Researching device predicate references means working backward from your device's product code and forward through FDA databases until you can justify each comparison claim. The predicate is the legal anchor for a 510(k), but a usable reference file also includes the cleared indications, technology descriptions, and any warning signs that could weaken substantial equivalence. FDA's general advice is to begin with classification, then search the 510(k) database one field at a time [1][2]. The harder work comes after: distinguishing valid predicates from weak ones, separating true predicates from reference devices, and documenting why the selected devices remain defensible.
Start from product code rather than trade names
FDA identifies the device classification and product code as the essential starting point for predicate research. The Product Code Classification Database supplies the classification panel, common name, product code, and applicable CFR regulation for a device type when final classification exists [1]. For companies preparing a first 510(k), Emergo by UL recommends entering that product code in the 510(k) database search field and leaving other fields blank as the initial step [2]. This approach avoids the common failure of searching by a trade name that may not appear in the database.
After product code, useful search dimensions include the names and manufacturers of similar devices, marketing status, 510(k) numbers, and classification details [1]. FDA cautions that trying several search strategies is usually necessary, and that completing only one box at a time tends to work best [1]. Hyphens or spaces in manufacturer names can change results, so alternate name formats may be required [1]. The 510(k) database also preserves original application information only, meaning the name of the original applicant and original trade name may differ from the current market identity [1].
Confirm legal marketing status and intended use
A predicate must be a legally marketed device that is not subject to PMA. FDA's 510(k) page states that a legally marketed device can be a preamendments device, a device reclassified from Class III to Class II or I, a device found substantially equivalent through the 510(k) process, or a device granted marketing authorization through De Novo classification [5]. The predicate cannot be one that is or was in violation of the FD&C Act, even if it is no longer marketed in the U.S. [1]. This distinction matters because a recently cleared 510(k) device is usually used, but not required [1].
Intended use is the boundary condition that separates a valid comparison from a new device. For preamendments devices, a firm must demonstrate that the device was labeled, promoted, and distributed in interstate commerce for a specific intended use and that the intended use has not changed [1]. If the device is labeled for a new intended use, FDA considers it a new device requiring a 510(k) [1]. The same principle applies when evaluating candidate predicates: differences in indications under the same product code cannot simply be ignored, as illustrated by FDA's monitoring system examples [2].
Documenting the predicate decision
Keep a working table that lists each candidate's K number, manufacturer, product code, clearance date, intended use, and the source where each item was verified. For postamendments devices, the 510(k) number links directly to the database record [1]. For reference devices, record the specific scientific role, such as supplying standard reference values or methodology support after Decision Point 4 [2][3]. A reference device is a legally marketed device intended to supply scientific or technical information, such as test methodology, to help address the safety and effectiveness of a new technological characteristic [2]. A comment from DiMe states that without a singular database to reference, manufacturers must navigate a patchwork of information and may need to piece together insights from multiple sources, so noting the source and date of every data point prevents later confusion about what was verified where [8].
Separate reference devices from predicates
A reference device is not a predicate and cannot resolve the first four decision points on the 510(k) Decision-Making Flowchart. FDA's draft best practices guidance states that reference devices can support a submission beyond Decision Point 4, for example by supplying scientific methodology or standard reference values at Decision Point 5a [3]. Emergo by UL makes the same distinction in its white paper, explaining that reference devices cannot be used to address decision points 1 through 4 and may only appear after those points [2]. This means a research file should label each device by its intended role.
When a reference device is used, FDA expects it to be identified in the appropriate submission sections, including the 510(k) Summary [2]. The supporting role can involve standard reference values or test methodology that does not establish the predicate relationship itself. Researchers should therefore ask a specific question for every candidate: does this device serve as the legal comparator, or does it merely support a test method or reference range? Conflating the two roles can create an unsupported substantial equivalence argument.
Screen candidates for methods, safety, and recall history
The 2023 draft guidance proposes best practice criteria that go beyond legal market status. FDA recommends prioritizing predicate devices with methods developed within a consensus environment, and those subject to public comment or peer review [3]. Submitters should consider how much information is available about the test methods supporting the predicate's clearance and whether those methods remain appropriate for the subject device [3]. This shifts research from simply finding a cleared device to evaluating the evidence quality behind that clearance.
Safety screening is explicitly recommended once a list of valid predicates exists. FDA advises searching the MAUDE database, the Medical Device Reporting database, and the MedSun Reports database for reported injuries, deaths, or malfunctions associated with candidate predicates [3]. The Digital Medicine Society's comment to the docket reiterates the four draft criteria: cleared using well-established methods, meeting or exceeding expected safety and performance, lacking unmitigated use-related or design-related safety issues, and having no recall due to design, manufacturing, or labeling defects [8]. That organization also argues that existing public resources are insufficient for manufacturers to achieve these practices [8].
A suggested workflow for a first-time evaluation
As a demonstration, suppose a regulatory team is preparing a 510(k) for a continuous patient monitoring accessory. The team could begin by locating the product code through the Product Code Classification Database and recording the classification regulation and common name [1]. The next step would be to run the product code alone in the 510(k) database, then repeat with likely manufacturer names, and save the K numbers and clearance dates [1][2]. This first list should be treated as candidates, not yet as predicates.
The team would then check each candidate against the legal marketing categories and the intended use statement in its 510(k) summary or clearance record [1][5]. The 510(k) database contains original application information only, so the original applicant and original trade name may differ from the current market identity [1]. Any device with an unclear or mismatched indication should be set aside. For the remaining candidates, the team could run MAUDE, MDR, and MedSun searches and note whether any reports indicate recurring design-related failures [3]. The final predicate would be the candidate with the closest intended use and the cleanest method and safety record, while clearly different devices supported only by test references would be designated as reference devices, not predicates [2][3]. The team would also record whether any candidate has a new intended use or new technological characteristics that differ from other devices cleared under the same classification product code, because such a device may be a candidate for a De Novo Classification Request [2].
Sources & references
- How to Find and Effectively Use Predicate Devices | FDA · www.fda.gov
- U.S. FDA Medical Device Predicate Selection for 510(k) Submissions · www.emergobyul.com
- Best Practices for Selecting a Predicate Device to Support a Premarket Notification [510(k)] Submission - Draft Guidance for Industry and Food and Drug Administration Staff · www.fda.gov
- Best Practices for Selecting a Predicate Device to Support a Premarket Notification [510(k)] Submission | FDA · www.fda.gov
- Premarket Notification 510(k) · U.S. Food and Drug Administration
- FDA Continues to Take Steps to Strengthen the Premarket Notification [510(k)] Program - Program Updates | FDA · www.fda.gov
- Identification of predicate creep under the 510(k) process: A case study of a robotic surgical device | PLOS One · journals.plos.org
- 90 Canal St., 4th Floor, Boston, MA 02214 | +1-765-234-3463 | www.dimesociety.org December 6, 2023 Dockets Management Staff Food and · downloads.regulations.gov