On January 7, 2025, the FDA published draft guidance for developers of AI-enabled medical devices to ensure safe and effective development and use of such devices throughout their lifecycles (“Draft Lifecycle Guidance”). This draft guidance constitutes the FDA’s most detailed view to date on this topic. The FDA will accept stakeholder feedback on the Draft Lifecycle Guidance until April 7, 2025. 

Context

Since 2018, the FDA has been executing on its Medical Device Safety Action Plan. A major theme of that plan is focusing the FDA’s safety assurance process on the type of medical device, rather than the device’s lifecycle stage.

The prior emphasis on specific lifecycle stages (such as premarket review and postmarket surveillance) was being challenged by AI-enabled medical devices, which can change meaningfully after their deployment, among other factors.

A major element of this regulatory transition is the Total Product Life Cycle (TPLC) approach, which involves reorganizing the FDA’s lifecycle-oriented teams (for example, for premarket review and postmarket surveillance) into teams organized by product type. The TPLC database similarly organizes information by medical device, bringing together data from separate FDA databases for Premarket Approvals, Premarket Notifications, De Novo Classification Requests, Adverse Event Reports, and Medical Device Recalls, respectively.

In April 2019, the FDA released a discussion paper on AI/ML-enabled Software as a Medical Device, which recommended the inclusion of Predetermined Change Control Plans (PCCPs) in premarket applications to help device manufacturers secure the FDA’s approval for predetermined changes. This would enable the implementation of these changes without a new premarket application to the FDA.

Building upon significant prior guidance related to AI-enabled medical devices, the FDA finalized its guidance on PCCPs in AI-enabled medical device software on December 4, 2024. The Draft Lifecycle Guidance released on January 7 complements this finalized guidance on PCCPs. 

Contents of the Draft Lifecycle Guidance

 The Draft Lifecycle Guidance includes recommendations that map to the lifecycle stages of AI-enabled medical devices:

• Development: Risk Assessment, Data Management, and Model Description and Development

• Validation: Data Management and Validation

• Description of the Final Device: – Device Description, Model Description and Development, User Interface and Labeling, Public Submission Summary

• Postmarket Management: Device Performance Monitoring and Cybersecurity

Two important topics that the Draft Lifecycle Guidance covers in depth are bias mitigation and transparency.

• Transparency: The guidance proposes the use of model cards (with two examples in the appendices) and usability evaluations (with further discussion in an appendix), to bolster transparency and user understanding.

• Bias Mitigation: The guidance highlights the importance of using high-quality and representative datasets, so that devices can be safely used across diverse demographic groups.

To help device manufacturers and software developers demonstrate their considerations of these factors, the guidance also outlines the FDA’s documentation expectations, including those related to risk assessments, device descriptions, user interfaces, and labeling, among other topics.

 Though the Draft Lifecycle Guidance attempts to describe the factors that device manufacturers and software developers should consider from the earliest stages of the development lifecycle, it also encourages early engagement with the FDA to help navigate the regulatory process.

Enzai is here to help

Enzai’s AI GRC platform can help your company deploy AI in accordance with best practices and emerging regulations, standards and frameworks, such as EU AI Act, the Colorado AI Act, the NIST AI RMF and ISO/IEC 42001. To learn more, get in touch here.