USFDA Guidance: Predetermined Change Control Plans (PCCPs) for AI/ML Medical Devices

Artificial Intelligence and Machine Learning (AI/ML) are increasingly used in medical devices—from diagnostic imaging software to digital pathology and decision support systems. Unlike traditional devices, AI/ML-enabled products can evolve over time as they “learn” from new data.

Historically, each modification required a new FDA submission, even for minor algorithm updates. This slowed down innovation and patient access to improved technology. To address this challenge, the FDA introduced the Predetermined Change Control Plan (PCCP) framework—first outlined in 2019 as a concept and now finalized in this 2025 guidance. PCCPs provide a proactive regulatory pathway for manufacturers to anticipate and manage device changes without repeated submissions.

• AI refers to machine-based systems that predict, recommend, or make healthcare-related decisions, learning from data and adapting over time.

• ML (Machine Learning), a subset of AI, enables systems to improve automatically with new data.

• AI-Enabled Device Software Function (AI-DSF): A software function that incorporates artificial intelligence (AI), and is classified as a regulated medical device function under U.S. law.

What is a Predetermined Change Control Plan (PCCP)?

A PCCP is a documented plan detailing what future device modifications are planned (that would otherwise require an FDA submission) and how changes will be made, evaluated, and communicated to ensure ongoing safety and effectiveness., in advance:

• The types of changes expected in an AI/ML-enabled device,

• The methods for implementing and controlling those changes, and

• How safety and effectiveness will be maintained post-change.

In essence, a PCCP functions like a “roadmap” of planned future modifications. Once FDA approves it, manufacturers can make those updates without resubmitting each change, provided they stay within the approved scope.

When and How Does PCCP Apply?

• Applies to:

  • AI-enabled devices manufacturers plan to update periodically (e.g., model retraining, expanded compatibility), whether updates are automatic or manual.

• Channels:

  • Premarket Approval (PMA)

  • 510(k) Notification/Clearance

  • De Novo Classification

• Not for changes that wouldn’t require a submission under existing device modification policies (these are handled via Quality System Regulations).

  
  

The guidance outlines three essential components of a PCCP:

1. Description of Modifications

   • Specifies the anticipated changes (e.g., expanding input data types, enhancing algorithm performance, adapting to new patient populations).

   • Must be clear, limited, and well-defined.

2. Modification Protocol

   • Provides detailed methods for implementing changes, including validation and verification steps.

   • Must ensure that modifications are scientifically sound and do not introduce unacceptable risks.

3. Impact Assessment

   • Explains how changes will be monitored and evaluated to ensure continued device safety and effectiveness.

   • Includes a risk management framework aligned with FDA’s Quality System Regulation (21 CFR Part 820).

Marketing Submission Recommendations

When submitting a PCCP as part of a marketing application (510(k), De Novo, or PMA), FDA recommends:

• Clearly distinguishing PCCP content from the “static” parts of the submission.

• Using traceability matrices to link modifications with protocols and risk assessments.

• Providing clinical evidence where appropriate to support the anticipated changes.

• Ensuring all Good Machine Learning Practices (GMLP) are followed.

This helps reviewers assess the predictability and reliability of the proposed PCCP.

The FDA’s final guidance on Predetermined Change Control Plans for AI/ML-enabled devices marks a pivotal shift in regulatory science. It balances innovation with patient safety, giving manufacturers flexibility while ensuring robust oversight. For full details, refer to the official FDA resources:

• FDA Guidance Page

• Full Guidance PDF