FDA Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions
Overview of FDA Predetermined Change Control Plan
It’s no news that medical devices incorporating artificial intelligence and machine learning (AI/ML) technology have been on the rise over the past several years. In 2021, the FDA authorized a record 115 submissions, an 83% increase from 2018.
While AI/ML has been busy disrupting the medical device market, the FDA has tasked itself with establishing regulation to address this evolving technology. Regulating this technology has introduced a unique set of challenges for the FDA, from how to handle device security to software updates. As is expected with a software product, the medical device software that incorporates AI/ML can - and should be - routinely updated. The nature of AI/ML software is that it has the ability to to learn and, ideally, improve. This has the potential to provide even more benefit to patients and users. However, there is a fine line between software improvements and affecting the safety or effectiveness of the device. With this April 2023 guidance [1], the FDA is seeking to outline best practices on how to make software/device improvements without affecting device safety or effectiveness, all while minimizing the regulatory burden. The agency presents the Predetermined Change Control Plan (PCCP) as an approach for achieving this.
By including a PCCP in a marketing submission, manufacturers can proactively pre-specify and seek premarket authorization for intended modifications (and their method of implementation) to an ML-DSF without needing additional marketing submissions for each modification delineated and implemented in accordance with the PCCP. In other words, a PCCP, as part of a marketing submission, is intended to provide a means to implement modifications to an ML-DSF that generally would otherwise require additional marketing submissions prior to implementation. This guidance outlines the FDA’s recommendations and best practices on marketing submission content for a PCCP.
Elements and Content of a Predetermined Change Control Plan for AI
It is important to note that a PCCP for device modifications should be included in the original premarket submission for that device; be it a 510(k) Premarket Notification, a De Novo Classification Request, or a Premarket approval (PMA). It should have its own standalone section within the submission application and be identified in the Table of Contents as “Predetermined Change Control Plan”. Additionally, it should be discussed in the submission cover letter, device description, labeling, and any other relevant sections. Once cleared/granted/approved, the PCCP will fall under the technological characteristics of the device.
Three essential components of a Predetermined Change Control Plan for AI:
A Description of Modifications, which includes a detailed description of the specific, planned device modifications. This section should list the individual proposed device modifications to be included in the PCCP as well as a rationale for each change. Manufacturers should keep in mind that each listed modification should be able to be verified and validated within their existing quality system. This section should clearly state whether modifications will be implemented automatically or manually, and if they will be implemented uniformly across all devices on the market or differently on different devices on the market. Finally, any proposed modifications should maintain the device within its original intended use and indications for use. Otherwise, a change in intended use or indications for use would trigger the need for a new marketing submission.
A Modification Protocol, which is the associated methodology to develop, validate, and implement those modifications in a manner that ensures the continued safety and effectiveness of the device across relevant patient populations. While the Description of Modifications section provides a list of the proposed modifications, the Modification Protocol details how those changes will actually be implemented. There are four essential buckets of information that this section should include:
- Data Management, in which the manufacturer describes how new data will be collected, annotated, curated, processed, stored, retained, controlled, and used to support each modification.
- Re-training Practices, which should identify the objective of the re-training process, provide a description of the ML model, identify the device components that may be modified, outline the practices that will be followed, and identify any triggers for re-training.
- Performance Evaluation, which should describe the specific processes that will be followed to verify and validate that any proposed changes meet specifications identified as part of that change. Verification and validation plans for individual modifications as well as in aggregate, to cover all implemented modifications, should be captured.
- Update Procedures, where the manufacturer describes how they will update their devices to implement any proposed changes, transparency and communication to users regarding labeling, post-market monitoring, and user training (if applicable)
An Impact Assessment, which provides an assessment of the benefits and risks of the planned modifications and risk mitigations. This should be performed within the context of the manufacturer’s existing quality system. Manufacturers should consider how the individual modifications proposed per the PCCP impact not only specific device software functions, but also the overall device functionality, other software functions, and device hardware (if applicable).
Conclusion
The FDA recognizes that there is a pressing need to support the iterative improvements of AI/ML-enabled devices. Having to submit an entirely new premarket submission as a result of an improvement to a software function would create a huge burden not only for manufacturers, but also for the agency. The implementation of a PCCP is a reasonable approach for avoiding this burden. It will allow for more flexibility for manufacturers to introduce post-market changes, ultimately expediting device improvements and benefiting patients.
The FDA is accepting public comments on this draft guidance until July 3, 2023 [2].
REFERENCES
2. https://www.regulations.gov/docket/FDA-2022-D-2628
Disclaimer - This post intended for informational purposes and does not constitute legal information or advice. The materials are provided in consultation with US federal law and may not encompass state or local law.
