This audio is automatically generated. Please let us know if you have any comments.
The Food and Drug Administration released a much-anticipated update on Tuesday detailing how it plans to regulate changes to artificial intelligence-based medical devices after the agency authorizes a product.
The agency shared its final guidance on what information it wants from manufacturers who propose predetermined change control plans (PCCPs), a new type of regulatory framework that allows certain modifications to a device after it has been placing on the market.
The FDA first introduced the concept of PCCP in a 2019 Discussion Paper on changes to devices based on AI or machine learning. Currently, devices on the market that use AI or machine learning are “locked,” meaning the algorithm cannot adapt or change over time.
PCCPs are intended to address this problem by enabling performance improvements through iterative changes, with manufacturers sharing a framework for proposed device modifications as part of premarket submissions to the FDA.
The agency outlined PCCPs for AI-enabled devices in a 2023 document. draft guidelines and extended the concept to other types of medical devices, such as in vitro diagnostics, in a August Draft Guidelines.
The final guidance is largely consistent with the 2023 draft, but includes a new section on version control and maintenance. The FDA explained that a PCCP may require revisions before the agency can determine its safety and effectiveness. The FDA may request additional information through an interactive review or deficiency letters, but if deficiencies remain unresolved, the agency could authorize the device after PCCP withdrawal.
So far, the FDA has not authorized any AI-enabled adaptive devices, said Jessica Paulsen, associate director for digital health in the FDA’s Office of Product Quality and Evaluation, during a press conference. Digital Health Advisory Committee Meeting in November. However, manufacturers have submitted PCCPs and the FDA has authorized them for devices following the 510(k), de novo, and premarket approval pathways.
Key elements of a PCCP
PCCPs could give device makers more leeway to update AI-enabled devices, but these changes must remain within certain limits. Any modifications must comply with the PCCP authorized by the FDA and must be within the intended use of the device.
A PCCP must include three elements: a description of the planned changes, a modification protocol explaining the criteria by which manufacturers will determine that a device remains safe and effective with the planned changes, and an impact assessment that also details the benefits and the risks of PCCP. as risk mitigation plans.
Any planned changes must be specific and must be capable of verification and validation, the FDA said. Manufacturers should clarify whether they plan to implement the changes automatically or manually. They must also indicate whether the planned changes will be implemented uniformly across all devices in the market or implemented differently, based on the unique characteristics of a clinical site or individual patients.
The device labeling should explain that it incorporates machine learning and has an authorized PCCP, so users are aware that they may need to perform software updates and that the Device performance may change.
One example shared by the FDA is software used in an intensive care unit that extracts electrocardiogram, blood pressure, and pulse oximetry data to detect when a patient is at the beginning of physiological instability and issue an alarm. A manufacturer could offer to retrain the AI model with more data to reduce the false alarm rate while maintaining or increasing its sensitivity.
Under a PCCP, that manufacturer could collect data, retrain the AI model, and test it to demonstrate that false alarms were significantly reduced. The company would update the device label and communicate the changes to users. Since these changes would be made in accordance with the PCCP, the manufacturer would not need to submit a new trade presentation.
However, changes to the AI model to allow it to predict physiological instability before its onset, which the previous version could not do, would fall outside the scope of a PCCP and require a new marketing application .
Changes to users of a device are also outside the scope of a PCCP.
For example, an algorithm that analyzes images of skin lesions to help primary care providers make a diagnosis could be updated to work with other types of cameras as part of a PCCP, the FDA said. However, a company will need to submit a new pre-market application to modify the algorithm to be patient-oriented.