What the FDA’s Draft Guidance for AI-Enabled Medical Devices Means

The FDA’s new draft guidance for developers of AI-enabled medical devices marks a critical milestone in addressing the complexities of transparency, bias and safety throughout the lifecycle of these advanced technologies.

Tailored for manufacturers and stakeholders, it outlines strategies to ensure devices meet high standards of efficacy and inclusivity. Learnings from more than 1,000 FDA-authorized AI-enabled devices have informed this guidance to better support developers.

A recent scoping review highlighted critical gaps in AI device approvals: only 3.6% reported race or ethnicity data, fewer than 10% included socioeconomic details and just 1.9% linked to scientific validations. Such omissions hinder transparency, restrict performance evaluations and risk exacerbating health disparities among underrepresented populations.

This announcement aligns with the FDA’s release of the draft guidance for AI in drug and biological product development, signaling a unified effort across healthcare.

Fundamentally, the guidance brings forth the importance of a Total Product Lifecycle (TPLC) approach, a comprehensive strategy that considers every stage of a device’s journey — design, development, market approval and real-world use — to ensure long-term safety and effectiveness.

The guidance further highlights the need to address the unique challenges posed by AI models, including their reliance on complex algorithms and the potential for opaque decision-making processes, ensuring that devices perform consistently and reliably.

XTALKS WEBINAR: AI in the Workplace: How to Reskill and Train Your Team

Live and On-Demand: Tuesday, January 14, 2025, at 1pm EST (10am PST)

Register for this free webinar to learn tactics for training employees to utilize AI tools efficiently and more.

Key elements of the guidance focus on transparency and bias. Developers are urged to document evidence that their devices perform consistently across diverse demographic groups. Addressing “AI bias,” where skewed outcomes arise from unrepresentative training data, is critical to reducing disparities in AI performance. The guidance also stresses clear documentation in marketing submissions, requiring detailed descriptions of device inputs, outputs and workflows, empowering both regulatory evaluation and user understanding.

Another key area is data management. The FDA encourages developers to use diverse, high-quality datasets for training and testing their AI systems. Sourcing data from multiple clinical sites helps reduce the risk of “data drift,” where differences between training data and real-world data could impact performance.

Notably, this guidance complements the recently finalized guidance on Predetermined Change Control Plans (PCCPs), which focuses on proactive planning for post-market updates to AI-enabled devices. Together, these guidances provide a cohesive framework, uniting principles of lifecycle safety and post-market adaptability.

The guidance for AI-enabled medical devices highlights that failing to meet expectations for transparency or risk management — such as not proving consistent performance across patient groups — could lead to delays or additional scrutiny during the approval process. By addressing these issues, the guidance ensures AI innovations are safe, effective and equitable.

To support stakeholders, the FDA will host a webinar on February 18, 2025, offering insights and addressing questions on this important initiative, enabling a step toward in fostering trust and inclusivity in AI-enabled medical technologies.