De Novo vs 510(k) for AI Medical Devices — Which Pathway?

The decision between De Novo and 510(k) is one of the most consequential regulatory strategy choices an AI device company makes. Get it wrong and you either spend 18 months on a De Novo you didn't need, or you submit a 510(k) that gets rejected as not substantially equivalent — losing a year to a problem that was foreseeable.

This framework is built from analysis of AI device clearances across both pathways.

The Core Difference

510(k) requires substantial equivalence to a predicate device. If you can find a cleared device with the same intended use and either the same technology or different technology that doesn't raise new safety questions, you have a path.

De Novo is for devices that are novel — no valid predicate exists — but present low to moderate risk. FDA evaluates the device on its own merits, establishes special controls if appropriate, and creates a new product code. Future devices in the same category can then use your De Novo device as a predicate for 510(k).

When 510(k) Is Right for AI Devices

510(k) is appropriate when a valid predicate exists in your product code. For most AI device indications, a predicate exists because the category has been opened by a prior De Novo. The relevant question is whether your device is substantially equivalent to cleared predicates.

Strong indicators that 510(k) is the right path:

• A product code exists for your indication (QMF, OZO, QFP, etc.)
• Devices have been cleared under that product code in the last 3 years
• Your intended use substantially overlaps with cleared predicates
• Your algorithm type (CNN, transformer, etc.) has been used in cleared devices
• You can meet the special controls established by the governing De Novo

When De Novo Is Right for AI Devices

De Novo is appropriate when no valid predicate exists. For AI devices, this typically means you're addressing an indication where no AI device has been cleared, or your device is so technologically distinct that existing predicates don't constitute valid comparators.

Strong indicators that De Novo may be required:

• No product code exists for your specific indication
• Your device is autonomous — makes diagnostic decisions without clinician interpretation
• Your device uses modalities or inputs that haven't been cleared in AI devices
• FDA has issued a Not Substantially Equivalent determination for similar devices trying to use 510(k)
• Your intended use is genuinely novel with no analogous cleared device

The Strategic Calculus

Beyond the technical question of which pathway you're eligible for, there's a strategic question about which is better for your business.

De Novo has a significant strategic upside: if granted, you establish the product code and special controls for your category. Future competitors must either use your device as their predicate (benefiting your position as first-mover) or pursue their own De Novo. You shape the regulatory expectations for your indication.

The costs: De Novo takes 18-24 months versus 9-14 months for 510(k). It requires more extensive clinical data. It requires FDA to agree on the special controls — a negotiation that can be time-consuming. And there's no guarantee FDA grants it; if FDA determines higher risk than expected, you could end up referred to PMA.

The Hybrid Strategy

Some AI device companies pursue both pathways simultaneously or sequentially. They submit a 510(k) for a more limited version of their intended use where a predicate exists, clear quickly, and generate real-world evidence. They then use that real-world evidence to support a De Novo for the broader, more novel intended use they ultimately want.

This strategy reduces time to first revenue while building toward the regulatory position they want long-term. It works when the 510(k)-clearable version is a genuine, commercial product — not a stripped-down regulatory placeholder.

The Q-Sub Decision

Before committing to either pathway, request a pre-submission meeting (Q-sub) with FDA, particularly for De Novo or for 510(k)s in novel indications. FDA's feedback in a Q-sub can tell you definitively whether they'll accept your predicate strategy or whether De Novo is required.

The Q-sub costs 2-3 months. The alternative — submitting and getting an NSE determination 9 months later — costs much more.