Predetermined Change Control Plan: How to Write a PCCP

Clinician and engineer reviewing an AI scan analysis with a change control cycle for model updates

FDA's guidance Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions is final. It was first issued on 4 December 2024, and the current version is dated 18 August 2025. The legal basis is section 515C of the FD&C Act, added by section 3308 of the Food and Drug Omnibus Reform Act of 2022.

An AI model that is retrained after clearance is, in most cases, a changed device. Without a predetermined change control plan (PCCP), a significant change generally needs a new marketing submission. This article covers what a PCCP is, which submissions can establish one, its three components, which changes fit, labelling, and how to use and change an authorised PCCP.

What a PCCP does

According to the guidance, an authorised PCCP specifies planned modifications that would otherwise need a new marketing submission under 21 CFR 807.81(a)(3) or 21 CFR 814.39(a). Because FDA reviewed those modifications in the original submission, they can be implemented without a new one, as long as they follow the plan.

A PCCP is optional. FDA does not authorise a PCCP in a Pre-Submission, but encourages manufacturers to use the Q-Submission Program to get feedback on a proposed PCCP first.

Our article on FDA AI guidance covers the wider submission content for AI-enabled devices.

Which submissions can establish a PCCP

Pathway Submission types FDA considers appropriate
PMA Original PMA; modular PMA (PCCP as a module); 180-day supplement; panel-track supplement; real-time supplement (minor changes, by agreement)
510(k) Traditional 510(k); Abbreviated 510(k)
De Novo Original De Novo request

Source: PCCP guidance (PDF), Section V.B.

For a 510(k), FDA treats the PCCP as part of the device's technological characteristics. A new device must be compared with the predicate as cleared, before any changes made under the predicate's PCCP. Our articles on 510(k) submissions and De Novo review cover those routes.

The three components

Component What it contains Guidance section
Description of Modifications Each planned modification, with the specifications for its characteristics and performance and a rationale VI
Modification Protocol The methods to develop, validate and implement each modification, with pre-defined acceptance criteria VII
Impact Assessment The benefits and risks of each modification, their mitigations, and the combined effect VIII

Source: PCCP guidance, Sections V.A and VI to VIII.

Description of Modifications

FDA recommends a limited number of specific modifications that can be verified and validated. For each one, the description should state:

  • whether it is implemented automatically, manually, or both;

  • whether it applies to all devices in the same way (global) or differs by site or patient (local), and, for local changes, what triggers them;

  • how often updates are expected, from rarely (a mostly "locked" device) to continuously.

Modification Protocol

The guidance names four components, to be addressed for each modification:

  1. Data management: how new data are collected, annotated, stored and controlled, and how test data are kept separate from training and tuning data.

  2. Re-training: which processing steps may change, the methods used, any architecture changes with a rationale, and the triggers for re-training.

  3. Performance evaluation: study design, metrics, pre-defined acceptance criteria and statistical tests, comparing the modified device with both the original and the last modified version.

  4. Update procedures: how updates are deployed, how users are informed, how legacy users are affected, labelling updates and post-market monitoring.

If a modification fails its acceptance criteria and the failure cannot be resolved, the protocol should state that the failure will be recorded and the modification will not be implemented. FDA suggests a traceability table linking each modification to the methods that support it.

Impact Assessment

The Impact Assessment should:

  • compare each modification individually with the unmodified device;

  • discuss the benefits and risks of each modification, including unintended bias;

  • explain how the Modification Protocol keeps the device safe and effective;

  • discuss how modifications affect each other;

  • describe the cumulative impact of all modifications.

In our view, write the Modification Protocol from your actual development pipeline. A protocol the team cannot follow becomes a deviation risk on the first retraining cycle.

Which modifications fit

According to Section VI.C, modifications should maintain or improve safety or effectiveness. Types that may fit include:

  • changes to quantitative performance, such as retraining on new data from the intended use population with the same type of input;

  • changes to inputs, such as new makes or models of an acquisition system, limited new input types, or compatible hardware, operating systems or cloud infrastructure;

  • certain changes to use and performance, such as a specific subpopulation within the original indication.

