FDA Draft Guidance For Predetermined Change Control Plans (PCCPs) for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions

Since the publication of the Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan in 2021, interested parties have been awaiting guidance from FDA that would allow for alterations to already approved medical devices that use machine learning. On April 3, 2023, FDA published a draft guidance document titled, “Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions.” In fact, Congress anticipated the need for this change and added section 515C to the Food, Drug, and Cosmetics Act (FD&C Act) as part of the Food and Drug Omnibus Reform Act of 2022. This new section provides FDA with express authority to allow Predetermined Change Control Plans (PCCPs).

The draft guidance document that followed in April 2023 is FDA’s answer to a long unanswered question. Traditionally, FDA will approve or clear a medical device based on their review of the application that details the device in its current state. When a manufacturer intends to alter the device to make an improvement or modification, the manufacturer is required to submit a premarket approval supplement, De Novo submission, or a new premarket notification. However, with the increasing number of marketing submissions and pre-submissions for devices using machine learning technologies that are modified over time, the typical review process is no longer sufficient.

AI/ML-based SaMD has turned this long-accepted concept of the static unchanging medical device on its head. The medical device in the hands of the end user should be the same medical device that underwent FDA’s rigorous review process. However, the device that received FDA approval or clearance can now be drastically different than the device that is in the end user’s hands. Medical devices that use machine learning based software functions continue to improve themselves in real time as they gather data through their use in the real world. This means that the device being used is not the same device that received approval or clearance from FDA.

The draft guidance document states that a PCCP may be used in lieu of premarket authorization. However, manufacturers must submit a marketing submission when the modification to the device could alter the “intended use of the device or could significantly affect the safety or effectiveness of the device.” In essence, this PCCP would allow for premarket authorization for pre-specified automatic and manual modifications without requiring the lengthy and expensive process of submitting the entire device for review. The PCCP will require the following parts in the submission: a Description of Modifications, a Modification Protocol, and an Impact Assessment. The PCCP will also impact labels and warnings. Labels will need to be updated to include the incorporation of language that warns the user that the device incorporates machine learning technology and of the existence of the PCCP. Interested stakeholders should submit comments on the draft guidance document by July 3, 2023.

Submitting a PCCP will have implications for products liability. Manufacturers are encouraged to consult with legal counsel to ensure that they remain compliant with labelling requirements and to ensure that the PCCP is correctly drafted to cover all planned modifications.