510k vs PMA Difference

De Novo sits between the 510(k) and the PMA. It is for a novel device that has no predicate, so the law automatically places it in Class III, but whose risk is low or moderate enough that general controls (or general plus special controls) can assure safety and effectiveness. If the request is granted, the device goes into Class I or II instead of needing a PMA, and it becomes a predicate that later devices can cite in their 510(k)s.

Approach: I worked from the FDA's 2021 De Novo Classification Process guidance and the 2022 guidance on FDA and industry actions on De Novo requests. I didn't open 21 CFR Part 860 Subpart D or the FD&C Act itself.

Why it exists

• A Class III device can come about in two ways. The FDA can put a device type into Class III by regulation or order because of its risk; those devices can't use De Novo ^[9]. Or a new type of device is automatically Class III under section 513(f)(1) simply because it has no predicate, "regardless of the level of risk they pose" ^[9].
• Congress created section 513(f)(2) in FDAMA (1997) for the second case. The aim was to avoid requiring a PMA for devices where general or special controls are enough ^[9].

Two ways in

1. After a 510(k) found the device not substantially equivalent (NSE). This was originally the only route ^[9]. The NSE has to be due to:

   • no identifiable predicate,
   • a new intended use, or
   • different technology that raises different questions of safety and effectiveness.

   An NSE based only on inadequate performance data generally doesn't qualify ^[10]. The 21st Century Cures Act removed the old rule that the request had to be filed within 30 days of the NSE ^[9].

2. "Direct De Novo" (added by FDASIA section 607). If you conclude there is no legally marketed predicate, you can file without a 510(k) first ^[9][11].

When De Novo isn't available

• If a likely predicate exists, or the device falls under a Class III classification regulation, the request is declined ^[12].
• If the device type already has a classification regulation or approved PMAs, the way to get it into Class I or II is reclassification under section 513(e) or 513(f)(3) ^[10].

How it compares with the other two routes

	510(k)	De Novo	PMA
Basis	Substantial equivalence to a predicate	Classification under section 513(a)(1)(A) or (B) [11]	Reasonable assurance of safety and effectiveness, Class III
Evidence standard	Same as the predicate	Like a PMA, it requires a demonstration of reasonable assurance of safety and effectiveness [13]. The request must show that benefit outweighs risk under general or special controls, and must lay out each risk and its mitigation [14][10]	Full showing of safety and effectiveness
Outcome	"Cleared"	"Granted" [15]	"Approved"
Effect	Device can be marketed	Device can be marketed, a new classification regulation is created, and the device can serve as a predicate [11]	Approval for that device only

Process and timelines

• Pre-Submission: not required, but "strongly recommended." It lets the FDA give early feedback on eligibility, whether a predicate exists, the likely special controls and the data needed ^[10].
• Acceptance review: the FDA decides within 15 days whether to accept the request. If it doesn't decide in time, the request counts as accepted ^[12].
• Additional information: if the FDA asks for more information, you have 180 days to answer completely, or the request is treated as withdrawn ^[12].
• Statutory deadline: the classification order is due within 120 days of receipt ^[11].
• MDUFA V goal: a decision (grant, decline or withdraw) within 150 FDA Days for 70% of requests in FY 2023-2025. In FY 2026-2027 the target can rise to 80% or 90% ^[16].
• If granted: you can market the device immediately under the order. The FDA publishes a final order in the Federal Register within 30 days, and the classification is codified in 21 CFR Parts 862-892 ^[15].
• If declined: the device stays in Class III and can't be marketed. You can either file a PMA or file a new De Novo request with more data ^[15].
• Competing requests: if two De Novo requests for the same device type are under review and the first is granted, the second is declined. That sponsor can then refile as a 510(k), referencing its earlier submission ^[15].

In practice, if you have a predicate, use a 510(k). If you have a novel device whose risks you can control without a PMA, use De Novo. If the device type is already Class III by regulation, or controls can't manage its risks, you need a PMA.

Pick your predicate in two passes. First, list every legally marketed device that can serve as a predicate: one with the same intended use as yours, where any differences in technology don't raise different questions of safety and effectiveness. Then narrow that list to the best one, using the four best practices in the FDA's September 2023 draft guidance, and explain your choice in the submission ^[17].

Approach: I used the FDA's 2014 510(k) Program guidance (final) and the 2023 draft predicate-selection guidance, and confirmed on the FDA guidance page that it is still a draft (docket FDA-2023-D-3134, content current as of 09/07/2023). Its recommendations are non-binding and "not for implementation."

Step 1: Is the device eligible to be a predicate at all?

