Predicate Device and Substantial Equivalence: Choosing and Justifying a 510(k) Predicate

Scope of this page: how to choose and justify a predicate device for a 510(k) submission — the two-part substantial equivalence (SE) test, a practical decision tree for working through it, what changes when you cite more than one predicate, and why a candidate predicate’s recall or safety history matters before you commit to it. This assumes you already know a 510(k) is the right pathway for your device; if you need that background first, see the 510(k) premarket notification guide.

What a predicate device actually is

A predicate device is a device already legally marketed in the United States that a 510(k) sponsor cites as the benchmark for substantial equivalence. The requirement comes from 21 CFR 807.92(a)(3): a 510(k) summary must identify the predicate(s) the submission relies on. A predicate is typically one of two things — a device already cleared through its own 510(k), or a device legally marketed before May 28, 1976 (the date the Medical Device Amendments took effect) that was never required to go through premarket review at all. A device that reached the market through De Novo classification can also serve as a predicate for later 510(k)s, since De Novo creates a new classification regulation rather than granting equivalence to something else — see the De Novo classification entry if your device has no existing predicate at all.

What a predicate is not: it doesn’t have to be the newest device on the market, the most similar-looking device, or a device from the same manufacturer. It has to be the device that best supports the specific equivalence argument you’re making, which is a narrower and more deliberate choice than it sounds.

The two-part test: intended use, then technological characteristics

Substantial equivalence is not a single yes/no judgment — it’s two sequential questions, and the second one only gets asked if the first is answered correctly.

1. Same intended use. Your device and the predicate must have the same intended use. This is compared first because it’s a threshold question: if the intended uses differ, FDA generally will not proceed to compare technological characteristics at all — a different intended use is its own basis for finding the device not substantially equivalent (NSE), regardless of how similar the technology is.
2. Same technological characteristics, or different characteristics that don’t raise new questions of safety or effectiveness. Once intended use is established as the same, FDA compares design, materials, energy source, performance specifications, and other technological characteristics. If they’re the same, that alone supports SE. If they’re different, SE is still available — but the sponsor has to show the differences don’t raise new questions of safety or effectiveness, and where they might, has to provide performance data (bench, animal, or clinical, as appropriate) demonstrating the device is at least as safe and effective as the predicate.

“Intended use” and “indications for use” are related but not identical for this purpose: intended use is the general purpose of the device (what it does, in what context), while indications for use is the specific disease, condition, or patient population. FDA’s SE comparison is anchored to intended use; a broader or narrower indication than the predicate can still support equivalence if the underlying intended use matches, but it’s a common place submissions get RTA (Refuse to Accept) holds when the intended use statement drifts from what the predicate was actually cleared for.

The substantial equivalence decision tree

Worked through in order, the logic FDA reviewers apply looks like this:

1. Does the candidate device have the same intended use as the proposed predicate?
   • No → not substantially equivalent on this predicate. Look for a different predicate with a matching intended use, or consider whether your device needs to go through De Novo or PMA instead.
   • Yes → continue.
2. Does it have the same technological characteristics as the predicate?
   • Yes → substantially equivalent. No performance testing is required to establish SE itself (though testing may still be needed for other reasons, e.g. design controls or a recognized-standard declaration of conformity).
   • No → continue.
3. Do the different technological characteristics raise new questions of safety or effectiveness?
   • No → continue to the next question.
   • Yes → continue, but expect FDA to require more extensive performance data before it will accept the comparison.
4. Are there accepted scientific methods available to evaluate the effect of the differences, and does the submitted performance data show the device is at least as safe and effective as the predicate?
   • Yes → substantially equivalent, with the technological differences supported by the data submitted.
   • No (the differences can’t be adequately evaluated, or the data doesn’t support equivalence) → not substantially equivalent (NSE).

An NSE determination isn’t a paperwork setback — by operation of law it places the device into Class III automatically. From there the sponsor’s remaining options are a Premarket Approval (PMA) application, a De Novo request (if the device is actually low-to-moderate risk despite lacking a predicate), or a reclassification petition. That downstream cost is exactly why predicate selection deserves real diligence before a 510(k) is drafted, not after an NSE letter arrives.

Choosing a predicate that will actually hold up

Several predicates may technically be eligible on paper; not all of them are equally defensible in review. Considerations that consistently matter in practice:

• Well-characterized technology. A predicate whose technological characteristics are clearly documented in its own 510(k) summary and labeling makes the comparison easier to write and easier for a reviewer to accept. A predicate with a thin public summary forces you to reconstruct characteristics indirectly, which weakens the comparison.
• Avoid predicate creep. A predicate that is itself several generations removed from the original device it was compared to — where each step introduced a small technological change — can accumulate differences from the true origin device that no single 510(k) in the chain ever had to justify all at once. FDA has flagged this pattern (sometimes called predicate creep) as a concern, particularly where the cumulative drift touches safety-relevant characteristics. Where a more recent, more directly comparable predicate exists, it’s usually the stronger choice over an older device several links back in a chain.
• Currency. A predicate cleared many years ago under older performance standards or a since-updated recognized consensus standard can require you to bridge the gap yourself (showing your device still meets the current expectations even though the predicate was cleared under older ones). A more recently cleared predicate against current standards is usually less work to defend.
• Real market presence. A predicate that was cleared but never actually commercially marketed, or has since been discontinued for reasons unrelated to performance, is still legally eligible, but reviewers increasingly probe why a still-marketed alternative wasn’t used instead.

