A reference device isn't a predicate

A 510(k)’s substantial-equivalence argument is usually described in shorthand: device versus predicate, pick one cleared device close enough in intended use and technology and build the comparison table against it. FDA’s own 2014 guidance on evaluating substantial equivalence names a second role that shorthand skips past entirely — the reference device — and treats blending the two roles the wrong way as the one failure mode worth naming outright.

The predicate carries the whole claim

Substantial equivalence, as 21 U.S.C. 360c(i) defines it, is one finding about one relationship: the new device has the same intended use as the predicate, and either the same technological characteristics or different ones that raise no new questions of safety or effectiveness the performance data doesn’t answer. FDA’s guidance allows more than one predicate to carry that finding together — common when a new device combines features from two prior devices — but requires the filer to name a primary predicate, the one whose indications and technology are the closest match, anchoring the comparison the rest of the submission builds on.

A reference device proves one fact, not the relationship itself

A reference device plays a different role: a separate, legally marketed device cited not to establish substantial equivalence, but to support one technological characteristic or test method the new device shares with it. FDA’s own example is concrete — a knee implant carrying a new antimicrobial coating can point to a hip implant already cleared with that same coating to support the coating’s safety and performance, without the hip implant ever becoming the device the knee implant claims equivalence to. The reference device still has to be legally marketed, the same baseline a predicate has to clear, but it never substitutes for one, and a submission can cite more than one without turning any of them into a second predicate.

A reference device can support a technological characteristic. It cannot carry the finding that makes the submission a clearance.

The guidance bans only one combination

What the guidance forbids by name is the “split predicate” — one device cited for sameness of intended use, a second, different device cited for sameness of technological characteristics, with neither one actually carrying both halves of the comparison against the subject device. FDA had been rejecting that structure in review practice for years before 2014; the guidance made it an explicit, final policy rather than a reviewer-by-reviewer judgment call. Multiple predicates sharing one finding together are still allowed. A reference device supporting a single technological detail is still allowed. A predicate for intended use and an unrelated device standing in for the technology comparison is the one arrangement the guidance now names and refuses.

The practice

Reading a 510(k) strategy from the outside means reading past the shorthand, because the public record carries no field for either role: openFDA’s 510(k) endpoint returns a K-number, applicant, product code, decision and date — nothing that names what the filer compared to what, or which cited device played which part. The distinction exists only in the submission’s own free text, in whichever summary or statement the applicant filed under 807.92 or 807.93. Reading a competitor’s predicate strategy, or building your own, means opening that document and checking which role each cited device actually plays — not assuming a device named in the file carries the whole substantial-equivalence argument just because it appears in the same section. This is regulatory intelligence about a documented FDA distinction, never advice about whether a specific predicate or reference device would hold up for a specific submission.