Real-world evidence still has to earn the name

Real-world evidence gets treated, in a lot of regulatory strategy conversations, as a category of data — claims records, registry entries, electronic health record extracts, anything gathered outside a controlled investigation. FDA's own framework doesn't define it that way, and the distinction is the whole point of the guidance. Real-world data is the raw material: any data relating to patient health or device use collected outside a traditional clinical trial. Real-world evidence is something narrower — the clinical evidence a sponsor derives from that data through an analysis adequate enough for FDA to actually rely on it. A dataset doesn't become evidence just by being real-world in origin; it has to earn that status through the analysis built on top of it.

The gap between data and evidence is where a submission actually lives

FDA's finalized guidance, Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices, published December 18, 2025, replaces the agency's 2017 framework and finalizes a draft FDA circulated in 2023. Its core structure hasn't changed across that revision: FDA evaluates whether a given source of real-world data is of sufficient relevance and reliability to generate evidence usable in a specific regulatory decision, and that evaluation is the actual gate. The underlying standard is still the same one that governs every other kind of clinical evidence a device submission relies on — the general definition of valid scientific evidence in 21 CFR 860.7(c)(2), which recognizes well-documented case histories and significant human experience with a device alongside controlled investigations. Real-world evidence isn't a separate, lower bar; it's real-world data being tested against the same bar other evidence has always had to clear.

Relevance and reliability are the sponsor's burden to document

The December 2025 guidance is more directive than its predecessor about who does that evaluation first: it directs sponsors to conduct and submit their own relevance-and-reliability assessment of the real-world data source, the study design built on it, and the analytic methods applied — not to hand FDA a dataset and let the review team reconstruct that case from scratch. Relevance asks whether the data actually captures the population, exposure, and outcomes the regulatory question is about; reliability asks whether the data was accurately and completely captured and can be verified. A submission that cites a large registry without addressing either question directly hasn't supplied real-world evidence yet — it's supplied real-world data and left the harder half of the guidance's own test unanswered.

What the 2025 update actually loosened

The most consequential change isn't conceptual, and it's worth naming plainly: earlier practice generally expected a submission relying on real-world data to include the underlying identifiable individual patient-level records for FDA to review directly. The finalized guidance moves away from that default for certain device submission types, allowing FDA to rely on a sponsor's documented relevance-and-reliability assessment and analysis without always requiring the patient-level data itself to be submitted alongside it. That's a real loosening of a practical barrier, not a loosening of the underlying standard — the assessment still has to be built and still has to hold up, and FDA gave itself and industry a stated runway, roughly 60 days from publication, before expecting submissions to reflect the new recommendations.

The data was already flowing in; this names what happens to it

None of the sources this guidance is written around are new to a device's file. A post-market surveillance system is already reaching into registries and PMCF data, and Clause 8.2.1's feedback process is already pulling in field and production data neither system originally built with an FDA submission in mind. What the RWE guidance adds isn't a new pipe for that data — it's a documented test the data has to pass before whatever it shows can be cited as evidence in a premarket or postmarket regulatory decision, rather than treated as background context a reviewer takes on faith.

Where this meets the file

A real-world-evidence worksheet built around the guidance's own test — the data source named, relevance addressed against the specific regulatory question, reliability addressed against how the data was captured and verified — is previewed in the launch catalog. If your program documents this assessment differently, the shelf takes that correction directly.