The De Novo Pathway: First-in-Class Devices

When no predicate exists for your novel algorithm. The De Novo process, strategic advantages of creating a new classification, and how it differs from 510(k).


The De Novo pathway exists for devices that don’t fit into existing FDA classifications because they’re genuinely novel, or because no suitable predicate device exists for the 510(k) pathway. It’s less commonly used than 510(k), but it carries a hidden strategic advantage: when you go through De Novo, you’re not just getting approval for your device. You’re creating a new device classification that becomes the predicate for every competitor who comes after you.

For AI devices, this is a big deal. If you have a truly novel algorithm (an autonomous diabetic retinopathy screener, a first-of-its-kind pathology tissue classifier, a unique natural language processing tool for clinical notes), De Novo might be your path. It takes longer, costs more, and requires more evidence than a 510(k), but if you’re the first, the strategic advantage of setting the classification standard can be worth it.

De Novo vs. 510(k): The Key Differences

The 510(k) pathway asks the FDA, “Is my device equivalent to something you’ve already approved?” De Novo asks a different question: “Can you classify my device and set the standards for what devices like mine need to demonstrate?”

When the FDA clears a 510(k), you get marketing approval based on equivalence. You’re done. When the FDA clears a De Novo, two things happen simultaneously: your device gets approved, and a new device classification is established. That classification automatically becomes the predicate for all future 510(k) submissions in that category. This is why the first-mover advantage is real.

Consider IDx-DR, a famous AI example. It was the first autonomous diabetic retinopathy screening system approved by the FDA, cleared via De Novo in 2018. The FDA established a new classification for autonomous AI-based diabetic retinopathy detection. Now, any company building a similar system can use IDx-DR as their predicate and go through 510(k), which is faster and cheaper. IDx got the hard part out of the way. Everyone else benefits from the roadmap they created.

When De Novo Makes Sense

You should consider De Novo if:

No suitable predicate exists: You’ve searched the FDA 510(k) database thoroughly and either found nothing similar or found devices that are so different in intended use or technology that claiming equivalence would be a real stretch. Maybe you’re doing something truly novel, or maybe existing predicates are proprietary/protected in ways that make them hard to reference.

Your intended use is novel: You’re solving a clinical problem in a way that hasn’t been done before, or you’re applying an algorithm to a new clinical context. For example, if you’re building the first AI system to autonomously grade histopathology slides end-to-end (from slide image to diagnosis), and that doesn’t fit existing pathology device classifications, De Novo is likely in your future.

You have first-mover advantage: You want to establish the standard for your category. This is strategic. If you know competitors are coming and you want to set the bar for what your type of device needs to demonstrate, going first with De Novo is a move to consider.

The 510(k) predicate question is murky: You’ve done a Q-Submission and the FDA said, “We’re not sure a 510(k) predicate applies here. Consider De Novo.” This is actually good feedback. It means you’re in novel territory, and the FDA is steering you toward the pathway that fits.

The De Novo Process

A De Novo submission follows a similar structure to a 510(k), with one major addition: you’re also proposing a classification for your device and suggesting what special controls should apply to future devices in that class.

Here’s the flow:

  1. Pre-Submission Q-Meeting: You are going to do a pre-submission Q-Meeting, aren’t you? Before you invest in a full De Novo, do a Q-Submission asking the FDA if De Novo is the right pathway and what they’d expect to see. This clarifies the path and de-risks your strategy.

  2. Classification Proposal: You propose a device classification. Will this be a Class I (general controls only), Class II (general controls plus special controls), or Class III device (PMA required)? For most novel AI devices, sponsors propose Class II. You’ll have to explain your rationale.

  3. Special Controls Proposal: For Class II De Novo submissions, you propose what special controls should apply. Special controls are device-specific regulatory requirements. For AI systems, these might include: performance testing across subpopulations, continuous monitoring for model drift, cybersecurity validation, or specific documentation of the algorithm’s limitations. The FDA works with you to refine these.

  4. Full Submission Package: You submit substantial evidence supporting your proposed classification and special controls. This is more comprehensive than a typical 510(k) in some ways (you need to really justify the classification), and similar in others (you still need analytical validation, clinical validation, software documentation, and risk analysis).

  5. FDA Review: The FDA reviews your submission and decides whether to establish the classification. If approved, they issue an order establishing the new classification, and your device is cleared to market. If there are issues, they send a deficiency letter or request a meeting to discuss.

  6. Public Record: Unlike 510(k) submissions, the fact that a De Novo was submitted and approved is published in the Federal Register, so the public and competitors know a new device classification exists.

