How to Verify a Medtech Claim in the Public Record

← Back to All Insights

A deck crosses your desk. It says the device is FDA cleared, the pivotal trial is enrolling, the IP is granted, and a strategic is circling. Four claims, four sentences, no citations. An associate has a day.

Most of that is checkable. Not all of it, and the difference between the two is the part worth being careful about. Below is the sequence we use, which systems settle which claims, and — more useful — what each system structurally cannot tell you, because that is where a confident-sounding answer goes wrong.

The claim is a search key, not a sentence

The first move is translation. “FDA cleared” is not a fact you can look up; it is a compressed assertion containing at least four separable facts: that a submission exists, that it was cleared rather than merely filed, who the clearance belongs to, and what indication it actually covers. Each of those resolves in a different field of the same record, and they come apart more often than you would expect. A clearance held by a predecessor entity, or covering a narrower indication than the one being pitched, is not a false statement — it is a true statement doing more work than it can carry.

So before searching, write down what would have to be true. Then go find it.

The six systems, and what each one settles

1. FDA device clearances and approvals

The 510(k) Premarket Notification database is the authority for a clearance claim. A K-number resolves to a decision date, an applicant name, a device name and a product code, and, where FDA has posted them, to the 510(k) summary or statement that sets out the indications for use. FDA maintains a broader index of its public databases covering PMA approvals and De Novo grants alongside it.

Settles: whether a clearance exists, its date, and whose name is on it. Does not settle: whether the marketed product is the cleared product. Devices evolve after clearance; the record is a snapshot at decision, not a running description of what ships today.

2. Recalls, adverse events, and enforcement

The Medical Device Recalls database and MAUDE carry the safety history. Both are also available programmatically through the openFDA device APIs, which matters if you are checking a portfolio rather than a company.

Settles: that a specific recall or reported event occurred, its classification, and its date. Does not settle: rate, or comparative safety. MAUDE is a passive surveillance system, and FDA states that its reports are not intended to evaluate rates of adverse events, changes in event rates over time, or differences in event rates between devices. Report volume reflects reporting behavior and how widely a device is used as well as how it performs, so a raw event count is not a safety signal on its own.

3. Establishment registration and device listing

The Establishment Registration & Device Listing database answers a quieter question: is this company registered with FDA in the role it describes, and at the facility it describes? Manufacturers, contract manufacturers, specification developers and initial importers are among the establishment types that must register, and the obligations differ by type.

Settles: the declared role and site. Does not settle: quality-system condition. Registration is a filing, not an inspection outcome.

4. ClinicalTrials.gov

A trial claim resolves to an NCT number and a record with a recruitment status, an enrollment figure, a primary completion date, and — separately — a results section. The gap between registration and results is the interesting part, and for many trials it has a regulatory clock attached; we cover that clock in Reading a ClinicalTrials.gov Record for Diligence.

Settles: that a trial is registered, its declared design, and its stated status as of the last update. Does not settle: that enrollment is actually proceeding. A status field reflects the last time a sponsor updated it.

5. SEC EDGAR

EDGAR full-text search covers filings submitted electronically since 2001 and supports exact-phrase searches and operators such as OR and NOT. Its use in early-stage biomedical diligence is not the target — most targets do not file — but the counterparties: the strategic said to be circling, the acquirer of a comparable, the public company a founder previously ran. A transaction that was material to a public counterparty generally left a filing.

Settles: what a public counterparty told its own shareholders, under liability, on a date. Does not settle: anything about a private company. Absence from EDGAR is absence of an obligation to file, not absence of activity — a distinction worth holding firmly, because it is easy to read a blank search as a negative finding when it is not a finding at all.

6. USPTO

Patent Public Search covers U.S. patents and published applications; Patent Center shows application status and documents, and since September 2025 it requires identity-verified sign-in. “Patented technology” in a deck can mean an issued patent, a published application, a provisional, or a license to somebody else's estate. These are four different assets.

Settles: what exists, who is named as assignee, and where prosecution stands. Does not settle: whether the claims cover the commercial product, or whether anyone else's claims read on it. That is a claim-chart question, not a database question.

Reading the silences

The hardest discipline in public-record work is refusing to convert an empty result into a finding. These systems have structural gaps that have nothing to do with the company in front of you: a foreign manufacturer with no US registration obligation, a subsidiary whose filings sit under a parent's name, a company that changed names after an acquisition and left its history under the old one, a device regulated under a pathway you searched the wrong database for.

An empty result is a prompt, not a conclusion. The honest formulation is “the public record does not address this,” and the correct next step is to ask the company, in writing, for the identifier. A company that has the K-number can send it quickly. A company that cannot produce one has told you something — but it has told you through the request, not through your search.

Where the name itself is the problem

One practical note saves a lot of time. Corporate identity is not consistent across these systems: FDA records carry the applicant name at the time of the submission, EDGAR organizes filings by registrant, and USPTO shows the assignee as recorded. A rename, reincorporation or acquisition can leave part of a company's history under an older or different name. Searching one spelling of one name and finding nothing is an easy way to produce a false negative. Search the predecessor, the parent, the abbreviated form, and the founder as an individual inventor before concluding anything.

Where this stops scaling

For one company and four claims, the sequence above is an afternoon, and an associate who runs it will know more about the target than the deck disclosed. It holds up.

What it does not survive is volume. Run it across a pipeline — thirty companies a quarter, each with a dozen checkable assertions, each requiring the name-variant discipline above and a cross-reference against comparable companies to mean anything — and the afternoon becomes the job. That is a structural limit of doing it by hand, not a failure of the method, and it is worth naming plainly rather than discovering it in month four.

Vantage assembles these systems into one substrate and cites every factual claim in a report back to the record it came from, so a reader can check us the same way we would check a deck. We do not forecast whether a company succeeds; the record does not support that and we do not claim it. What the record supports is a complete, sourced account of what is actually documented — including, explicitly, where it is silent.