7..1e-9 / Extremely Improbable <=1e-9); maps to Design Assurance Levels DAL A-E. SOURCE https://www.faa.gov/documentLibrary/media/Advisory_Circular/AC_25.1309-1B.pdf NUMERICAL-ERROR ESTIMATORS (Roache GCI, per NIST IR 8298): GCI21 = Fs*|ea21|/(r21^p - 1); r21=h2/h1 grid ratio, p=observed order, Fs=1.25 structured / 3.0 unstructured, >=3 grids; GCI ~ 95% interval, u_h = GCI/k (k~2 Gaussian). SOURCE https://nvlpubs.nist.gov/nistpubs/ir/2020/NIST.IR.8298.pdf === APPLICATION TO THE NISHI SIM (nx_sim_risk_cou_gate) === Credibility is COU-RELATIVE. Our orbit sim: COU-A education/visualization (reproduce a KNOWN analytic orbit) = LOW risk, gating factor Verification(achieved 3) >= goal 1 => MEETS. COU-B engineering trade study = MED risk, UQ(2) >= goal 2 => MEETS. COU-C clinical/safety novel-real-world decision = HIGH risk, Validation(0) < goal 3 => GAP (documented rationale required; needs real-system experimental validation). This is the honest, regulator-aligned reading: credible-enough for low/med-risk use, NOT for high-risk regulated decisions.