• It must be legally marketed. That means one of:

  • a device on the market before May 28, 1976;
  • a device reclassified from Class III to Class II or I; or
  • a device found substantially equivalent through a 510(k) ^[18].

  Any legally marketed Class I or II device qualifies, not only recent clearances ^[19].

• Some devices are excluded. You can't use a device that FDA removed from the market, or one a court found misbranded or adulterated (section 513(i)(2)) ^[18].

• It must have the same intended use as your device. The technology must also be the same, or different without raising different questions of safety and effectiveness ^[20].

• "Different technological characteristics" means a significant change in materials, design, energy source or other features (section 513(i)(1)(B)) ^[18].

• Watch for changes in indications. These are more likely to create a new intended use:

  • moving from a functional indication to a treatment or aesthetic one;
  • moving between diagnostic and screening;
  • a different anatomical site;
  • a different patient population;
  • a different setting, such as hospital versus home use ^[21].

Where to look: start in FDA's 510(k) Premarket Notification Database. Search by trade name, manufacturer, 510(k) number, product code or classification regulation, then read each candidate's 510(k) Summary and Indications for Use ^[22][23].

Step 2: Narrow the list with the four best practices

Best practice	What to check	Example from the FDA
A. Cleared using well-established methods	Was it tested under FDA-recognised consensus standards, FDA guidance, a qualified MDDT, or published, accepted methods? Prefer methods developed by consensus or peer review [23]	Between two similar candidates, pick the one tested to FDA guidance or recognised standards [24]
B. Meets or exceeds expected safety and performance	Search MAUDE, the MDR database and MedSun for injuries, deaths and malfunctions [24]	Frequent battery failures causing serious injury on a predicate infusion pump suggest a design problem, so choose another predicate [24]
C. No unmitigated use-related or design-related safety issues	Check emerging signals and safety communications on FDA's Medical Device Safety pages and CBER Safety & Availability pages [25]	The reprocessing signal for duodenoscopes [25]
D. No design-related recall	Search the Medical Device Recalls Database, which covers recalls classified since November 2002. A design-related recall can point to a fundamental design flaw [26]	A coronary catheter tip fracture caused by the material not meeting user needs [26][27]

Age isn't the test. The FDA moved away from its earlier idea of targeting predicates more than ten years old:

• For implants, older devices can bring long-term safety data.
• For software devices, a recent predicate may include modern cybersecurity and interoperability features ^[28][29].
• In the FDA's intervertebral fusion device example, the 15-year-old predicate is chosen over a 3-year-old one because of its longer safety record ^[30].

If no candidate passes all four checks:

• You can still use the best one available. Explain how your design or testing mitigates its known concerns ^[25].
• State in the 510(k) Summary that no predicate meeting the best practices was available, and describe the related testing in the Performance Data section ^[27][31].
• The FDA's bone sonometer example does exactly this with a single predicate that had a design-related recall ^[32][30].

Step 3: Decide how many predicates you need (2014 final guidance)

• A single predicate is best. If you cite more than one, name a primary predicate: the one whose indications and technology are most similar to your device ^[20].

• Multiple predicates are acceptable in some cases:

  • combining features from predicates that share the same intended use;
  • seeking more than one intended use;
  • seeking more than one indication under the same intended use.

  Every predicate you cite must have the same intended use as your device ^[20][6].

• Keep the number down. Limit predicates to the minimum needed and explain why each one is necessary ^[19].

• Added features. You can add a convenience feature, but the device must still meet the substantial-equivalence standard against at least one predicate from the same classification regulation ^[33].

• No "split predicates." You can't take the intended use from one device and the technology from another device with a different intended use; this is inconsistent with the 510(k) standard. FDA must be able to work through Decision Points 1-4 of the flowchart using one predicate ^[20].

• Reference devices are not predicates. They can only support test methods or standard reference values at Decision Point 5a, after you have passed Decision Point 4 with a real predicate, and you need a scientific rationale for each ^[33].

Step 4: Document the choice

• In the submission: show how you applied the best practices ^[22].
• In the 510(k) Summary: include a narrative explaining why you chose the predicate ^[27].
• Comparison table: the FDA's coronary guidewire example rates four candidates against criteria A-D and picks the only one that is clean on all four ^[31][32].
• Citations: every predicate relied on must be cited accurately in the 510(k) Summary under 21 CFR 807.92(a)(3) ^[19].
• The FDA decides in the end whether your proposed predicate is valid ^[17].

If you can't find a predicate with the same intended use, you're in De Novo territory, or PMA territory for high-risk devices. A Pre-Submission is encouraged to confirm with the FDA whether a valid predicate exists ^[19].