The FDA, Health Canada, and the British MHRA keep paving the way for continuous learning systems for medical devices.
In April this year, the FDA had published a draft guidance on Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence/Machine Learning (AI/ML)-Enabled Device Software Functions.
A predetermined change control plan, PCCP, is a description of anticipated modifications (through retraining), and how they will be managed and revalidated. Once a PCCP is accepted as part of the pre-market assurance, modifications within its scope can be made without the involvement of the FDA.
Last week, the FDA, Health Canada, and MHRA have published five key guiding principles for Predetermined Change Control Plans. Though this publication is not a major milestone for AI/ML regulation, it does open the door for continuous learning in medical devices a little more, and it demonstrates the FDA's intentions and level of activity in this field.
The published guiding principles for PCCPs are as follows:
1. Focused and Bounded
The PCCP clearly defines the changes a manufacturer intends to make, limiting alterations to the original purpose of the MLMD.
It outlines the scope of proposed changes, methods for verifying and validating these changes, and steps to halt or revert changes that don't meet the set performance standards.
Promotes a risk-managed approach in the design and execution of a PCCP, ensuring risk assessment throughout the product’s lifecycle.
Highlights the need for evidence collected over the device's lifecycle to ensure ongoing safety and effectiveness, showing that benefits outweigh associated risks, and confirming that risks are well-managed and controlled.
Advocates for clear and detailed plans to maintain transparency with users and other stakeholders, enabling them to stay informed about the device’s performance before and after changes are made.
5. Lifecycle Perspective
Encourages the creation and use of a PCCP from a lifecycle viewpoint, which includes continually considering the perspectives of all stakeholders as well as risk management practices to ensure device safety.