3 Important Things to Know about the FDA Predetermined Change Control Plan (PCCP) for ML/AI in medical devices

Software development is in itself an iterative process. Particularly, medical software incorporating AI/ML (machine learning-enabled device software functions or ML-DSF) has become an integral part of many medical devices.

To ensure the safety and effectiveness of such devices and support iterative improvement through ML-DSF modifications, the FDA has recently issued a draft guidance containing a recommendation for a Predetermined Change Control Plan (PCCP) for Artificial Intelligence/ Machine Learning (AI/ML).

Here are the most important things to know about the FDA Predetermined Change Control Plan (PCCP).

 

In general, the FDA requires manufacturers to submit a premarket approval supplement, De Novo submission, or a new premarket notification if ML-DSF modifications affect the intended use or the safety and effectiveness of the device. To circumvent this, the new guidance proposes the submission of a PCCP together with the original submission. A PCCP can be considered for ML-DSFs for which modifications to the ML model are implemented either automatically or manually.

 

 

By including a PCCP in a marketing submission, manufacturers can proactively specify and get premarket authorization for the planned modifications they intend to perform on their devices. The PCCP is reviewed as part of a marketing submission through the appropriate pathway (510(k), PMA, De Novo) and it must be reviewed and approved prior to the manufacturer implementing any modifications listed under the PCCP (authorized PCCP).

 

An authorized PCCP is a technological characteristic of the authorized device for which it was established. An authorized PCCP should be followed to ensure device safety and effectiveness across patient populations. In instances when the PCCP cannot be followed or deviations occur, a new marketing submission for the change is required.

 

For previously authorized devices without a PCCP, a marketing submission must be submitted that includes the usual information and the PCCP. The FDA review will in this case focus on the aspects of the device that were significantly modified since the last submission.

 

 

The PCCP shall be a standalone section in the submission that is outlined in the table of contents. It should be addressed in the cover letter, as part of the device description, in the labeling, and in the appropriate sections that discuss substantial equivalence, safety, and effectiveness of the device. Any information throughout the submission belonging to the PCCP must be referenced in the PCCP section itself.

 

In public-facing documents such as the submission summary, details of the PCCP should be included in enough detail to ensure transparency.

 

Importantly, the labeling of the device should clearly state that said device incorporates machine learning and has a PCCP, in order for users to be aware that they need to perform updates that might modify the device's performance, inputs, or use.

 

 

In general, the PCCP needs to be evaluated within the existing risk management framework of the device and implemented in accordance with the manufacturer’s quality system.

 

The guidance provides ample descriptions, examples, and how-to information on structuring the PCCP. In general, three sections are required/recommended:

 

Description of the modifications – detailed descriptions of each planned modification to the ML-DSF. Modifications should be intended to maintain or improve the safety and effectiveness of the device. The recommendation is to include a limited number of changes that are specific and can be validated and verified. Each modification shall be traceable to a specific performance evaluation activity in the modification protocol section.

 

Modification protocol – step-by-step descriptions of the methods that will be used during the development, verification, validation, and implementation of the proposed changes. The modification protocol shall identify test methods, analyses, and acceptance criteria for all proposed changes in order to ensure the safety and effectiveness of the device. The changes shall be documented in compliance with the quality system of the manufacturer. Broadly, the modification protocol shall discuss data management, re-training practices, performance evaluation, and updated procedures.

 

Impact assessment – the benefits, risks, and mitigation of the risks of implementing the changes outlined in the PCCP shall be thoroughly discussed. The impact of individual changes shall be addressed alongside a discussion of whether the combination of proposed modifications introduce additional unmitigated risks. The impact of the modifications on other software functions as well as on the device hardware shall also be considered in the impact assessment.

For some devices, it might be easier to include the impact assessment with the modification protocol.

 

While preparing the PCCP, it is important to refer to the original content of the FDA guidance and to use the Q-Submission process to receive FDA feedback on your PCCP before submitting your marketing application.

 

If you need help with preparation of your PCCP, please contact us at: info@quaregia.com. We have 20 years of experience with medical device software in all its facets.


Last updated 2023-04-18