Where do I actually see which predicate a cleared device used?

In its 510(k) Summary PDF, linked from the FDA record (and from each device page on this site). FDA does not publish predicates as a structured database field, so the summary document is the primary source. If the applicant filed a 510(k) Statement instead of a Summary, the predicate is not public and you would have to request the submission via FOIA.

Can I use a recalled device as a predicate?

Sometimes, but carefully. A recall does not automatically disqualify a predicate — the device must be legally marketed and must not have been removed from the market by FDA or found misbranded/adulterated in ways that defeat equivalence. A predicate withdrawn for safety reasons is off the table, and reviewers read recall history. Check the recall record and its reason before committing.

Does the predicate have to share my product code?

Usually it will, because the product code encodes intended use and technology type — but the code is a filing convention, not the legal test. The legal test is same intended use plus technological characteristics that do not raise new questions of safety and effectiveness. Devices under a neighboring code can occasionally serve; expect to justify the choice more heavily.

How many predicates should I cite?

One primary predicate is the cleanest submission. Multiple predicates are allowed and common when no single device covers every characteristic, but each added predicate expands the comparison table and the surface a reviewer can question. Reference devices — used to support specific performance points without carrying the equivalence claim — are the better tool for secondary support.

How recent should a predicate be?

There is no regulatory age limit, and old predicates are legally valid. Practically, recent predicates tend to make cleaner submissions: their summaries reflect current testing expectations and recognized standards, so your comparison inherits fewer gaps. FDA has repeatedly signaled a preference for modern predicates, even though no cutoff exists in law.

By Andrew Weston — Installation Systems Engineer

How to Find a Predicate Device

This is the working procedure, not the theory. If you want the concept of substantial equivalence and how predicate chains form, read Understanding Predicate Devices first. This guide assumes you have a device in development and need candidate predicates you can defend.

Step 1 — Write the intended use before you search

Predicate research fails most often at the start, by searching for devices that look like yours instead of devices used like yours. The equivalence test runs on intended use first. Write one sentence: what the device does, for which patient population, in which setting. Every candidate gets measured against that sentence, and a candidate that matches your technology but not your intended use is not a predicate — it is a trap that costs a submission cycle.

Step 2 — Find the product code

The three-letter product code is the index key for everything that follows. Two honest ways to find it:

From a known competitor. If any marketed device does roughly what yours does, look up its 510(k) record and take the product code from there. This is the fast path and the one practitioners actually use.
From the classification database. Search FDA's classification database by device type keywords, then confirm the regulation number and device class match your expectations.

Check the code's landscape before trusting it: a product-code page shows the device class, regulation, advisory panel, how many devices have cleared under it, and who files there. If the recent clearances under the code describe devices that are plainly not your category, you have the wrong code — a sibling code under the same regulation is usually the right one. Some device types legitimately straddle two codes; resolve that now, not after you have built a comparison table.

Step 3 — Build the candidate list from recent clearances

From the product-code page, work through recent clearances newest-first. For each plausible device, open its 510(k) record and pull the summary PDF. You are collecting three things per candidate: the stated intended use / indications for use, the technological description, and — critically — which predicate it cited. That third field is the shortcut: summaries name predicates, so three or four recent summaries under your code sketch the code's live predicate lineage, and the devices that appear repeatedly as predicates are the load-bearing members of that lineage. They are popular for a reason: reviewers have accepted them many times.

Expect friction with the documents themselves. Some summaries are image-only scans, older clearances may have no published summary at all, and a minority of applicants filed statements instead of summaries — those predicates are simply not public. Budget for reading PDFs; there is no structured predicate field in openFDA to script your way around.

Step 4 — Vet before you commit

For each shortlisted candidate, verify:

Legally marketed. The clearance exists and the device was actually brought to market. A cleared but never-marketed device is a weak foundation.
Intended use matches yours — the sentence from Step 1, prong by prong. Same disease, same population, same clinical context.
Technology is comparable, or your differences demonstrably raise no new questions of safety and effectiveness. Every difference becomes a row in your comparison table and possibly a test in your bench plan.
Regulatory history is clean. Check the recall record for the device and its manufacturer. A safety-driven market removal defeats a predicate; a lesser recall is context a reviewer will still weigh.

Step 5 — Choose, and document the losers

Pick the primary predicate with the cleanest intended-use match, not the newest gadget. Keep notes on the candidates you rejected and why — pre-submission meetings and reviewer questions routinely probe predicate choice, and "we considered K-numbers X and Y and rejected them because…" is a much stronger position than reconstructing the reasoning a year later. If genuinely no predicate exists, stop: you are looking at De Novo or PMA territory, and no amount of creative searching changes that.

Total honest effort for a well-behaved product code: a few hours. For a crowded or messy code, a few days of PDF reading. Either way it is cheap insurance against the expensive failure mode — a Not Substantially Equivalent decision built on a predicate that never fit.

Frequently asked questions

Where do I actually see which predicate a cleared device used?: In its 510(k) Summary PDF, linked from the FDA record (and from each device page on this site). FDA does not publish predicates as a structured database field, so the summary document is the primary source. If the applicant filed a 510(k) Statement instead of a Summary, the predicate is not public and you would have to request the submission via FOIA.
Can I use a recalled device as a predicate?: Sometimes, but carefully. A recall does not automatically disqualify a predicate — the device must be legally marketed and must not have been removed from the market by FDA or found misbranded/adulterated in ways that defeat equivalence. A predicate withdrawn for safety reasons is off the table, and reviewers read recall history. Check the recall record and its reason before committing.
Does the predicate have to share my product code?: Usually it will, because the product code encodes intended use and technology type — but the code is a filing convention, not the legal test. The legal test is same intended use plus technological characteristics that do not raise new questions of safety and effectiveness. Devices under a neighboring code can occasionally serve; expect to justify the choice more heavily.
How many predicates should I cite?: One primary predicate is the cleanest submission. Multiple predicates are allowed and common when no single device covers every characteristic, but each added predicate expands the comparison table and the surface a reviewer can question. Reference devices — used to support specific performance points without carrying the equivalence claim — are the better tool for secondary support.
How recent should a predicate be?: There is no regulatory age limit, and old predicates are legally valid. Practically, recent predicates tend to make cleaner submissions: their summaries reflect current testing expectations and recognized standards, so your comparison inherits fewer gaps. FDA has repeatedly signaled a preference for modern predicates, even though no cutoff exists in law.

Last updated 2026-08-04.