The FDA released a guidance to regulate artificial intelligence (AI) enabled medical devices focusing on the iterative process of improving device software through modifications. The guidance introduces a Predetermined Change Control Plan (PCCP) that manufacturers can submit to outline planned modifications to AI-driven devices.
The development of the guidance begins with the publication of the FDA’s 2019 “Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML) BAsed Software as a Medical Device”. The framework outlines a foundation for premarket review and modifications of AI/ML-driven devices. It was built upon the FDA’s existing premarket practices and integrated international risk categorization principles.
In response to the growing interest and potential iterative updates to AI models through real-world data, the FDA then proposed the introduction of the “Predetermined Change Control Plan” in 2019. The plan laid out a comprehensive approach for regulating AI/ML devices.
In 2022, the White House introduced the Blueprint for an AI Bill of Rights, outlining principles like data privacy and algorithmic discrimination protections. Finally, the Food and Drug Omnibus Reform Act of 2022 (FDORA) added section 515C to the FD&C Act which requires provisions for Predetermined Change Control Plans for devices.
The legislation enacted in December 2022 allows manufacturers to modify their devices under an FDA-approved PCCP without the need for new premarket approvals. The FDA’s guidance has evolved in response to these legislative, public, and industry developments.
The guidance allows manufacturers to submit a PCCP as part of their marketing submissions which outlines planned modifications to AI-enabled devices without the need for separate submissions for each change. The approach allows healthcare organizations to benefit from continuous and clear improvements in AI applications like enhanced diagnostic capabilities and personalized treatment options.
Through the use of iterative updates, healthcare providers can implement advanced tools that make use of real-world data. The emphasis the guidance places on safety and efficiency also ensures that advancements do not compromise patient care standards and reinforce the progression of AI in healthcare by promoting trust.
The guidance stems from the FDORA. It simply puts into perspective the relevance of PCCPs which allows manufacturers to implement modifications without separate marketing submissions. The guidance therefore establishes boundaries related to the FDORA and provides clear guidelines to medical device manufacturers to Section 515 that prevents any potential vagueness.
Related: HIPAA Compliant Email: The Definitive Guide
The function of the FD&C Act is to regulate the safety and efficacy of food, drugs, and medical devices.
The oversight mechanisms imposed by the guidance include requiring manufacturers to document planned modifications, implement a verification and validation plan, conduct risk assessments, and ensure that changes do not compromise device safety.
Machine learning differs from AI in that machine learning is a subset of AI-focused specifically on algorithms that allow systems to learn from data and improve their performance over time without being explicitly programmed for each task.