Is a Predetermined Change Control Plan (PCCP) Right For Your AI-Enabled Device?
Purpose
If you are a medical device manufacturer looking to bring an AI-enabled medical device to the US market, you probably have many questions as this is a much uncharted territory. One such question may be whether a Predetermined Change Control Plan (PCCP) would be a beneficial part of your regulatory strategy. If that is a question you have, read on.
Background
The draft FDA guidance on Predetermined Change Control Plans for Machine Learning-Enabled Medical Devices: Guiding Principles was released April 2023. The Predetermined Change Control Plan outlines the types of changes that may be made to the device, as well as the processes and procedures that must be followed to ensure that these changes are implemented in a safe and effective manner. This includes identifying potential risks associated with the changes, assessing the impact on the device's performance and safety, and determining the need for additional testing or validation.
By having a Predetermined Change Control Plan in place, manufacturers can demonstrate to the FDA that they have thoughtfully considered the potential impact of changes to their devices and have put in place measures to mitigate any risks. This can help streamline the regulatory review process and ensure that AI enabled devices continue to meet the FDA's safety and effectiveness standards.
Overall, the FDA guidance for Predetermined Change Control Plan for AI enabled devices provides manufacturers with a roadmap for managing changes to their devices in a systematic and transparent manner, ultimately helping to ensure the safety and effectiveness of these innovative technologies.
Assess Roadmap
PCCPs are designed for business cases where there are a set of pre-planned changes defined. Therefore, the first factor to consider is whether or not you have an adequately-defined roadmap. In this assessment, there are two important factors to consider: Definition Detail and Timeline.
Definition Detail
The guidance lays out criteria for the level of detail that is required in order to get a PCCP approved. The "Description of Modifications" section of a Predetermined Change Control Plan (PCCP) for AI/ML-enabled medical device software functions has specific requirements according to the FDA draft guidance. Here's an expanded overview based on the document:
Detailed Descriptions: Each planned modification must be described in detail, including how changes to device characteristics and performance will result from the implementation of the modifications.
Modifications should be clearly documented and justified within the PCCP, linking them to specific performance evaluation activities as detailed in the Modification Protocol section.
Rationale for Changes: The specific reasons for each change to the device should be articulated, alongside the expected impacts on the device's safety and effectiveness.
Modifications that are appropriate for a PCCP include those that are intended to maintain or improve the safety or effectiveness of the device.
Examples of acceptable types of modifications can include:
Changes related to quantitative measures of performance,
Modifications to the inputs of the device, or
Adaptations for specific subpopulations, provided they can be verified and validated within the existing quality system.
Automatic vs. Manual: The document should clarify whether modifications are deployed automatically by software or manually, which involves human input. This distinction affects how the FDA evaluates the modifications.
Global vs. Local: The FDA would like to know whether the proposed modifications will be deployed to all devices uniformly or there will be a deployment variation across devices in the market (e.g. certain geographies, certain unique user groups)
Limited Number of Modifications: The FDA recommends including only a limited number of specific modifications in a PCCP that can be verified and validated.
Known Modification Protocol: Verification and validation as well as implementation plans must be pre-defined. More concretely, the following four components should be known for the Modification Protocol:
Verification and Validation Plan:
Data management practices,
Re-training practices,
Performance evaluation protocols. Acceptance criteria for the V&V must be known and approved by the FDA. Each modification should be linked to a specific performance evaluation activity within the Modification Protocol.
Implementation Plan:
Update procedures, including communication and transparency to users and real-world monitoring plans
Note that upon reviewing a PCCP, the FDA may determine that some, but not all, modifications meet the criteria for inclusion in the authorized PCCP. Only those modifications that are found to be substantially equivalent or ensure safety and effectiveness will be included. This is why a pre-sub meeting to discuss proposed changes with the FDA is highly recommended.
Once approved, modifications should remain within a "range of FDA-authorized specifications" that were established during the initial premarket approval. This ensures that modifications, once verified and validated, can be implemented without requiring a new marketing submission.
Timeline
The longer your defined roadmap, the more value there is in a PCCP. While the exact break-even point varies on a case-by-case basis, a multi-year roadmap is much more likely to benefit from a PCCP than anything shorter.
Assess Alternate Regulatory Pathways
The next step will be to understand the alternate regulatory pathways available to you if you were to not include a PCCP. In particular when looking at your roadmap:
Do the proposed modifications to AI-enabled features fall under device software functions as defined in Content of Premarket Submissions for Device Software Functions Guidance?
If an AI-enabled function does not qualify as a Device Software Function, it is out of scope of the medical device and therefore the PCCP. An example of such a feature may be using deep learning to provide logistical decision support around an operating room to hospital administration.
Will your changes rise in significance to the level that requires another submission as opposed to being handled through internal documentation per the FDA guidance?
Will your changes fit within a special 510(k)? While traditional 510(k)s take an average 140 days for a final decision, Special 510(k)s take a maximum of 90 days to reach a final decision.
Is there a predicate with PCCP already listed as special controls? If not, will use of PCCP impact your regulatory pathway at all? For example will your initial clearance require a De Novo submission as opposed to a traditional 510(k) based on this factor? On average, De Novo submissions take 338 number of days and require more resources.
Consider competitive landscape and business strategies. A PCCP may not be as beneficial in highly competitive landscapes where long term roadmaps change based on customer and market needs.
It may also be useful to consider whether your device will qualify for a Breakthrough Device designation and/or benefit from enrolling in the FDA’s Total Product Life Cycle Advisory Program (TAP) as that would impact timelines for the scenarios outlined above.
Additionally, it is beneficial to utilize the pre-sub process to ensure that the FDA is aligned with your assessment of all regulatory pathways.
Determine the Least Burdensome Approach
Once you have a defined roadmap and identified alternate regulatory pathways the decision on whether or not a PCCP is the right approach for you comes down to determining the least burdensome approach for your business and your company. In particular, the overall impact on resources for your business.
Assess Resourcing Impact
It is important to understand that the PCCP provides a trade-off between pre-commercialization vs. post-commercialization resource needs of design changes. An organization invests more pre-market resources into the initial clearance of the device to then use less resources post-commercialization in bringing device revisions to market. How this trade-off looks in exact and absolute terms will vary depending on the business and must be evaluated on a case-by-case basis.
Conclusion
As a manufacturer of an AI-enabled medical device, it is important to perform a business-specific assessment of the value a PCCP may bring to your regulatory strategy. Depending on the existence and nature of your device’s roadmap, existence of predicate devices and availability of alternate regulatory pathways, as well as the specific circumstances surrounding your business’s resources, a PCCP may or may not be a part of the most optimal path to market for your device.
References
FDA Guidance: Predetermined Change Control Plans for Machine Learning-Enabled Medical Devices: Guiding Principles. October 2023
Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions. Draft Guidance for Industry and Food and Drug Administration Staff. April 3, 2023.
FDA Guidance: Deciding When to Submit a 510(k) for a Software Change to an Existing Device. October 2017.
Aboy, M., Crespo, C. & Stern, A. Beyond the 510(k): The regulation of novel moderate-risk medical devices, intellectual property considerations, and innovation incentives in the FDA’s De Novo pathway. npj Digit. Med. 7, 29 (2024). https://doi.org/10.1038/s41746-024-01021-y
FDA Guidance: Content of Premarket Submissions for Device Software Functions. June 2023.
