Medical devices containing Artificial intelligence (AI) and Machine Learning (ML) technologies inherently require updates as the software self-trains on the data generated internally, improving their performance over time which, in turn, triggers the threshold of a new 510k or a PMA amendment creating regulatory burden for the manufacturer. A new FDA guidance document provides a solution to this conundrum.
AI technology is designed to solve problems, predict outcomes, or make decisions based on its specific algorithm and its ML component uses internal data to improve its performance over time. Unlike traditional systems that follow fixed rules, AI and ML can learn and adapt by analyzing new data, making them particularly useful for tackling complex problems. This ability has made AI/ML devices very popular to generate systems that mimic and expand on human capabilities. For example, an AI/ML-powered device might identify patterns in medical scans that a human doctor could miss, potentially saving lives by diagnosing illnesses sooner.
However, any improvement that changes the safety and effectiveness profile of a device requires regulatory submissions. The new FDA guidance document released this week provides a solution to avoid unnecessary submissions. The FDA understands that AI-powered devices often need updates as they improve over time. These updates might involve changes to how the AI model works or new features that enhance the device. To handle this, the FDA allows manufacturers to include a “Predetermined Change Control Plan” (PCCP) in their marketing submissions. A PCCP outlines how the device will be updated while ensuring it remains safe and effective.
A PCCP generally includes three main components:
- Description of Modifications: A detailed explanation of the specific updates planned for the device.
- Modification Protocol: A method for testing and implementing the changes, ensuring they don’t compromise safety or effectiveness.
- Impact Assessment: An analysis of the risks and benefits of the changes, along with any strategies to reduce risks.
By following these steps, manufacturers can make improvements without needing to submit new applications for every update. This process is especially important for AI models that rely on continuous learning, where the software automatically updates itself, or for systems that require manual changes with human involvement. These guidelines allow devices to stay up-to-date while maintaining high safety standards.
The FDA encourages manufacturers to engage with their review teams early in the development process, especially for high-risk devices like life-supporting or implantable technologies. This early interaction ensures that the planned updates are appropriate and meet regulatory standards. Manufacturers can use the FDA’s Q-Submission Program to get feedback on their PCCP before submitting a full marketing application.
Although the guidance focuses on the PCCP, it doesn’t replace other regulatory requirements for medical devices. Manufacturers still need to provide all necessary evidence to show that their devices are safe, effective, and meet FDA standards. These recommendations make it easier for companies to bring innovative AI-powered devices to market while protecting patients.
The FDA’s new guidance document for AI/ML devices strikes a balance between fostering innovation and ensuring safety. By allowing manufacturers to plan for updates in advance, the FDA supports the development of cutting-edge healthcare tools that can evolve with new data and improve patient outcomes over time.
