Among the wave of guidance documents issued by the U.S. Food and Drug Administration (“FDA” or the “Agency”) in the first week of 2025 were three notable draft guidance documents pertaining to medical devices (together, the “Draft Guidances”). The Draft Guidances hit on the topics of in vitro diagnostic (“IVD”) devices, artificial intelligence (“AI”) enabled device software functions, and pulse oximeters. This uncharacteristic deluge of guidance all within the span of a week illustrates the Agency’s desire to disseminate policy ahead of the incoming administration – especially as it relates to medical devices, which for a variety of reasons that any follower of this blog could intuit, have become a hot-button issue across the various corners of the healthcare and life sciences industries.
I. In Vitro Diagnostic Devices
On January 6, FDA released a draft guidance titled “Validation of Certain In Vitro Diagnostic Devices for Emerging Pathogens During a Section 564 Declared Emergency” (the “IVD Draft Guidance”).[1] This guidance aims to provide a framework for manufacturers to efficiently validate IVDs for emerging pathogens – part of FDA’s continuing effort to lay the groundwork for a timely and effective response to future public health emergencies. FDA is inviting comments to the Draft Guidance with a deadline set for March 7, 2025.
A. Background
The Food, Drug, and Cosmetic Act (“FD&C Act”) grants FDA authority to facilitate the availability and use of medical countermeasures (“MCMs”) to address chemical, biological, radiological, and nuclear threats to the nation’s public health.[2] This power is referred to as Emergency Use Authorization (“EUA”) and allows FDA to authorize the use of certain unapproved medical products if the Secretary of Health and Human Services (the “Secretary”) declares that justifying circumstances exist. FDA has used EUA to authorize emergency use of IVDs for eight infectious diseases over the years – most recently and notably, for COVID-19.
During COVID-19, FDA had to play catch-up by issuing enforcement discretion policies, through guidance, for certain unauthorized tests to help rapidly increase testing capacity on a nationwide scale – meaning certain tests were made available without EUA. Whether or not tests are authorized through EUA or described in enforcement discretion policies, the key concern for FDA is that these tests are properly validated. To this end, FDA can, and has, taken appropriate action against tests lacking the proper validation. In the IVD Draft Guidance, FDA provides recommendations for test validation so that IVD manufacturers can have a framework to efficiently secure authorization under EUA, and get much-needed treatments to the public, in the event of a new infectious disease outbreak.
B. Takeaways
The IVD Draft Guidance is clearly underscored by a desire to be better prepared to efficient, safe, and effective testing in the event of another disease outbreak like COVID-19 – in fact, FDA says as much in the guidance itself. What FDA does not explicitly say, but would could also underscore the Agency’s timing in issuing the guidance when it did, is a concern about how the incoming administration might handle such an outbreak in terms of testing and therapeutics, given some of the discourse we’ve heard to date.
Aside from emergency preparation, the IVD Draft Guidance also underscores FDA’s concerns about the efficacy of IVDs, generally, especially those that are subject to abbreviated validation standards. For example, last year, the Agency issued a lengthy (and controversial) final rule outlining a plan to end its previous policy of enforcement discretion for laboratory-developed tests (“LDTs”) – a subset of IVD – based on over a decade of concerns over the efficacy of these tests that have historically not been subject to any oversight, including validation standards, from FDA at all. The framework outlined in this IVD Draft Guide similarly addressed concerns over the efficacy of testing during emergency scenarios when manufacturers are subject to urgent time constraints and abbreviated EUA standards.
II. AI-Enabled Device Software
On January 7, FDA released a draft guidance titled “Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations” (the “AI Draft Guidance”).[3] This is only the third guidance FDA has released related to Artificial Intelligence (“AI”), and the second relating specifically to AI-enabled device software functions.[4] The AI Draft Guidance was expected, as it appeared on FDA’s Center for Devices and Radiological Health’s (“CDRH”) “A-list” of guidances to publish in Fiscal Year 2025.[5] The AI Draft Guidance signifies an acknowledgement by FDA of the need to keep pace – as best it can – with technological advancements in the medical device space, particularly in light of the attention and concern swirling around AI use. FDA is inviting comments to the Draft Guidance with a deadline set for April 7, 2025.
