Understanding FDA's Predetermined Change Control Plan for AI-Enabled Devices
- 2 days ago
- 4 min read
Artificial intelligence (AI) is transforming healthcare, especially through AI-enabled medical devices that learn and improve over time. But this rapid evolution poses a challenge for regulators who must ensure these devices remain safe and effective as they change. To address this, the U.S. Food and Drug Administration (FDA) introduced the Predetermined Change Control Plan (PCCP) guidance, finalized in August 2025. This guidance offers a clear path for manufacturers to propose and manage future modifications to AI and machine learning (ML) models without submitting a new marketing application for every update.
This post explains what the PCCP is, why it matters, and how it supports innovation while maintaining patient safety.
What Is the Predetermined Change Control Plan?
The Predetermined Change Control Plan is a framework the FDA uses to review and approve planned changes to AI-enabled device software functions before those changes happen. Instead of treating every software update as a new product requiring full regulatory review, the PCCP allows manufacturers to outline:
The types of modifications they expect to make
How they will develop and validate those changes
An assessment of the potential impact on safety and effectiveness
Once the FDA reviews and accepts the PCCP during the initial marketing submission, manufacturers can implement the planned changes within the agreed scope without submitting a new application each time.
Why the PCCP Matters for AI-Enabled Devices
AI and ML models in medical devices often improve through continuous learning and updates. Traditional regulatory pathways are not designed for this iterative process, which can slow down innovation and delay patient access to better technology.
The PCCP addresses this by:
Encouraging iterative improvement
Maintaining regulatory oversight
Reducing administrative burden for manufacturers
Providing clarity and predictability in the regulatory process
This approach aligns with the FDA’s broader commitment to modernize regulation for AI-enabled medical devices, balancing innovation with patient safety.
Key Components of a PCCP
To create a successful PCCP, manufacturers must include three main elements in their submission:
1. Description of Planned Modifications
This section outlines the types of changes the manufacturer expects to make to the AI/ML model. Examples include:
Updating training data sets
Changing algorithms or model architecture
Adjusting performance parameters
The description should be specific enough to define the scope of future changes but flexible enough to allow meaningful improvements.
2. Methodologies for Development and Validation
Manufacturers must explain how they will develop and test the planned modifications. This includes:
Data collection and management processes
Model training and retraining methods
Validation strategies to ensure the updated model performs safely and effectively
Clear methodologies help the FDA assess whether the manufacturer can reliably control changes.
3. Impact Assessment
This part evaluates how the planned changes might affect device safety and effectiveness. It should address:
Potential risks introduced by modifications
Measures to mitigate those risks
How the manufacturer will monitor device performance post-change
An honest and thorough impact assessment builds trust with regulators and supports patient safety.
How the FDA Reviews the PCCP
The FDA reviews the PCCP as part of the initial marketing submission for the AI-enabled device. The agency evaluates whether the planned changes and control methods are reasonable and sufficient to maintain safety and effectiveness.
If the FDA accepts the PCCP, the manufacturer can implement the described modifications without submitting a new marketing application for each change. However, if changes fall outside the PCCP scope or raise new safety concerns, the manufacturer must notify the FDA and may need to submit a new application.
Practical Examples of PCCP Use
Example 1: AI-Based Diagnostic Imaging Software
A company develops AI software that analyzes medical images to detect tumors. Their PCCP includes plans to update the model with new image data from diverse patient populations to improve accuracy. They describe how they will retrain the model, validate performance with clinical data, and monitor for false positives or negatives.
This approach allows the company to regularly improve the software while ensuring it remains safe and effective.
Example 2: AI-Enabled Wearable Health Monitor
A wearable device uses AI to detect irregular heart rhythms. The manufacturer’s PCCP outlines plans to refine the algorithm based on real-world data collected from users. They specify validation steps including simulated testing and clinical trials to confirm accuracy before deployment.
The PCCP helps the company adapt the device to new data trends without regulatory delays.
Benefits for Manufacturers and Patients
The PCCP offers several advantages:
Faster access to improved AI-enabled devices
Reduced regulatory workload for manufacturers
Clear expectations for managing AI/ML changes
Ongoing assurance of device safety and effectiveness
Patients benefit from devices that evolve with new data and technology, potentially leading to better health outcomes.
Challenges and Considerations
While the PCCP provides a useful framework, manufacturers should consider:
Defining the scope of changes carefully to avoid regulatory issues
Maintaining thorough documentation of development and validation processes
Monitoring device performance continuously after changes
Being prepared to submit new applications if changes exceed the PCCP scope
Effective communication with the FDA throughout the product lifecycle remains essential.
Moving Forward with AI-Enabled Device Innovation
The FDA’s Predetermined Change Control Plan guidance marks a significant step toward modernizing regulation for AI-enabled medical devices. It balances the need for innovation with patient safety by allowing planned, controlled updates to AI/ML models without repeated full reviews.
Manufacturers should start integrating PCCPs into their development and regulatory strategies to stay ahead in this evolving field. By doing so, they can bring safer, more effective AI-enabled devices to patients faster.