The Regulations and Special Controls

When the FDA establishes a new classification via De Novo, they’re also establishing what “special controls” will apply to that device type going forward. Special controls are the guardrails that keep devices in that class safe and effective. For an AI system, examples of special controls might be:

  • Performance validation: You must validate your algorithm on datasets representative of the intended patient population and clinical setting.
  • Software documentation: You must provide detailed documentation of your software development process, validation, and risk management.
  • Model monitoring: You must have a plan to monitor your model’s performance after deployment and detect degradation over time.
  • Cybersecurity: You must follow specific cybersecurity standards (like NIST) and document your approach.
  • Training data documentation: You must document the sources, composition, and potential biases in your training data.
  • Limitations statement: Your labeling must clearly state the limitations, populations the algorithm was tested on, and use cases it was not designed for.

These special controls become the floor for future competitors. Anyone else building a device in this new classification will need to meet these standards. That’s why the special controls negotiation during your De Novo is important. You’re shaping the regulatory standard for your entire market segment.

Timeline and Cost

De Novo submissions typically take 6-12 months from submission to clearance. This is longer than 510(k) because the FDA is doing more work (establishing a new classification, not just comparing to an existing one). The actual FDA review period is typically 6 months, but you often get deficiency letters and need additional meetings, which extends the timeline.

Preparation time is similar to a complex 510(k): expect 6-12 months of front-end work doing validation studies, assembling the submission, and preparing for meetings with the FDA.

Cost: The FDA user fee for a De Novo is approximately $40,000 for small businesses, higher for larger companies. The total out-of-pocket cost for a De Novo submission with external support is typically $200K-$500K, higher than 510(k) because you’re doing more work (more comprehensive validation, more extensive documentation, more FDA meetings).

Strategic Advantage: You Set the Predicate

Here’s where it gets interesting strategically. Once your De Novo is approved and the classification is established, you’ve created the predicate for everyone who comes after you. If a competitor wants to build a similar device, they can reference your device (the one you just got approved) and go through 510(k) instead of De Novo. This puts you first to market with a big head start.

Moreover, the special controls you negotiated become the regulatory bar for the entire category. If you negotiated thoughtfully and proposed controls that you can easily meet but your competitors might struggle with, you’ve created a moat. This is why De Novo strategy is really important. You’re not just getting approval for one device, you’re shaping the market.

Real example: The IDx-DR device created a classification for “autonomous AI-based diabetic retinopathy detection.” The special controls they agreed to include performance validation across different patient populations and different imaging equipment. These are things IDx could already do (they’d done it as part of their development). But if a smaller competitor wants to compete, they also need to meet these controls. This doesn’t guarantee IDx’s success, but it raises the bar for everyone else.

When You Propose De Novo (And It Gets Rejected)

Occasionally the FDA looks at your De Novo proposal and says, “Actually, we think this device fits into an existing classification. We can’t establish a new one, but a 510(k) to this device would be appropriate.” This is called a “De Novo Denial” (confusing terminology, but that’s the FDA for you). It’s not a rejection of your device; it’s a reclassification. You then pivot to 510(k) using the predicate the FDA suggested.

This isn’t the worst outcome. You’ve gotten FDA guidance on what predicate to use, and you go through 510(k) from there. It takes longer than if you’d done Q-Submission to nail down the pathway upfront, but your device still gets approved.

After Approval: Maintaining Your Classification

Once your De Novo is approved and your classification is established, you have a responsibility to maintain the integrity of that classification. If you discover that your device doesn’t meet the special controls you promised, or if you make substantial modifications to your algorithm, you might need to go back to the FDA for a new submission (a supplement). The FDA may also revisit the classification if they see problems in the market.

The point: you don’t get approved and then forget about regulation. You maintain your device, monitor its performance, and stay compliant with the special controls you agreed to. When competitors come in via 510(k), they’re comparing to your device, so your performance and safety track record matter.


Key Takeaways

  • De Novo is for novel devices where no suitable predicate exists for 510(k)
  • De Novo creates a new device classification that becomes the predicate for future 510(k)s (first-mover advantage)
  • You propose a classification (usually Class II) and special controls that will apply to your device and all future devices in that category
  • Timeline: 6-12 months submission to clearance, plus 6-12 months prep work
  • Cost: ~$35K user fee, plus $100K-$250K total for submission and validation
  • De Novo is strategically valuable if you’re truly first-to-market; you get to shape the regulatory standard for your category
  • Use Q-Submission before committing to De Novo to confirm it’s the right pathway
  • The special controls you negotiate become the regulatory bar for your entire market segment

This article is part of the AI in Clinical Research Knowledge Base. For a glossary of regulatory terms, see Appendix A.