510(k) predicate search: how to find and defend a predicate device
A practical method for 510(k) predicate search: what makes a valid predicate, how to use K-numbers and product codes, and how to document substantial equivalence.
A 510(k) stands or falls on the predicate. Substantial equivalence is a comparison, so the device you compare against decides how much evidence you need and how likely the submission is to survive review.
What makes a predicate valid
A predicate is a legally marketed device: previously cleared through 510(k), pre-amendments, reclassified into Class II, or granted through De Novo. It must have the same intended use as your device. Technological characteristics may differ, but different characteristics must not raise different questions of safety and effectiveness — and where they do differ, performance data has to show the device is as safe and effective as the predicate.
A device subject to a design-related recall, or one that has been withdrawn, is a poor choice even when it is technically eligible.
Running the search
Start from the product code you selected during classification, not from a company you admire. Pull clearances under that code, read the indications statements rather than the marketing names, and keep the ones whose intended use language you could match without stretching. Note the K-number, decision date, and any special controls or guidance cited.
Recent clearances are usually better than old ones: the review expectations they reflect are closer to what you will face. Where no single device matches, teams sometimes cite a primary predicate plus a reference device for a specific technological feature — that is acceptable, but the primary predicate still has to carry the intended use.
Documenting substantial equivalence
Build a side-by-side table: intended use, indications, patient population, user, environment of use, energy source or algorithm, materials, performance specifications. For every difference, state the question it raises and the data that answers it. Reviewers read the differences column first.
Software and AI-enabled devices
For Software as a Medical Device, predicate selection is often harder because indications are narrow and evolve quickly. Expect to address the software documentation level, cybersecurity, and — for adaptive algorithms — how change will be managed after clearance.
How DevicePath helps
DevicePath returns real 510(k) predicate candidates with K-numbers alongside the likely class, candidate product codes and probable pathway, so a predicate shortlist takes minutes rather than days. It is open source, and intended as the first pass before regulatory counsel refines the strategy.