Industry groups call for changes to FDA’s guidance on managing AI-enabled medical devices

Industry stakeholders are asking the US Food and Drug Administration (FDA) to revise its draft guidance on the development and management of AI-enabled medical devices. They are seeking a more tailored approach that considers the different types of devices that utilize AI, rather than applying a one-size-fits-all strategy. As it currently stands, the guidance focuses on complex AI-enabled devices but does not address simpler systems.

The two major medical device industry associations, the Advanced Medical Technology Association (AdvaMed) and the Medical Device Manufacturers Association (MDMA), also urged FDA to align this document with its other guidances on AI-enabled medical devices.

These comments were in response to the FDA’s draft guidance on managing the lifecycle of AI-enabled device software functions. The document outlines a set of recommendations on the contents of marketing submissions for devices that include AI-enabled device software functions. (RELATED: FDA issues draft guidance on developing and managing AI-enabled devices, Regulatory Focus 6 January 2025)

FDA officials have said the agency has authorized over 1,000 AI-enabled devices through established premarket pathways.

The agency received 46 comments in response to the draft guidance. The deadline for commenting was 7 April.

AdvaMed and MDMA recommended a comprehensive revision of the document. MDMA stated that “due to the significant concerns with this draft guidance, we recommend issuing a revised draft guidance that reflects the feedback included in this comment letter. The revised draft guidance would be subject to an additional comment period before being finalized.”

Guidance tailored to complex devices

AdvaMed told FDA to avoid a “one-size-fits-all approach” to AI regulatory recommendations; the group said that different frameworks are needed depending on the type of AI system.

The group states that “AI-enabled medical devices vary widely in complexity and risk, and regulatory expectations should reflect these differences. A rigid, one-size-fits-all approach could impose excessive burdens and confusing expectations for lower-risk AI applications and simpler technologies while failing to provide meaningful oversight where it is most needed.”

The group suggests revising the scope of the guidance to concentrate specifically on deep learning or deep neural networks, as the draft recommendations already addresses these systems. “Many of the recommendations in the AI Lifecycle draft guidance are not relevant for less complex AI-machine learning (ML) models,” AdvaMed wrote.

MDMA agreed and stated, “The scope of the document should be narrowed to address more complex AI, in accordance with FDA’s least burdensome principles and to ensure a risk-based approach. The recommendations in this guidance are most relevant and applicable to devices that incorporate the subset of AI technologies known as deep learning or deep neural networks … Many of the recommendations in this guidance may be less relevant to other less complex machine learning models.”

Both AdvaMed and MDMA stated that the FDA's existing guidance titled “Content of Premarket Software Submissions for Device Software Functions” is more suitable for simpler AI models that utilize linear regression and decision trees.

Leverage existing regulatory frameworks

Both groups requested that the FDA align this document with its other guidance on AI.

MDMA said that “the draft guidance presents information that overlaps with current FDA guidance documents … For example, topics such as labeling, cybersecurity, software validation, postmarket monitoring, and human factors engineering/user interface are covered in existing guidance materials, leading to a duplication of requirements. We suggest eliminating the redundant information from this guidance. For example, the cybersecurity section of the draft guidance is redundant with FDA’s cybersecurity guidance and contains definitions and concepts that are currently being developed in consensus standards.”

“The draft guidance does not fully acknowledge the existing regulatory frameworks that already govern key aspects of AI oversight such as risk management and quality system regulations that govern the total product lifecycle (TPLC), from design and validation through postmarket monitoring. Better alignment of AI oversight with these established systems would enhance regulatory efficiency without compromising patient safety,” AdvaMed stated.

Draft does not address product lifecycle

Various other organizations raised concerns about the guidance. The Combination Products Coalition (CPC) indicated that the guidance omits important information regarding the product lifecycle.

CPC recommends that the final guidance “clarify which portions of the document are applicable to the internal processes device manufacturers must comply with as part of the TPLC approach as it is currently not clear.”

The coalition also suggested the agency take a more nuanced approach in its submission requirements for combination products such as how it applies the draft guidance’s marketing submission content recommendations to device-led combination products and biologic-led combination products.

The American College of Radiology (ACR) recommended that the guidance address post-market monitoring requirements.

The group states that “FDA's draft guidance encourages, but does not require, post-market monitoring as a TPLC risk control … Specific post-market monitoring mechanisms, such as periodic reporting or third-party audits, are not clearly outlined.”

The American Medical Association (AMA) provided largely positive feedback on the guidance, stating that it introduces needed transparency for physicians using AI medical devices.

The group stated that the guidance “proposes much-needed recommendations on critical topics for physician use of AI, such as enhanced transparency through updated labeling recommendations, guidelines for communication of performance validation information, among others. We urge the FDA to move towards finalization of this guidance.

AI comments

Industry groups call for changes to FDA’s guidance on managing AI-enabled medical devices

Related topics

Recommended content

Welcome to the new RAPS Digital Experience