Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning

Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning

On April 2 the Food and Drug Administration (FDA) released a discussion paper and request for feedback from stakeholders by June 3 on a proposed regulatory framework for modifications to artificial intelligence/machine learning (AI/ML) -based software as a medical device (SaMD). SaMD is software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device. Under the Federal Food, Drug & Cosmetic Act, FDA considers medical purpose as those purposes that are intended to treat, diagnose, cure, mitigate, or prevent disease or other conditions. The foundation of this paper builds on the fact that AI, specifically ML, are techniques used to design and train software algorithms to learn from and act on data. The proposed regulatory framework speaks to the reality that adaptive AI/ML technologies are constantly evolving, and have the potential to adapt and optimize device performance in real-time to continuously improve healthcare for patients.

The underlying issue raised is, what happens when these medical devices are continuously learning to the extent that there is a modification that may require a premarket submission for an algorithm change? This inquiry prompted the FDA to reimagine an approach to premarket review for AI/ML based SaMD, while allowing the software to continue to learn and evolve over time to advance and improve patient care without the delay of traditional means of medical device regulation that are misaligned and are slow to keep up with the speed of developing adaptive AI/ML technologies. The FDA points out the need for a new total product lifecycle (TPLC) regulatory approach that facilitates a rapid cycle of product improvement and allows these devices to continually improve while providing effective safeguards. While speaking to the quality systems and good machine learning practices expected of every medical device manufacturer, the FDA additionally proposes a framework for modifications to AI/ML-based SaMD that build on existing principles related to Medical device risk safeguards as well as describes an innovative approach that may require additional statutory authority to implement completely.

The overarching objective set out by this proposal embraces improvement of AI/ML-SaMD with modernized FDA oversight that acknowledges and addresses rapid changes in innovation. FDA envisions this will occur while assuring that patient safety is maintained, and furthermore that ongoing algorithm changes are implemented according to pre-specified performance objectives, defined protocols, and validation processes consistently focused on improving performance, safety and effectiveness of AI/ML software.

FDA anticipates that many modifications to AI/ML-based SaMD involve algorithm architecture modifications and re-training with new data sets, which under the existing software modifications guidance would be subject to premarket review. The TPLC regulatory review for AI/ML-Based SaMD enables the evaluation and monitoring of a software product from its premarket development to postmarket performance, along with continued demonstration of the organization’s excellence. FDA’s TPLC approach is based on a series of principles that balance the benefits and risks, and provide access to safe and effective AI/ML-based SaMD.

As learning, adaptation, and optimization are inherent to AI/ML-based SaMD, these capabilities would be considered modifications to SaMD after they have received market authorization from FDA. FDA proposes a “Predetermined change control plan” to refine SaMD Pre-Specifications (SPS) or “Algorithm Change Protocol” (ACP), based on the real-world learning and training for the same intended use of AI/ML SaMD model. This would streamline the processes required of modifications that may not necessitate another FDA premarket review or focused FDA review of SPS and ACP. The paper discusses that this approach does not come without scope and limitations to the types of changes that fall under the pre-specified and managed category.

As a final note, FDA notes a commitment to transparency could be achieved through a variety of mechanisms. In regard to modifications in the SPS and ACP, manufacturers would ensure that labeling changes accurately and completely describe the modification, including its rationale, any change in inputs, and the updated performance of the SaMD.

Published on