Multiple predicates and split predicates

21 CFR 807.92(a)(3) refers to the predicate device(s) plural — a 510(k) is not limited to a single predicate. Two related but distinct patterns show up in practice:

• Multiple predicates for the same intended use. A sponsor can cite two or more devices that each independently support the same intended use and technological comparison, strengthening the argument that the intended use and characteristics claimed are genuinely established in the marketplace rather than resting on one narrow prior clearance.
• Split predicates (a primary predicate plus reference devices). This is the more common real-world pattern for devices that combine established elements in a new way: the sponsor cites one predicate to establish the intended use, and separately cites one or more reference devices to support specific technological characteristics the primary predicate doesn’t itself have. The reference device isn’t claimed as the intended-use predicate — it’s cited narrowly, to show a particular design element, material, or energy source is already accepted in a legally marketed device. This only works when the intended use itself still traces back to a genuine predicate; reference devices can support individual characteristics, but they cannot substitute for the intended-use predicate at the first step of the decision tree.

The practical risk with either pattern is overreach: stitching together technological characteristics from several unrelated devices to construct a combination that doesn’t actually exist in any single marketed product invites an NSE finding, because FDA is evaluating whether your specific device — as a whole — is substantially equivalent to something real, not whether each individual feature has appeared somewhere before. Keep the primary predicate doing the intended-use work, and use reference devices for well-bounded, specific characteristics only.

Predicates subject to recall or unresolved safety issues

A device that is legally marketed can still be a poor predicate choice if it carries safety history that undercuts the comparison you’re trying to make. Before committing to a predicate, check its status:

• Class I recalls (reasonable probability of serious adverse health consequences or death) on the candidate predicate are the strongest reason to look elsewhere. Citing a device with an unresolved Class I recall as your safety benchmark undermines the core premise of a 510(k) — that the predicate is a safe and effective baseline you’re equivalent to.
• Class II and Class III recalls don’t automatically disqualify a device as a predicate, but the reason for the recall matters: a recall tied to a labeling error or a supply issue unrelated to the device’s core technological characteristics is a different situation from a recall tied to the exact design element you’re relying on for the equivalence argument.
• MDRs (adverse event reports) and warning letters against the predicate manufacturer for that device are worth checking even absent a formal recall — an accumulating pattern of reported problems is the kind of signal a reviewer may independently notice and ask about, even if it hasn’t triggered a recall.
• Withdrawn or expired clearances matter too: a predicate whose clearance was later withdrawn, or that is no longer legally marketed for reasons tied to safety or effectiveness, is not a sound choice even though the original 510(k) technically still exists in FDA’s database.

FDA maintains public device recall data, and checking a candidate predicate against it before filing costs little relative to the risk of an SE argument built on a device FDA already has open safety concerns about. For the broader vocabulary around recalls and how it maps across the FDA/EU MDR frameworks, see field safety corrective action, recalls, and field safety notices.

When there isn’t a defensible predicate

Sometimes the honest answer, after working through the decision tree above, is that no existing device is a good match — the intended use is genuinely novel, or every technologically similar device carries characteristics you can’t adequately bridge with data. Forcing a 510(k) onto a weak predicate at that point usually just delays an NSE finding rather than avoiding it. If the device is nonetheless low-to-moderate risk, De Novo classification lets FDA create a new classification regulation directly, without requiring a predicate at all — see the 510(k) vs. PMA comparison and the device classification and pathways guide for how the three routes (510(k), De Novo, PMA) relate to each other.

Frequently asked questions

What is a predicate device?

A device already legally marketed in the U.S. — either previously cleared through its own 510(k), marketed before the 1976 Medical Device Amendments took effect, or reached the market via De Novo — that a 510(k) sponsor cites under 21 CFR 807.92(a)(3) as the benchmark for demonstrating substantial equivalence.

Can a 510(k) cite more than one predicate device?

Yes. Sponsors commonly cite a primary predicate to establish intended use plus one or more reference devices to support specific technological characteristics (sometimes called a split predicate), or multiple independent predicates supporting the same comparison. The intended-use step of the decision tree still has to trace back to a genuine predicate, not just a collection of reference devices.

Can a device with a recall history be used as a predicate?

It’s legally possible if the device is still legally marketed, but it’s a weak choice, especially where the recall (particularly a Class I recall) is tied to the same technological characteristics the equivalence argument relies on. Checking a candidate predicate’s recall and adverse-event history before filing is a routine, low-cost step worth doing early.

What happens if FDA finds a device not substantially equivalent (NSE)?

An NSE determination places the device into Class III by operation of law. The sponsor’s remaining options are a PMA submission, a De Novo request if the device is actually low-to-moderate risk, or a reclassification petition.

What’s the difference between intended use and technological characteristics in the SE test?

Intended use is compared first and is a threshold question — if it differs from the predicate, FDA generally does not proceed to compare technological characteristics at all. Technological characteristics (design, materials, energy source, performance specifications) are compared second, and can differ from the predicate as long as the differences don’t raise new questions of safety or effectiveness that aren’t resolved by performance data.