FDA Issues Final Guidance on Predetermined Change Control Plans for AI-Enabled Devices

Manufacturers are generally required to submit new marketing applications to the FDA for review when a software modification could significantly alter that device’s safety and effectiveness. PCCPs are supplemental documentation that manufacturers of medical devices may include as part of typical marketing submissions, and they provide manufacturers with an avenue to avoid that burdensome process entirely. Comprising at least three sections – a description of modifications, a modification protocol, and an impact assessment – PCCPs are intended to notify the FDA of anticipated modifications to a device that may alter its safety and effectiveness. If the FDA authorizes a manufacturer’s PCCP, that manufacturer can implement modifications in line with the PCCP without having to submit new marketing applications.

The FDA specifies that a device manufacturer stays within the bounds of an authorized PCCP when it:

• Only implements modifications contemplated by the relevant PCCP
• Deploys the authorized modifications in a manner that is consistent with the relevant PCCP.

The final guidance finalizes previous draft guidance, which we covered in April 2023, and clarifies the FDA’s view of PCCPs in the AI-enabled medical device context. The guidance also discusses how manufacturers should leverage the PCCP process and the content that should be included in their PCCPs. Because many of the FDA’s substantive recommendations remain unchanged from the draft guidance, this On the Subject highlights changes that the FDA made in the final guidance, rather than providing a comprehensive outline of the guidance’s contents. Affected manufacturers should review the final guidance to see the FDA’s full suite of suggestions.

DEFINITIONS

The FDA broadened the scope of its guidance to cover all AI-enabled medical devices, rather than just machine-learning-enabled devices, a subset that the draft guidance focused on exclusively.

In keeping with this change, the FDA defines terms such as “artificial intelligence” and “machine learning”:

• Artificial Intelligence: “A machine based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. Artificial intelligence systems use machine- and human-based inputs to perceive real and virtual environments; abstract such perceptions into models through analysis in an automated manner; and use model inference to formulate options for information or action.”
• Machine Learning: “A set of techniques that can be used to train AI algorithms to improve performance at a task based on data.”

GENERAL CHANGES TO THE GUIDANCE

Throughout the guidance, the FDA added recommendations that manufacturers consider the unique characteristics of their device’s intended use populations (such as ethnicity, gender, and disease severity) and intended environments in their PCCP.

The FDA also newly recommends that manufacturers update a device’s labeling and public-facing documents as the device is modified. Suggested updates include detailing when modifications are implemented, describing how they were implemented, and explaining how users were informed of the modifications.

COMPONENTS OF A PCCP

The recommended structure of a PCCP remains largely unchanged from the FDA’s previous draft guidance. The FDA still recommends that three sections specifically be included:

• A description of modifications
• A modification protocol
• An impact assessment.

DESCRIPTION OF MODIFICATIONS

This section of the PCCP should generally “identify the specific, planned modifications” that the manufacturer intends to apply to the device following clearance or approval. The FDA has previously suggested what information should be included, such as descriptions of the planned modifications and modes of implementation.

The FDA added to these recommendations in its final guidance, specifying that manufacturers are expected to detail the projected frequency of updates and raise effective guardrails against the range of automatic updates.

MODIFICATION PROTOCOL

The FDA recommends that the modification protocol section of the PCCP detail “the methods that will be followed when developing, validating, and implementing modifications” described in the previous section.

The final guidance recommends additional content that manufacturers should include in their PCCP submissions, such as:

• Information on how to mitigate bias with respect to data-management practices
• Information on how reference standards are determined, including when clinician interpretation is leveraged
• Information on unresolvable failures in performance evaluation of a proposed modification
• Information on the manufacturer’s post-market surveillance strategies, especially related to post-modification safety and effectiveness monitoring.
• The FDA suggests that manufacturers adhere to their preexisting risk-management strategies when completing the modification protocol section of the PCCP.

IMPACT ASSESSMENT

In the impact assessment section of the PCCP, manufacturers should weigh the benefits and risks of using a PCCP for the device in question and discuss how the identified risks might be mitigated.

In its final guidance, the FDA clarifies what information should be included in a PCCP for a combination product. Relevant considerations are:

• How modifications to the device constituent part might impact the non-device constituent parts
• How modifications to the device constituent part may affect the combination product in its entirety.

The final guidance also revises the FDA’s previous recommendations on how the impact assessment should be structured. The FDA previously suggested that elements of the impact assessment could be incorporated freely into other sections of the PCCP where relevant. Although the final guidance still allows this practice, it specifies that clear references to that information must be included in the impact assessment section itself.

APPENDICES

As part of the guidance, the FDA provides two helpful resources to aid manufacturers in navigating the PCCP drafting process. First, the FDA lists a considerable number of sample questions for manufacturers to contemplate and discuss in their PCCP submissions. Second, the guidance provides several hypothetical scenarios that describe when PCCPs might be used in lieu of new marketing submissions. These include a new reference scenario related to a combination product titled, “Optical Imaging System Co-Packaged With Imaging Drugs.”

KEY TAKEAWAYS

This final guidance provides a useful framework for manufacturers and developers to address projected modifications to AI-enabled medical devices. However, it also introduces potential implementation challenges that will require manufacturers to assess the impact of these recommendations on existing quality system processes such as design control, data management practices, risk management plans, adverse event reporting, and labeling.

Manufacturers of AI-enabled medical devices should review the final guidance in its entirety to see if PCCPs might be useful in their marketing submission process. The FDA will host a webinar to answer questions about the final guidance on January 14, 2025.

For more information on recent FDA guidance on AI in medical products, contact the authors of this article or any other member of McDermott’s Food, Drug & Medical Device Regulatory Group. Additional resources are also available on McDermott’s AI in Healthcare resource center.