A Global Medical Device Risk Grammar for Classification Interoperability

This record deposits a preprint methods paper introducing the Global Medical Device Risk Grammar (G-MDRG), a non-binding semantic-risk framework for medical-device classification interoperability. The paper addresses a recurring problem in global medical-device regulation: jurisdictional class labels, product codes, nomenclature systems, and classification rules can compress or obscure classification-relevant meaning. Such meaning may include intended use, clinical consequence, exposure and invasiveness, software autonomy, AI/adaptivity, cybersecurity dependency, evidence adequacy, lifecycle volatility, and vulnerable-population context. G-MDRG is proposed as an interpretive and documentation-oriented layer, not as a legal classifier, conformity-assessment pathway, regulatory approval tool, clinical decision system, or substitute for jurisdiction-specific expert, legal, regulatory-authority, FDA, notified-body, or competent-authority review. The framework separates baseline control intensity, lifecycle-governance concerns, and evidence-gap signalling through the Global Control Band (GCB), Lifecycle Governance Modifier (LGM), and Evidence Gap Modifier (EGM). It also introduces semantic-loss flags to identify cases where local classifications or product codes may not preserve risk-relevant meaning. A practical use of G-MDRG is early regional classification intelligence. For a given device, the framework can help identify which attributes are likely to drive classification concern across FDA, EU MDR, Swiss/EU-aligned, and SaMD/AI contexts. The resulting GCB, LGM, EGM, and semantic-loss outputs can support classification-rationale preparation, region-specific uncertainty mapping, evidence-gap identification, and prioritization of cases requiring expert, legal, or notified-body review. This is classification support and briefing, not binding classification determination. The manuscript includes feasibility-stage proxy evidence and an internal dry-run workflow. These analyses are intended to evaluate operational feasibility, rule stability, triage behaviour, semantic-loss detection, and expert-panel preparation. They do not constitute external expert validation, legal classification validation, regulatory validation, or evidence of global regulatory acceptance. Accordingly, the paper frames G-MDRG as a pre-normative regulatory-science contribution requiring future blinded expert adjudication, benchmark-corpus testing, usability evaluation, cross-jurisdictional validation, and governed codebook refinement before any stronger claims of validity, regulatory utility, or implementation readiness can be made.