De Novo trades the predicate for a benefit-risk case

No predicate table to fill in

A traditional 510(k) is built around a comparison: intended use, technological characteristics, and performance data laid out against a predicate device already on the market, closing with the claim that any differences don't raise new questions of safety or effectiveness. That structure is why a 510(k) submission can move relatively fast — FDA isn't deciding whether the device is safe and effective in the abstract, only whether it's different enough from something already cleared to matter. A De Novo request has nothing to put in that column. The device is novel enough, or different enough from anything on the market, that a predicate comparison isn't available — the requester either gets a not-substantially-equivalent finding on a prior 510(k) first, or goes to De Novo directly when no predicate exists to try against in the first place.

Since the 21st Century Cures Act amended the pathway in 2016, a requester doesn't even have to clear that first step: the direct De Novo option lets a manufacturer file straight into 21 CFR Part 860 Subpart D once it has determined, on the information available to it, that no legally marketed predicate exists — without first submitting a 510(k) and waiting for FDA to agree there's nothing to compare it against. That change removed a procedural delay, not the underlying asymmetry. Whichever door a request comes through, it still arrives with no comparator to lean on.

What 860.220 asks for instead

The content requirements at 21 CFR 860.220 read less like a comparison table and more like a dossier. Administrative information and a device description are the easy parts. The section that does the actual work is 860.220(a)(14): a discussion showing that the data and information in the request constitute valid scientific evidence, and that the probable benefit to health from using the device — under general controls, or general controls plus special controls the requester proposes — outweighs the probable injury or illness that use could cause. There's no shortcut through a comparator here. The evidence has to describe the disease or condition the device addresses and, per 860.220(a)(11), what alternative practices or procedures already exist for it, since that alternative-practices context is part of what a reviewer weighs the benefit-risk case against.

Proposing the controls that will govern the whole category

The harder asymmetry with a 510(k) is that the requester doesn't just describe the device — it proposes the special controls that would make general controls alone sufficient to manage the risks it identified: specific performance testing, labeling requirements, postmarket data collection, whatever closes the gap between novel and reasonably assured to be safe and effective. A 510(k) inherits its risk framework from the predicate it cites; a De Novo request writes that framework for the first time, for every device that will later cite this one as its own predicate. Getting the proposed controls wrong doesn't just risk this request — it risks handing the whole device type either controls too loose to defend or so conservative that the category never gets used.

Automated intake doesn't shorten the argument

FDA's acceptance review under 860.230 runs on a 15-day clock, and the eSTAR template that reshaped 510(k) intake now covers De Novo requests too, checking that the required sections are present before substantive review even starts. That automation checks structure, not substance — it confirms the benefit-risk discussion exists, not that it's convincing. A checklist built for this pathway has to walk through 860.220's paragraphs one at a time, not prompt generically for “supporting evidence,” because the paragraph a request is missing is exactly what an acceptance review is built to catch. That's the shape of the De Novo checklist previewed in the launch catalog, and it's the kind of gap a team is welcome to flag through a template contribution once the shelf is open.