How to read a predetermined change control plan

A device that ships today is not necessarily the device that ships next year. Software gets retrained, manufacturing processes get refined, algorithms get retuned — and until recently, a change substantial enough to affect what FDA cleared or approved meant a new submission, reviewed from scratch, no matter how well the sponsor could already predict and control it. A predetermined change control plan, or PCCP, is FDA’s answer to that mismatch: a defined set of future changes, pre-authorized inside the original submission, so implementing one later doesn’t mean starting over.

Why this exists

For years before Section 515C existed, FDA was already negotiating informal update plans for adaptive algorithms case by case, through the ordinary submission process, because a machine-learning model that legitimately warrants frequent retraining doesn’t fit a submission cycle built around occasional, discrete changes. Section 515C didn’t invent that idea — it gave it a durable statutory home instead of leaving every sponsor to negotiate the same arrangement from scratch. The underlying tension is the same one that shows up anywhere a device changes over its life: 21 CFR § 807.81(a)(3) already requires a new 510(k) when a change could significantly affect safety or effectiveness, and PMA holders face a comparable supplement requirement. A PCCP doesn’t remove that rule. It lets a sponsor satisfy it once, in advance, for a specific, bounded category of future changes.

What’s actually in a plan

The description of modifications has to be specific enough to review, not just aspirational — closer to “the model may be retrained on new data within stated performance bounds, for the same intended use” than “the algorithm will be improved over time.” The methodology is the commitment underneath that description: what gets verified before a change ships under the plan, what the acceptance criteria are, what happens if a check fails, and how the change gets documented and communicated once it’s out — to FDA, and depending on the change, to the people using the device. The impact assessment is the benefit-risk case for the modifications as described, including a look at what happens if a change performs worse than expected in the field, not just better. All three have to hold together — a tightly written modification description with a vague methodology behind it is not a plan FDA is likely to authorize.

Where the authority stops

A PCCP only covers what’s actually inside its bounded description. A change that’s adjacent to the plan but not named in it isn’t covered by the earlier authorization — it needs its own path, the same as any unplanned device change would. Authorization of a plan also doesn’t touch FDA’s other authorities: inspection, adverse event reporting, and recall requirements all apply exactly as they would without one. It helps to read a PCCP less as a blanket exemption and more as a narrow, pre-negotiated exception carved out for specific, named changes — everything outside of it reverts to the default rules, including the plain fact that a device modified outside its plan’s stated bounds is, for regulatory purposes, an unauthorized change like any other.

A PCCP sits next to the same pathway decisions covered elsewhere here — whether a change needs a new submission at all is exactly the judgment call this option is built to preempt, for a narrow, pre-identified set of changes. It’s worth understanding as an addition to the standard 510(k) submission types, not a replacement for them, and the device files it lives in are the same files that eventually produce a De Novo order or a 510(k) clearance for the next sponsor to read.