FDA Outlines Predetermined Change Control Plans for AI-Enabled Medical Devices

FDA Guidance Details Plans for Updates to AI Software Used in Medical Devices

The U.S. Food and Drug Administration has expanded its primary resource on artificial intelligence and machine learning tools cleared for medical use. The page brings together documents that describe a risk-based total product lifecycle approach for software classified as a medical device. Officials highlight predetermined change control plans as one way manufacturers can map out future modifications to adaptive algorithms ahead of time.

These steps address how AI systems often improve or shift once deployed in clinics. The FDA, along with international partners, has also set out shared expectations for development practices. The materials draw on the agency's 2021 action plan and later clarifications to give developers more concrete direction.

The FDA regulates AI/ML technologies in medical devices as Software as a Medical Device. Its framework centers on clinical evaluation, quality systems, and post-market performance monitoring. [1]

The agency, Health Canada, and the UK MHRA released 10 guiding principles for good machine learning practice in 2021. Those principles target data quality, model robustness, clinical integration, and lifecycle management. [2]

Guidance updated in April 2023 spells out how manufacturers should prepare Predetermined Change Control Plans. Such plans must list the kinds of changes expected, show how risks will be assessed and controlled, and note which modifications would still require a fresh premarket submission. [3]

The FDA said these plans form part of a broader push to track device performance after clearance. Real-world data collection remains a consistent theme across the documents.

What this means

Documents show regulators expect AI tools in medicine to keep changing after approval. Predetermined plans create a structured lane for some updates so full re-reviews are not needed for every tweak. The materials also flag the importance of clear records on training data and steps taken to limit bias.

Key takeaways

• The FDA regulates AI/ML technologies in medical devices as Software as a Medical Device and follows a risk-based approach focused on clinical evaluation, quality systems, and post-market performance monitoring. [1]
• In 2021 the FDA, Health Canada, and the UK MHRA issued 10 guiding principles for good machine learning practice that address data quality, model robustness, clinical integration, and lifecycle management. [2]
• Manufacturers must submit Predetermined Change Control Plans for AI/ML device software functions that describe anticipated modifications, risk assessments, and mitigation steps to avoid new premarket submissions for planned updates. [3]
• A total product lifecycle approach accounts for the adaptive nature of AI algorithms. [1]
• Predetermined Change Control Plans enable managed evolution of devices while maintaining regulatory oversight. [3]

Limitations

FDA guidance documents are not legally binding regulations and represent current agency thinking only. Rapid advancements in foundational AI models may require frequent revisions. Real-world performance monitoring data remains limited for many authorized devices.

FAQ

What are the guiding principles for good machine learning practice in medical device development according to FDA?
The ten principles developed with Health Canada and the UK MHRA cover topics such as data quality, model robustness, clinical integration, and lifecycle management for AI/ML medical devices. [2]

How should manufacturers submit predetermined change control plans for AI/ML-enabled devices?
Manufacturers include the plans in their marketing submissions. Each plan details the kinds of changes expected, the risk assessment process, and the controls that will keep the device safe and effective. [3]

What regulatory framework does the FDA apply to modifications in AI/ML-based SaMD?
The agency uses a risk-based total product lifecycle approach. Predetermined Change Control Plans help manage modifications while preserving oversight. [1][3]

Sources / References

1. Artificial Intelligence and Machine Learning Software as a Medical Device. U.S. Food and Drug Administration (FDA). https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device

2. Good Machine Learning Practice for Medical Device Development: Guiding Principles. U.S. Food and Drug Administration (FDA). https://www.fda.gov/medical-devices/software-medical-device-samd/good-machine-learning-practice-medical-device-development-guiding-principles

3. Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence and Machine Learning-Enabled Device Software Functions. U.S. Food and Drug Administration (FDA). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/marketing-submission-recommendations-predetermined-change-control-plan-artificial-intelligence-and