A. Background
The rapid advance of AI technology in recent years has significantly influenced the development and implementation of medical device software functions. Device manufacturers are increasingly integrating these AI-enabled device software functions (“AI-DSFs”) to enhance diagnostic, monitoring, and treatment capabilities. In recent years, FDA has focused on promoting a Total Product Life Cycle (“TPLC”) approach to the oversight of AI-DSF, which emphasizes ongoing management and continuous improvement of AI-enabled devices both pre- and post-market according to guiding principles for Good Machine Learning Practice (“GMLP”), to ensure that AI-DSF remain safe and effective from design through decommissioning. In the AI Draft Guidance, FDA continues its effort by establishing lifecycle management and marketing submission recommendations for AI-DSF and, as always, encouraging early interaction with FDA to ensure the development of a safe and effective product for patients.
B. Takeaways
In its AI Draft Guidance, FDA makes clear that the integration of AI modeling into medical device software is being scrutinized in a way that at least parallels, or even exceeds, the oversight given to general device software functions. The AI Draft Guidance sets forth many FDA recommendations for lifecycle management, suggesting that a large overhaul is needed to come up-to-speed with rapidly evolving AI development. Significantly, the scope of the AI Draft Guidance includes the device itself and any device constituent parts of a combination product, which may include AI-DSFs.
Much of the AI Draft Guidance focuses on premarket notification (e.g., 510(k)) submissions for devices that include AI-DSFs, notably requiring a thorough explanation of how AI is integrated into the device. In light of all the uncertainties surrounding AI use and its seemingly unlimited application, FDA seems to be looking for some level of assurance that AI-DSF developers will properly leverage parameters and safeguards so that this “unlimited” use potential does not transform a device beyond its cleared and/or approved intended use.
Another key focus of the AI Draft Guidance is transparency and bias reduction. This is a typical, and growing, area of concern for FDA when it comes to devices that collect and store information; however, incorporating AI complicates the issue because of unknown risk of bias. Specifically, FDA notes that AI models may rely on data correlations and other machine learning-derived processes that do not connect to biologically plausible mechanisms of action. Therefore, while AI’s strength is its adaptability, risk lies in the fact that its decision-making processes are not fully predictable. To mitigate this risk, FDA provides a recommended design approach to transparency throughout the product lifecycle, especially with respect to involving data collection and monitoring.
Another key focus of the AI Draft Guidance – and another growing area concern for FDA and stakeholders alike – is cybersecurity. Here, FDA builds off of its 2023 guidance (“2023 Guidance”) which addressed, more generally, cybersecurity in medical devices,[6] to contextualize it within the AI-sphere. The application of AI to medical device software adds a new layer of security concern because, if AI systems are hacked/accessed, the consequences can be much more widespread. To mitigate this risk, FDA provides comprehensive, AI-specific recommendations for handling cybersecurity threats, while also deferring to the 2023 Guidance for the complete framework that should be implemented prior to marketing submission.
The emergence of AI necessitates that FDA alter its long-standing framework for ensuring the safety and efficacy of medical devices in light of the unique way that AI-enabled device functions operate – and the AI Draft Guidance illustrates FDA’s continued recognition of, and response to this need.
III. Pulse Oximeters for Medical Purposes
On January 7, FDA released a draft guidance titled “Pulse Oximeters for Medical Purposes – Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations.” (the “PO Draft Guidance”),[7] which provides recommendations for performance testing, labeling, and premarket submissions of pulse oximeters. Once finalized, the PO Draft Guidance FDA’s existing pulse oximeter guidance, which was issued on March 4, 2013 (“2013 Guidance”).[8] FDA is inviting comments to the PO Draft Guidance with a deadline set for March 10, 2025.
