GUIDE
Europe gives you twenty-two rules and expects you to apply them. The United States gives you a catalogue of roughly six thousand generic device types and expects you to find yours. Everything that follows — the class, the review pathway, whether you need a predicate — is read off the entry you land on, which makes finding the right one the whole job.
Written with Francesca Di Giuseppe, Senior QA/RA Specialist (LinkedIn)
Under review by Francesca Di Giuseppe, Senior QA/RA Specialist — this guide has not yet been through that review.
European classification is deductive. You take the intended purpose, apply the rules of Annex VIII in turn, and the class is whatever the strictest applicable rule produces. Two competent people working from the same intended purpose should reach the same class, and where they do not, they can argue about which rule governs.
American classification is not like that. Under section 513 of the Federal Food, Drug, and Cosmetic Act, the FDA classified existing devices by generic type and has gone on adding types ever since. Each type sits in a classification regulation in 21 CFR Parts 862 to 892 and carries a class. Your device is not classified by reasoning; it is classified by being recognised as an instance of a type that was classified already.
In Europe the difficulty is interpretive: does this device channel or store liquids within the meaning of Rule 2, does this software drive a therapeutic decision. In the United States the difficulty is one of identification: which of roughly six thousand generic types is mine, and is it really mine.
That is why this site classifies for the EU and searches for the US. It is not a feature gap. There is no American cascade to encode, and a tool that invented one would be describing a system that does not exist.
| CLASS | CONTROLS | USUAL PATHWAY TO MARKET |
|---|---|---|
| I | General controls | Most types are exempt from premarket notification. Exempt does not mean unregulated: registration, listing, labelling, and the Quality System Regulation still apply, and the exemption has limits written into the classification regulation itself. |
| II | General and special controls | 510(k) premarket notification, demonstrating substantial equivalence to a legally marketed predicate. Some Class II types are also exempt; the classification regulation says which. |
| III | Premarket approval | PMA — an independent demonstration of safety and effectiveness, not a comparison. The most demanding route, and the one where clinical evidence is normally unavoidable. |
NOTE —
A product code is three letters — FRN for an infusion pump, LKK for an electroencephalograph. It looks like an index number and behaves like a decision. Behind it sits an entry in the FDA's classification database holding, at minimum:
A minority of codes have no classification regulation behind them. These are types the FDA has recognised for administrative purposes without a codified classification, and the class shown against them is best read as proposed rather than settled. Our search marks these explicitly, because a class with no regulation to cite is not a class you can rely on in a submission.
FDA device names are written noun-first with the qualifiers trailing: Pump, Infusion, not Infusion pump. Catheter, Intravascular, Therapeutic, Short-term Less Than 30 Days, not the phrase any clinician would use.
This matters more than it sounds. Search for the words in the order you would say them and the base type — the one you probably want — ranks below its own derivatives, because the derivatives happen to contain your phrase and the base type does not. It is the single most common way people end up on the wrong code, and it is the reason our search treats the inverted form as a first-class match rather than a coincidence.
A device name is a label; the classification regulation is a definition. Two codes with similar names can have definitions that differ on exactly the point that matters — duration of contact, whether the device is powered, which part of the body. Landing on a plausible name and stopping there is how a submission gets a refuse-to-accept letter months later.
The two are routinely confused. Finding the right product code establishes that the FDA recognises your generic type and tells you which pathway applies. It does not establish that a suitable predicate device exists, and for a 510(k) you need one: a specific, legally marketed device whose intended use and technological characteristics you can show yours to be substantially equivalent to. That comparison is most of the submission, and no classification search can do it for you.
If no generic type describes your device, the 510(k) route is closed by definition — there is nothing to be equivalent to. Novel devices of low or moderate risk are handled by the De Novo pathway, in which you ask the FDA to classify the device into Class I or II on the basis of risk, creating a new generic type rather than joining an existing one. A granted De Novo produces a new classification regulation and a new product code, which subsequent devices can then use as a predicate.
Where a device is genuinely high-risk and novel, the route is PMA rather than De Novo, and the evidentiary burden is of a different order.
Section 513(g) allows a formal request to the FDA for information on the class and requirements applicable to a device. It produces the agency's view rather than a guess, and it is a legitimate step for a product sitting on a boundary. It is not free and it is not instant, but neither is choosing wrongly.
The recurring error is assuming the American class follows from the European one. Class IIa sounds like Class II, the numeral is the same, and there is a strong pull towards treating one as a translation of the other.
It is not. The European class is a conclusion about your device drawn from its intended purpose. The American class is a property of a generic type that the FDA classified, in some cases decades ago, on the evidence and risk understanding of the time. The two are answers to different questions, and the fact that they use overlapping numerals is an accident of drafting rather than a correspondence.
NOTE —
NOTE —