Regulatory Intelligence

De Novo: How the FDA Classifies Devices With No Precedent

A data pull of 480 FDA De Novo decisions shows grant volume roughly tripling since the early 2010s — and reveals how one novel-risk review seeds an entire downstream product category.

By Knitify Data Intelligence Team • Regulatory & Real-World Data Analytics

July 14, 2026 • 6 min read

Quick Takeaways

What's the difference between a De Novo request and a 510(k) clearance?

A 510(k) requires the sponsor to show their device is substantially equivalent to an already-legally-marketed predicate device. A De Novo request is for devices with no predicate at all — the FDA evaluates the device on its own risk profile and, if appropriate, creates a brand-new Class I or Class II regulatory category (and product code) for it, which future devices can then cite as a predicate via 510(k).

Why did De Novo grant volume jump in 2018 and again in 2023–24?

The FDA's own De Novo decision database shows grants rising from roughly 18–22 per year in the early 2010s to 44 in 2018, then settling at 47 in both 2023 and 2024. The underlying driver documented in FDA policy history is the agency's 2018 rule change streamlining the De Novo process (final rule effective 2021) combined with a surge in software- and AI-based devices, which frequently have no hardware predicate to point to.

Does a high De Novo count mean the FDA is approving riskier devices?

No — De Novo is explicitly reserved for low-to-moderate-risk devices (Class I or II). The pathway exists specifically so novel but not-high-risk devices avoid being defaulted into the Class III/PMA track that was originally the only option for anything without a predicate.

In 2018, the FDA granted 44 De Novo classifications — nearly as many as the pathway's entire first decade (1998–2007) combined, which totaled 45. By 2023 and 2024, annual grants had risen further, to 47 in each year, according to a full pull of the FDA's own De Novo database (480 decisions total, 1998–2026). Every single decision in that dataset is coded "granted" — the index captures only devices that made it through, which matters for how you read everything that follows.

The De Novo pathway exists because, before 1997, any device with no legally marketed predicate was automatically dropped into Class III — the same bucket as pacemakers and heart valves — regardless of actual risk. The FDA Modernization Act of 1997 created an escape hatch: a sponsor with a genuinely novel, low-to-moderate-risk device could ask the agency to evaluate it on its own merits and slot it into Class I or Class II, creating a brand-new regulatory category (and product code) that later entrants could then cite as a predicate via the ordinary 510(k) route. That second-order effect is the mechanism worth dwelling on: a single De Novo grant doesn't just clear one device, it seeds an entire product-code lineage. Every 510(k) "me-too" device in that category downstream owes its existence to one first mover absorbing the harder, un-precedented review.

Where the growth concentrates

Aggregating by review committee across all 480 records, three areas dominate: Microbiology (62 grants, largely infectious-disease diagnostics), Neurology (55, increasingly neuromodulation and AI-based software), and Gastroenterology/Urology (47). General & Plastic Surgery (44) and Cardiovascular (36) round out the top five. That mix tracks a broader shift the pathway was arguably built for but didn't originally anticipate: software- and AI-driven devices with no hardware predicate at all.

A concrete case: Neurovalens' Modius Spero (DEN250013), granted De Novo status in 2026 for PTSD, is a non-invasive vestibular nerve stimulation headset with no predecessor device to be substantially equivalent to. It followed the company's earlier De Novo grant for Modius Lean (DEN240076), a weight-management version of the same underlying stimulation platform — illustrating the seeding effect directly: one novel-risk review created a product code that the company's own follow-on device, and potentially competitors', can now reference.

De Novo volume roughly doubled from the early-2010s baseline (~18–22/yr) to 2018 and again settled at a higher plateau (~44–47/yr) in 2023–2024, per FDA's public De Novo decision database (n=480).

Caveats

This index only records granted requests — declined or withdrawn De Novo petitions aren't in scope, so the true submission volume (and rejection rate) is invisible here, and the true year-over-year growth rate could be inflated if the agency is also declining more requests. Recent years (2025–2026) are partial or reflect an unusually current data pull relative to typical publication lag, so the last bar should be read as directional, not final. And "De Novo" itself is a heterogeneous bucket: a single-use surgical clip and an AI diagnostic algorithm both count as one grant, so raw counts don't distinguish incremental novelty from category-defining novelty.

Why this matters: De Novo volume is one of the few public, real-time proxies for where genuinely new device categories — not incremental 510(k) variants — are being created. For companies scouting licensing or in-licensing targets, or investors trying to spot a nascent product-code before it fills with competitors, tracking De Novo grants by committee is an earlier signal than tracking 510(k) volume, which mostly measures how fast a category that already exists is being copied.

References

FDA — De Novo Classification Request (pathway overview) View Source →
FDA — Evaluation of Automatic Class III Designation (De Novo) Summaries View Source →
Beyond the 510(k): The regulation of novel moderate-risk medical devices... in the FDA's De Novo pathway (PMC) View Source →
Neurovalens — FDA Grants World's First Neuromodulation Device to Treat Symptoms of PTSD View Source →
MassDevice — Neurovalens earns FDA nod for neuromod for PTSD View Source →

Tags:

Regulatory Intelligence FDA Medical Devices De Novo Pathway Device Classification Digital Health

← Back to Blog