A. Background
In recent years, there has been growing concern over the accuracy of readings from pulse oximeters, which are devices that measure the amount of oxygen in arterial blood and pulse rate.[9] In addressing this concern, FDA found that a host of factors affects the accuracy of pulse oximeter readings, especially person’s skin pigmentation. In light of this particular concern, FDA engaged interested parties, and partnered with the University of California San Francisco, as part of the Centers of Excellence in Regulatory Science and Innovation (“CERSI”) program, to conduct a study comparing pulse oximeter errors in clinical patients with varying skin tones. Based on results from this study, as well as input from interested stakeholders, FDA has created enhanced recommendations for marketing submissions to ensure that pulse oximeters used as standalone medical devices, or as part of a multi-parameter medical device, accurately fulfill their intended use. This comprehensive list of marketing submission recommendations is laid out in the new PO Draft Guidance and includes enhanced clinical performance testing procedures that specifically account for disparities in performance across different populations, such as diverse skin tones and pediatric populations.
FDA is showing that it is concerned not only with whether the device performs its intended function accurately, but whether that performance is consistent across all patient populations. Significantly, FDA is exhibiting concern regarding the diversity of patient populations across this country, urging manufacturers to ensure accuracy for all.
B. Takeaways
This new PO Draft Guidance underscores FDA’s continuing commitment to ensuring that regulated products are safe and effective for all individuals – not only those majority populations who have typically been the subject of clinical testing and validation. It is incumbent on manufacturers, FDA, and providers to ensure that medical devices perform properly for each and every person, irrespective of differences in identifying characteristics. And where a certain product has not been tested on and/or cannot be confirmed safe and effective for a certain population, FDA is clear that this limitation needs to be made known to prescribers and end users by limiting the product’s intended use and associated labeling. Bottom line – we can’t have patients relying on products that do not operate safely and/or specifically for them and others like them.
Conclusion
The common thread among these device-specific Draft Guidances is an emphasis on early collaboration with FDA to get ahead of certain identified issues that pose public health threats—an infectious disease emergency, an unmanageable and/or unsecure AI algorithm, or a test result that was not clinically verified for a certain patient’s skin tone. Now, we have heard this refrain before, especially on the drug side of the house—we often roll our eyes when we see it, given that the Agency holds the ultimate power in just about any facet of inquiry, decision-making, and enforcement. But given the blistering speed with which these technologies have been and will continue to develop, FDA’s entreaties here might mean something more.
Indeed, despite the myriad of other critical issues that FDA needs to address, it is clear that CDRH policymakers did not intend for devices to fall by the wayside as this administrations changed guard. Whatever happens over the coming months, all eyes in our industry will be on FDA policy—in guidance, enforcement, or otherwise.
FOOTNOTES
[1] IVD Draft Guidance available here: Validation of Certain In Vitro Diagnostic Devices for Emerging Pathogens During a Section 564 Declared Emergency | FDA
[2] FD&C Act Section 564 available here: 21 U.S.C. 360bbb-3.
[3] AI Draft Guidance available here: Artificial Intelligence-Enabled Device Software Functions: Lifecycle Management and Marketing Submission Recommendations | FDA
[4] See December 2024 guidance available here: Marketing Submission Recommendations for a Predetermined Change Control Plan for Artificial Intelligence-Enabled Device Software Functions | FDA; January 2025 draft guidance available here: Considerations for the Use of Artificial Intelligence To Support Regulatory Decision-Making for Drug and Biological Products | FDA
[5] 2025 A-List available here: CDRH Proposed Guidances for Fiscal Year 2025 (FY2025) | FDA
[6] 2023 Guidance available here: Cybersecurity in Medical Devices: Quality System Considerations and Content of Premarket Submissions | FDA
[7] PO Draft Guidance available here: Pulse Oximeters for Medical Purposes – Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations | FDA
[8] 2013 Guidance available here: Pulse Oximeters – Premarket Notification Submissions [510(k)s]: Guidance for Industry and Food and Drug Administration Staff | FDA
[9] See, e.g., Pulse Oximeter Accuracy and Limitations: FDA Safety Communication, FDA (Feb. 19, 2021).; Multistate Letter Urging FDA to Address Concerns about Dangerous Pulse Oximeter Inaccuracies Impacting Communities of Color, Cal. Atty. Gen. (Nov. 1, 2023).