FDA Outlines Predetermined Change Control Plans for AI Tools in Digital Health

FDA Draft Guidance Clarifies When AI Wellness Software Avoids Medical Device Rules

The FDA applies a risk-based framework that keeps many AI-powered wellness applications outside medical device regulation. Software focused only on physical fitness, relaxation or mental acuity typically stays unregulated provided it makes no references to disease prevention, diagnosis or treatment. The agency's Digital Health Center of Excellence coordinates these boundary decisions. Documents issued between 2019 and 2021 spell out the criteria and expectations for developers.

What this means

The materials supply specific examples of AI wellness use cases. Adaptive exercise apps and sleep analysis tools with coaching features can remain outside regulation if they avoid disease claims. Transparency in how algorithms reach decisions, real-world performance monitoring and clinical validation gain importance as claims move closer to medical territory.

Key takeaways

• FDA policy excludes low-risk general wellness AI from device regulation when claims focus only on physical fitness, relaxation or mental acuity without disease references [1].
• The 2021 AI/ML SaMD Action Plan outlines five actions that include a tailored regulatory framework, good machine learning practices and real-world performance monitoring [3].
• The Digital Health Center of Excellence established in 2020 serves as the central point for oversight of digital health technologies including AI wellness and SaMD boundary cases [2].
• Predetermined change control plans help developers document anticipated model changes so updates within approved bounds may avoid repeated premarket reviews.
• Specific claims about intended use determine whether an AI tool qualifies as software as a medical device.

The distinction between wellness products and regulated devices dates to 2019. The FDA said products that promote healthy habits without referencing diseases or conditions do not meet the definition of a medical device [1].

The agency created its Digital Health Center of Excellence in 2020. This office acts as the hub for regulatory questions on technologies that sit near the wellness and medical device line [2].

The following year the FDA issued its AI/ML SaMD Action Plan. It lists steps to build a tailored framework and support ongoing monitoring of how these tools perform after release [3].

Transparency forms a central part of the approach. Developers must describe how algorithms arrive at outputs. Clinical validation becomes relevant when wellness claims start to approach diagnostic or therapeutic ground.

Predetermined change control plans address a common trait of machine learning. Models change with new data. These plans let the FDA review the boundaries of acceptable updates in one go.

The draft includes concrete examples. One covers an app that modifies exercise suggestions based on user feedback. Another describes AI sleep analysis paired with coaching that stops short of claiming to treat insomnia.

Limitations

The document remains in draft form. Final version may change after public comments. It does not address international regulatory alignment with agencies such as the EMA or Health Canada. The rapid pace of AI advancement may necessitate frequent updates to good machine learning practice standards. Limited detail appears on cybersecurity and bias mitigation specific to wellness AI applications.

FAQ

When does an AI wellness application cross the threshold into FDA-regulated SaMD?
It crosses when claims shift from general fitness, relaxation or mental acuity to references involving disease prevention, diagnosis or treatment.

What transparency and validation requirements apply to AI algorithms under the guidance?
The draft calls for transparency in AI decision-making, real-world performance monitoring, and clinical validation in cases where claims approach diagnostic or therapeutic territory.

How should developers implement predetermined change control plans for AI wellness tools?
The plans require documentation of anticipated changes to the model upfront. Regulators then assess the plan once rather than review each update separately.

Sources / References

[1] General Wellness: Policy for Low Risk Devices https://www.fda.gov/regulatory-information/search-fda-guidance-documents/general-wellness-policy-low-risk-devices

[2] Digital Health Center of Excellence https://www.fda.gov/medical-devices/digital-health-center-excellence

[3] Artificial Intelligence and Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device