Have we exhausted the precedent — or are we still guessing?

FDA expectations for real-world evidence supporting a new indication or labeling change

Methodology

I searched the FDA guidance library (Drugs@FDA / CDER, CBER, and CDRH guidance documents) for the real-world evidence (RWE) and real-world data (RWD) policy set, plus adjacent evidentiary-standard guidances. Two broad discovery passes returned the guidance inventory with issue dates and draft/final status; I then ran eight targeted content queries against the corpus (new indication/effectiveness, labeling changes, data relevance and reliability, study design, registries, EHR and claims data, submission and agency engagement, and the RWE Program framework). I reviewed roughly 400 guidance chunks and cross-referenced each substantive claim to the specific guidance title and its draft/final status so that draft recommendations are not read as settled policy.

The governing documents (chronological)

Device track: Use of RWE to Support Regulatory Decision-Making for Medical Devices, first final 2017-08-31 ³², draft revision 2023-12-14 ⁵⁰, and re-finalized 2025-12-18 ⁵¹.

What FDA expects

1. Fit-for-purpose is the master test: relevance + reliability

RWD must be shown fit for the specific regulatory question through two lenses. Relevance = the data capture the key study variables (exposure, outcomes, covariates, follow-up) and enough representative patients for the intended-use population ^{278216222207288}. Reliability = accuracy, completeness, and traceability to provenance, with documented quality controls ^278216222207. The judgment is context-specific; when multiple sources are combined, each is assessed individually and in aggregate, and the assessment plus limitations are documented in the protocol and study report ^280281292.

2. RWE for a new indication / effectiveness is case-by-case, not a shortcut

RWE/non-interventional evidence can contribute to a demonstration of substantial evidence, but only with appropriate design and analysis features, a scientific justification, and early discussion with FDA ^317305303. The underlying substantial-evidence bar is unchanged: generally one highly persuasive AWC trial plus strong confirmatory evidence, or more than one AWC investigation ^298304296. RWE most often enters as the confirmatory piece, from a source relevant and reliable for the intended use with prespecified statistical methods ³⁰².

3. Study design must support an unbiased, interpretable comparison

FDA expects a protocol finalized before start, with prespecified data sources, eligibility, exposure definitions, endpoints, analysis plan, and missing-data/bias mitigation ⁹. Make treatment and external-control populations as similar as possible on outcome-related factors ⁷¹²; prespecify confounder and bias analysis ³; ensure a comparable index date / time zero to avoid immortal-time bias ⁵⁶; handle missing data with sensitivity analyses ⁴¹³. FDA is explicit that unmeasured confounding and lack of blinding cannot be eliminated, only assessed and addressed analytically ³⁴.

4. Source-specific expectations

• Registries: usable when structured, predefined, longitudinal, and fit-for-use; assess representativeness, completeness, and traceability; put governance, audit, data dictionaries, and validation in place early; ensure patient-level data can be provided to FDA even when a third party owns the registry ^220217222229. (Final ²⁷.)
• EHR and claims: choose sources that adequately capture populations, exposures, outcomes, covariates, and comparators, or explain gaps in the protocol ¹⁹⁵²⁰⁹. Claims are more accurate for filled outpatient prescriptions; EHRs for in-facility administration ²⁰². Validate outcome algorithms; protocols and SAPs should be submitted before conducting the study when data will support a decision ²⁰⁷. (Final ²⁶.)

5. Labeling changes: the coverage gap

The search did not surface a general FDA guidance stating that RWE broadly supports drug labeling changes. The clearest RWD-to-label statement is on the device side: RWD may support expansion of the indication for a cleared device ²⁶⁷. For drug labeling, related statements are indication- or safety-specific and reference clinical evidence generally rather than RWE ²⁶⁵²⁵¹.

How expectations have shifted

1. From framework to operational detail (2018 to 2024+). The 2018 Framework set the ambition while flagging RWD's bias limitations ¹⁴¹¹⁴⁸. FDA then built a source-by-source operational layer: registries (2023 ²⁷), EHR/claims (2024 ²⁶), data standards (2023 ⁵²). The center of gravity moved from "whether RWE can be used" to "how to demonstrate a source is fit-for-purpose."
2. Fit-for-purpose hardened into consistent vocabulary across every newer document ^278216222207.
3. Design rigor tightened through the draft externally-controlled-trials (2023 ³⁹) and non-interventional (2024 ³³) guidances; a 2024 draft signals a preference for embedding randomization into real-world settings ⁵⁶.
4. International harmonization for safety arrived with M14 (final 2026 ³⁶).
5. The effectiveness bar stayed put while RWE's confirmatory role got clarified ^306299298.

Bottom line

FDA will consider RWE for a new indication, but the sponsor must prove the data source is fit-for-purpose, use a prespecified design that controls bias, meet data-standards and transparency requirements, and engage the review division early. RWE most commonly lands as confirmatory evidence inside an unchanged substantial-evidence standard. There is no general drug-labeling RWE guidance; the explicit RWD-to-label pathway currently lives on the device side ²⁶⁷.