The FDA finalized its guidance on Predetermined Change Control Plans (PCCPs) in October 2025, providing a structured pathway for manufacturers of AI and machine learning-based medical devices to implement planned algorithm modifications without submitting a new 510(k) or PMA supplement for each change. The PCCP concept was introduced in the FDA's 2019 action plan for AI/ML-based Software as a Medical Device (SaMD) and has since been refined through multiple draft guidance iterations and public comment periods. The final guidance represents a significant regulatory innovation that acknowledges the iterative nature of AI/ML development.
What a PCCP Must Include
A PCCP is a component of a marketing submission - either a 510(k), De Novo, or PMA - that describes in advance the types of modifications the manufacturer plans to make to the device and the controls that will govern those modifications. According to the final guidance, a complete PCCP must include three core elements:
- Description of Modifications: A precise characterization of the algorithm changes anticipated, including the scope of retraining, the datasets to be used, and the performance parameters expected to change
- Modification Protocol: The specific testing, validation, and verification steps that will be executed before any approved modification is implemented
- Impact Assessment: An analysis of how each planned modification affects device safety, effectiveness, and the device's intended use
Types of Modifications Covered by a PCCP
The FDA distinguishes between modifications that require a new submission regardless of a PCCP (such as changes to intended use or fundamental technology) and those that can be managed under an approved PCCP. Modifications typically suited to PCCP coverage include performance improvements within the same intended use, retraining on expanded or updated datasets without changing the model architecture, and updates to output thresholds within pre-specified ranges validated through the modification protocol. Manufacturers must be careful not to exceed the scope of the approved PCCP; out-of-scope modifications continue to require a separate submission.
Integration with Total Product Lifecycle (TPLC) Framework
The PCCP guidance is designed to work within the FDA's broader Total Product Lifecycle (TPLC) approach to AI/ML device oversight. Manufacturers are expected to maintain robust post-market monitoring systems that can detect performance degradation or distributional shift in the real-world data environment. The PCCP must describe how post-market performance data will inform decisions about when and whether to implement a planned modification. This creates a formal feedback loop between real-world performance and the algorithm development cycle.
Submitting a PCCP: Practical Guidance
From a submission strategy standpoint, manufacturers should develop the PCCP in parallel with the device's clinical and analytical validation activities, not as an afterthought. The modification descriptions and protocols must be specific enough that FDA reviewers can assess their safety implications, while being broad enough to capture the range of intended improvements. Overly narrow PCCPs limit the manufacturer's flexibility, while overly broad ones risk rejection during review. Sequence Group's regulatory strategy team has experience structuring PCCPs that balance specificity with practical adaptability.
Key Takeaways for AI/ML Device Manufacturers
The finalized PCCP guidance is a meaningful development for the medical AI industry, reducing regulatory friction for iterative improvement cycles. Manufacturers of AI/ML-based SaMD should assess their current and pipeline products to identify where a PCCP would be beneficial, build internal governance processes that align with the modification protocol requirements, and ensure their post-market surveillance infrastructure can support the monitoring obligations that come with an approved PCCP.
Frequently Asked Questions
What is a Predetermined Change Control Plan (PCCP) and what does it allow?
A PCCP is a component of a marketing submission - 510(k), De Novo, or PMA - that describes in advance the types of algorithm modifications the manufacturer plans to make and the controls governing those modifications. An approved PCCP allows the manufacturer to implement the described modifications without submitting a new marketing authorization for each change, reducing regulatory friction for the iterative improvement cycles typical of AI/ML device development.
What types of algorithm changes can be covered by a PCCP?
The FDA guidance distinguishes between modifications requiring a new submission regardless of a PCCP - such as changes to intended use or fundamental technology - and those that can be managed under an approved PCCP. Changes typically suited to PCCP coverage include performance improvements within the same intended use, retraining on expanded or updated datasets without changing model architecture, and updates to output thresholds within pre-specified ranges validated through the modification protocol.
How should manufacturers develop a PCCP as part of their submission strategy?
Manufacturers should develop the PCCP in parallel with the device's clinical and analytical validation activities, not as an afterthought. Modification descriptions and protocols must be specific enough for FDA reviewers to assess safety implications, while being broad enough to capture the range of intended improvements. Overly narrow PCCPs limit flexibility; overly broad ones risk rejection during review. The PCCP must also describe how post-market performance monitoring will inform decisions about implementing planned modifications.