The FDA De Novo Pathway: How a Genuinely New Device Finds a Route to Market

The short answer

The De Novo pathway is how the FDA classifies a medical device that is genuinely new, poses low to moderate risk, and has no earlier device it can point to as a match. Rather than forcing that device through the most demanding review reserved for the highest-risk products, De Novo lets the agency create a fresh classification for it, set safeguards that fit its actual risk, and grant marketing authorization. A successful request does something quietly important: it establishes a category that later devices can use as a predicate. This article is educational and not legal advice.

Three doors, not one

Most people who follow medical devices have heard of two routes to the U.S. market. The first, 510(k), rests on a concept called substantial equivalence. A company shows that its device is comparable in intended use and technology to a device already legally sold, the predicate, and the review asks whether the new product is at least as safe and effective as that established comparison. The second, premarket approval (PMA), is the demanding route for high-risk devices, the ones that support or sustain life or carry significant potential for harm. PMA asks a company to demonstrate safety and effectiveness on its own evidence, often including clinical data, rather than by comparison.

De Novo is the third door, and it exists because the first two leave a real gap. Imagine a device that is low or moderate in risk but first of its kind. There is nothing legally marketed that resembles it, so 510(k) has no predicate to lean on. Yet the device is nowhere near the risk level that PMA is built for. Before this pathway existed, such a device fell into an awkward trap.

Why the pathway was created

Under the original statutory logic, a novel device with no predicate was automatically slotted into Class III, the highest-risk class, by default. That default had nothing to do with the device being dangerous. It reflected the fact that the classification machinery had no obvious shelf to place it on. The result was that a modest, sensible, low-risk innovation could face the same PMA burden as an implantable life-support device.

The Food and Drug Administration Modernization Act of 1997 introduced De Novo to fix exactly this. It gave the agency a way to look at a novel device on its own terms and ask a better question: given what this device actually does, do general controls, or general controls plus a tailored set of special controls, provide reasonable assurance of safety and effectiveness? If the answer is yes, the device can be classified as Class I or Class II instead of being stranded in Class III.

A later amendment, the Food and Drug Administration Safety and Innovation Act of 2012, streamlined the process further. Originally a company had to submit a 510(k), receive a finding that its device was not substantially equivalent to anything, and only then request De Novo classification. The 2012 change allowed a direct De Novo request without that preliminary detour, so a sponsor who already knows no predicate exists can go straight to the pathway built for that situation.

What De Novo actually establishes

Granting a De Novo request does more than clear one product. The FDA writes a classification order that defines a new device type, assigns it to Class I or Class II, and, for Class II, specifies special controls. Special controls are the tailored requirements the agency judges necessary to manage the specific risks of that device type. They can include performance testing standards, labeling requirements, software validation expectations, cybersecurity measures, biocompatibility criteria, or particular clinical or usability data. General controls, which apply broadly to devices, cover basics such as registration, quality systems, and reporting of adverse events.

The elegant part is what happens next. Once that classification exists, it becomes part of the regulatory map. A future company with a similar device can cite the De Novo device as a predicate and pursue a 510(k). The pathway that had no predicate has now created one. A single first-of-kind clearance can open an entire category, letting successors follow a lighter route while meeting the same special controls that defined the original risk profile.

Why this matters for first-of-kind AI tools

Software as a Medical Device, and especially software that uses machine learning, keeps producing devices with no honest predicate. When an algorithm does something no cleared device has done before, claiming substantial equivalence to an older product can be a poor fit. De Novo gives these tools a route that matches their novelty without treating them as if they were high-risk implants.

I hold a certificate in medical device regulation from KTH covering EU MDR, IVDR, the FDA framework, and Software as a Medical Device, and I completed FDA clinical investigator training, so I read these pathways as an explainer rather than as a regulator or a lawyer. What strikes me about De Novo in the AI context is how the special controls mechanism carries the real weight. For a novel algorithm, the special controls can name the very things that make such software trustworthy: the datasets used to establish performance, reporting on how the model behaves across relevant subgroups, human oversight expectations, and transparency in labeling. In recent years the agency has also worked to let device software functions that use AI describe planned modifications in advance through a predetermined change control plan, which acknowledges that a learning system may be refined over time. That mechanism can attach across pathways, De Novo among them.

A short mental model

Think of the three pathways as answers to two questions. First, how risky is the device? Second, is there something already on the market it genuinely resembles? High risk points toward PMA. Low to moderate risk with a real predicate points toward 510(k). A low to moderate risk device with no honest predicate is the space De Novo was built for. The statute gives the agency a defined window to reach a classification decision, and a granted request produces two results at once: authorization for this device, and a new category that later devices can build on.

None of this is a shortcut around evidence. A De Novo request still has to show that the chosen controls provide reasonable assurance of safety and effectiveness, and the evidence expected can be substantial. What the pathway offers is proportion: a way to hold a genuinely new, lower-risk device to requirements that fit what it is, while building a durable category that the next innovator can use.