0011 Regulatory Approved Use Is Not Standard Of Care

Regulatory Approved Use Is Not Standard of Care

Status: accepted

FDA labels are authoritative evidence for what a regulator approved and the limitations stated in that label. They are not, by themselves, evidence that a treatment is accepted professional standard of care. Legacy Data Gov used the label pipeline to populate rows named as guidelines and standard of care; the useful intent was disease-facing approved-use decomposition, but the model collapsed two different claims.

The regulatory-approvals slice owns FDA label capture, document/version identity, source sections, and approved-use evidence. The label asset is therefore sources/regulatory/fda_label_snapshots, its physical source tables are sources_regulatory.fda_label_*, and its normalized records are normalized.regulatory_label_*.

The dedicated regulatory_label_source resource and lib/sources/regulatory_labels/ adapter remain separate from the approval-download adapter because the label endpoint has a different API, configured cohort, paging limit, document-version selector, and activation scope. This is a runtime variation boundary inside one regulatory domain owner, not a second domain.

Label-derived treatment contexts may appear in q_soc_review_candidates with source_standard_of_care='not_asserted'. They may publish as SOC only after either:

• a versioned professional-guideline statement supports the assertion; or
• an explicit human_approved curation decision promotes the exact candidate with assertion_kind='standard_of_care'.

Current live openFDA label sections fail closed before treatment-approach projection. A governed approved-use decomposition task must be implemented and evaluated before those compound sections can produce review candidates. That task remains regulatory extraction; it must not auto-publish SOC.

Considered Options

• Keep calling every regulator-approved use “standard of care.”
• Move FDA labels into the SOC source family because SOC consumes some label-derived context.
• Parse compound live label sections heuristically and publish whatever can be extracted.
• Keep one regulatory owner, expose approved-use review candidates, and require independent professional or reviewed evidence for SOC publication.

Consequences

• q_approval_indications, label evidence, SOC review candidates, and SOC recommendations remain distinct grains.
• Capturing a newer FDA label cannot silently create or remove a clinical recommendation.
• The absence of a professional-guideline source and live approved-use decomposition is an explicit activation blocker, not hidden fixture parity.
• Authored fixture expectations may exercise the downstream review workflow but are never evidence that the live decomposition path works.