Modifications must keep the device within its intended use and, for a 510(k), substantially equivalent to the predicate. FDA expects most changes to the indications for use to be difficult to assess in advance, and recommends discussing them through a Q-Submission.

In our view, a narrow, specific PCCP that FDA can authorise is worth more than a broad one that is withdrawn during review.

Labelling and public summaries

According to Section V.C, the labelling should state that the device has an authorised PCCP and incorporates machine learning, so users know updates may change performance, inputs or use. When modifications are implemented, the labelling should describe them, how they were implemented and how users are informed. It should not describe modifications that have not yet been implemented.

The PCCP should also be summarised in the public 510(k) summary, De Novo decision summary or PMA summary of safety and effectiveness. A new version of a device that must bear a UDI needs a new device identifier.

Using and changing an authorised PCCP

A modification is consistent with the authorised PCCP when it is in the Description of Modifications and implemented according to the Modification Protocol. It is then documented in the quality system, without a new submission.

Situation Consequence
Modification in the PCCP and implemented per the protocol No new submission; document it in the quality system
Modification in the PCCP but not implemented per the protocol A new marketing submission is likely required
Modification not in the PCCP Assess it under the normal device modification rules
Change to the PCCP itself Generally needs a new marketing submission; a Special 510(k) may be appropriate in some cases

Source: PCCP guidance, Sections V.D and V.E.

The guidance refers to Quality System Regulation sections that pre-date the QMSR, which took effect on 2 February 2026. Implement the PCCP under your current quality management system.

In our view, put the PCCP under version control from the start. FDA references its title and version number in the authorisation letter, and every later change is traced back to it.

A hypothetical example

Hypothetical example: a Dutch start-up has an AI image-analysis SaMD and plans a Traditional 510(k) (the pathway is assumed).

  1. Modifications: it limits the PCCP to retraining on new data from the intended use population and adding scanners from other manufacturers, both implemented manually and globally.

  2. Protocol: for each modification, it defines data sequestration, re-training triggers, acceptance criteria against the original and last version, and update and user communication steps.

  3. Impact: it assesses each change alone and both together, including bias across subgroups.

  4. Feedback: it discusses the PCCP with FDA through a Q-Submission before filing.

  5. After clearance: each retraining cycle follows the protocol and is documented in the quality system; a change outside the PCCP is assessed separately.

Conclusion

A PCCP lets FDA authorise specific AI model changes in advance. It needs three components, must stay within the intended use, and only works if each change follows the Modification Protocol.

In our view, keep the PCCP narrow and specific, base it on your real retraining process, and get FDA feedback before submitting.

If you need help writing a PCCP, our software lifecycle management team can help, or contact MedQAIR.

Frequently asked questions

Is a PCCP mandatory?

No. A PCCP is optional. It lets FDA authorise planned modifications in advance, so they do not need a new marketing submission.

Can a PCCP be added to a device that is already on the market?

Yes. A PCCP can be established through a marketing submission for a device that has already been authorised, such as a PMA supplement or a 510(k).

Can FDA authorise a PCCP in a Pre-Submission?

No. FDA encourages Q-Submission feedback on a proposed PCCP, but a PCCP is only authorised as part of a marketing authorisation.

Can a PCCP include automatic updates?

FDA will consider PCCPs with automatically implemented modifications and recommends defining clear boundaries for them and discussing them through the Q-Submission Program.

What happens if a modification fails its acceptance criteria?

The failure should be recorded and that modification not implemented, unless a root cause analysis shows the failure is unrelated to the PCCP.

Latest Blogs

November 20, 2026

On 7 January 2025, FDA issued the draft guidance Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations.

November 19, 2026

Since 2 February 2026, FDA's Quality Management System Regulation (21 CFR Part 820) has incorporated ISO 13485:2016 by reference. Even

November 18, 2026

Since 2 February 2026, 21 CFR Part 820 has been the Quality Management System Regulation (QMSR), incorporating ISO 13485:2016 by

logo

Unlock Your Quick Guide to AI Act
Compliance!

Explore AI-enabled SaMD requirements with our easy step-by-step guide.

Cookies help us improve your experience on our website. By using our site, you consent to the use of cookies as described in this policy.