Real-world evidence (RWE), clinical evidence derived from data collected outside a traditional randomized controlled trial, claims, electronic health records, registries, has been part of FDA's regulatory toolkit since the 21st Century Cures Act mandated a formal program in 2016. The framework has now matured considerably, with new device-specific guidance finalized in late 2025. The useful question for anyone building an evidence strategy is not whether FDA accepts RWE in principle, it is exactly where and how it has actually moved regulatory decisions.
What changed with the 2025 device guidance?
FDA's Center for Devices and Radiological Health, jointly with the Center for Biologics Evaluation and Research, issued final guidance, "Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices," clarifying when real-world data sources are considered "fit for use" to support premarket submissions, expanded indications, or postmarket surveillance requirements. Source: FDA guidance, RWE for Medical Devices, Federal Register notice, December 2025.
This built on FDA's original Framework for FDA's Real-World Evidence Program, published in 2018, which first laid out the scope of RWE use under the Cures Act, including generating evidence for safety, supporting label expansions, and informing postmarket study requirements. Source: Framework for FDA's Real-World Evidence Program, 2018. The 2025 device-specific guidance is best understood as an operational refinement of that framework, not a new legal authority.
Where has RWE actually been decisive versus supportive?
FDA's own program materials distinguish several use cases where real-world data has carried different amounts of regulatory weight. Source: FDA use of RWE in regulatory decision-making.
Postmarket safety surveillance is where RWE is most established and most decisive. Adverse event detection, signal identification, and safety label updates routinely rely on real-world claims and registry data, and have for years, this is the least controversial and most mature use.
Label expansion for already-approved products is a growing but more conditional use case. RWE has supported expanded indications when a randomized trial for the new population would be impractical or unethical, but FDA has generally required the real-world data to meet a high bar for data quality, completeness, and freedom from confounding, closer to a registry-based trial than a simple claims analysis.
New product approval based primarily on RWE remains rare. The large majority of new drug and device approvals still rest on prospective, often randomized, trial data, with RWE playing a supportive or contextualizing role, such as providing an external control arm for a rare disease trial where randomization is infeasible, rather than serving as the primary evidence.
What does the inspection and quality literature say about the gap between guidance and practice?
A 2025 review in Therapeutic Innovation & Regulatory Science, "Keeping the End in Mind: Reviewing U.S. FDA Inspections of Submissions including Real-World Data," examined how FDA's own inspection processes have handled submissions incorporating real-world data, and found that data provenance, traceability, and quality control remain the most common points of regulatory friction, more so than the underlying clinical question the RWE was meant to answer. Source: Reviewing FDA Inspections of Submissions including Real-World Data, 2025.
This is a practically important finding for any company planning an RWE-based regulatory strategy: the bottleneck is frequently not whether the clinical question is answerable with real-world data in principle, it is whether the sponsor can demonstrate the data pipeline meets the same auditability standard as a clinical trial database, source data verification, consistent case definitions, documented data lineage from EHR extraction through analysis.
What does this mean practically for a device or drug sponsor?
Three implications follow directly from how RWE has actually been used, rather than how the guidance describes it in principle.
Postmarket and label-expansion use cases are the most tractable near-term opportunity. A device or drug already approved for a related indication, with a large observational dataset showing consistent effect in an adjacent population, is a stronger RWE case than trying to use real-world data as the sole basis for initial approval.
Data quality infrastructure is the actual gate, not the clinical argument. The inspection literature suggests sponsors should invest as much rigor in data provenance and audit trails as in the statistical analysis plan, since that is where submissions are most often challenged.
External control arms remain the most FDA-accepted RWE application for approval-track submissions, particularly in oncology and rare disease, where a concurrent randomized control is genuinely difficult to enroll.
What this does not prove
The 2025 guidance does not lower FDA's evidentiary standard for safety and effectiveness, it clarifies how real-world data can meet that standard, which is a different thing. It does not mean claims or EHR data of any quality will be accepted, fitness-for-use determinations remain data-source and question-specific. And the inspection literature reviewed here reflects a limited number of examined submissions, not the full universe of RWE-supported filings, so the friction points identified may not generalize to every therapeutic area or device class.
The takeaway
Real-world evidence has a genuine, expanding role at FDA, but it is concentrated in postmarket safety, label expansion, and external control arms, not primary approval evidence, and the newly finalized device guidance formalizes rather than loosens that pattern. The practical lesson from the inspection literature is that most RWE submissions fail on data infrastructure and provenance, not on the underlying clinical logic. Any company building an RWE strategy should treat data traceability as a regulatory requirement from day one, not a data-engineering afterthought.
