FDA is pushing out new guidance documents at a pace rarely seen, likely caused by the upcoming change in administration.  Today, we are sharing with you our thoughts on two of FDA’s guidance documents, but we have a few more coming.  We’ll need to keep an eye on whether the Trump administration retains or makes changes to any of these last-minute guidance documents. 

The two guidance documents we will discuss today are: 

• Communications from Firms to Health Care Providers Regarding Scientific Information on Unapproved Uses of Approved/Cleared Medical Products; and 
• Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations

We briefly summarize below the key takeaways for each guidance document. 

Scientific Information on Unapproved/Uncleared Uses 

FDAs’ Communications from Firms to Health Care Providers Regarding Scientific Information on Unapproved/Uncleared Uses Guidance (“new SIUU guidance”) revises and finalizes the agency’s October 2023 draft guidance (“2023 draft guidance”) setting forth FDA’s enforcement policy regarding certain firm-initiated communications of scientific information on unapproved/uncleared uses (“SIUU communications”) of approved/cleared medical products to health care providers (HCPs). In many ways, the new guidance mirrors the recommendations in the precursor 2023 draft guidance.  For instance, the new SIUU guidance reaffirms FDA’s position that a firm will not be creating a new intended use for a product based solely on a SIUU communication made in a manner consistent with the recommendations in the new guidance.  The new SIUU guidance also reifies the key tenet that firm initiated SIUU communications must be truthful, non-misleading, and provide appropriate context/disclosures to allow HCPs to evaluate strengths and weaknesses of the scientific information being shared with them. 

Despite FDA retaining much of the content from the 2023 draft guidance, the new SIUU guidance does include some notable additions.  For example, in the section discussing presentational considerations that firms should take into account for SIUU communications, FDA includes new, specific examples to further illustrate these recommendations in practice.  The new SIUU guidance also fleshes out key factors for firms to consider when assessing whether a source publication is appropriate to be included as an SIUU communication subject to FDA’s enforcement discretion, including providing additional insight for assessing whether a study is considered scientifically sound for particular products.  As another example, FDA significantly expands upon its recommendations discussing firm-generated presentations of SIUU communications that are provided with source publications.  This includes calling out specific marketing tactics that would place firm-generated presentations outside the scope of the enforcement discretion policy under the new SIUU guidance.  FDA also takes the position that “calls to value that pre-judge the benefits of the medical product for individual patients” would place firm-generated presentations outside the protective sphere of enforcement policy in the new guidance.   

Accordingly, drug and device companies should make sure they understand the requirements that must be satisfied when making SIUU communications to HCPs in order to implement safeguards to mitigate the risk that such communications fall outside the ambit of FDA’s enforcement discretion policy. 

Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations

FDA’s guidance focuses on the Total Product Life Cycle approach to marketing submissions, emphasizing manufacturers' need to integrate safety, effectiveness, and quality throughout the lifecycle of products to ensure they remain reliable and relevant as technology evolves - a key concern in AI devices that incorporate machine learning. According to FDA, manufacturers should include detailed documentation when submitting marketing applications for AI-enabled devices including:

Device Description Details:

• Explanation of how AI is integrated into the device's functionality
• Input/output details, calibration procedures, and operational workflows
• Description of intended users, environments, and automation levels

User Interface Details:

• Descriptions and graphics illustrating user-device interactions
• Clear labeling, including information about AI’s role, performance metrics, and limitations
• Recommendations for transparency to ensure usability and mitigate risks of misinterpretation

Risk Management:

• Comprehensive risk assessments that identify and mitigate potential hazards across the product lifecycle
• Emphasis on addressing risks specific to AI, such as biases in data, misaligned user expectations, or misinterpretation of outputs

Data Management:

• Detailed accounts of training, validation, and testing datasets
• Measures to ensure datasets are representative of intended populations and free from biases
• Explanation of data collection and preparation processes

Performance Validation:

• Documentation of performance metrics, including sensitivity, specificity, and predictive values
• Subgroup analysis to verify consistent performance across different demographic or clinical groups.

FDA also asks manufacturers to consider several key areas when preparing to market these products, including transparency around how the AI models function and their limitations, strategies to mitigate bias, and performance monitoring strategies that will be utilized to track device performance in the real-world following marketing approval. 

Importantly, FDA confirmed that manufacturers can predefine updates or improvements to AI models without requiring additional FDA review, as long as these changes align with the approved Predetermined Change Control Plan (“PCCP”). In December 2024, FDA addressed the PCCP framework for AI-enabled medical devices in a guidance document (draft guidance issued in April 2023), and outlined a general overview of the framework that is designed to limit the need for new marketing submissions with updates to a device. In the guidance, FDA recommended that PCCPs include a description of the modifications that are anticipated, protocol that outlines the validation, verification and safety criteria for the modifications, and an impact assessment that evaluates risks and benefits. PCCPs are submitted as part of a device’s marketing authorization and any modifications that are not covered under a PCCP require new submissions and re-evaluation. 

In the updated December 2024 guidance, FDA largely maintained the framework that it outlined in the April 2023 guidance.  However, the guidance included certain notable updates. For example, FDA expanded the scope of the guidance from artificial intelligence/machine-learning-enabled devices to all artificial intelligence-enabled devices, while adopting the below definitions. 

Artificial Intelligence (AI): A machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments. Artificial intelligence systems use machine- and human-based inputs to perceive real and virtual environments; abstract such perceptions into models through analysis in an automated manner; and use model inference to formulate options for information or action.

Machine Learning (ML): A set of techniques that can be used to train AI algorithms to improve performance at a task based on data.

It appears that this expansion reflects FDA’s understanding that the principles in the guidance are applicable to all artificial intelligence-enabled devices and their software functions, and that they do not need to be limited to machine learning-enabled devices. 

Importantly, FDA recognized in the guidance that certain modifications to the indications – although not to the intended use – may be acceptable.  See page 22 of the guidance (“As with modifications to the intended use, FDA believes that most modifications to the indications for use included in a PCCP would be difficult for FDA to assess prospectively to determine whether the device would remain safe and effective. However, there may be certain modifications to the indications for use (e.g., certain changes in the indications for use to specify use of the device with an additional device or component) that may be appropriate for inclusion in a PCCP”).  FDA then recommends that manufacturers seek individual meetings through the Q-Submission program with FDA to receive detailed guidance.  While this flexibility is welcome, one area of concern is that this may encourage further dispersion and lack of consistency with regard to the standards for the data that must be included in submissions to FDA. Indeed in 2021, a STAT News investigation had discovered that submissions to FDA were inconsistent and divergent (at least according to the publicly available information) in terms of the data included in the submissions, for example, with many lacking information on racial makeup and gender breakdown. 

FDA’s recent guidance documents are welcome, and provide helpful guidance to the industry in terms of product development and the issues that the industry must consider.  Yet, it remains to be seen whether these principles will be continued by the new administration, and if not, what changes will be proposed and implemented by the new administration.  We will continue to monitor the developments and provide updates.