Page 1 of 20Free preview · 1–5

About this document

FDA discussion paper proposing a total product lifecycle regulatory framework for modifications to AI/ML-based software as a medical device. It addresses changes to performance, inputs, and intended use; good machine learning practices; and predetermined change control plans comprising SaMD Pre-Specifications and an Algorithm Change Protocol. The proposal also covers premarket review, transparency, and real-world performance monitoring. Appendices provide hypothetical device modification scenarios and proposed protocol content, followed by questions for stakeholder feedback. The paper is for discussion only and is not draft or final guidance.

Free preview text

The first 5 pages, shown as text so search engines and readers can find this document.

Page 1

pg. 1 www.fda.gov

Page 2

pg. 2 www.fda.gov Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback I. Introduction Artificial intelligence (AI)- and machine learning (ML)-based technologies have the potential to transform healthcare by deriving new and important insights from the vast amount of data generated during the delivery of healthcare every day. Example high-value applications include earlier disease detection, more accurate diagnosis, identification of new observations or patterns on human physiology, and development of personalized diagnostics and therapeutics. One of the greatest benefits of AI/ML in software resides in its ability to learn from real-world use and experience, and its capability to improve its performance. The ability for AI/ML software to learn from real-world feedback (training) and improve its performance (adaptation) makes these technologies uniquely situated among software as a medical device (SaMD)1 and a rapidly expanding area of research and development. Our vision is that with appropriately tailored regulatory oversight, AI/ML-based SaMD will deliver safe and effective software functionality that improves the quality of care that patients receive. FDA has made significant strides in developing policies2, 3 that are appropriately tailored for SaMD to ensure that safe and effective technology reaches users, including patients and healthcare professionals. Manufacturers submit a marketing application to FDA prior to initial distribution of their medical device, with the submission type and data requirements based on the risk of the SaMD (510(k) notification, De Novo, or premarket approval application (PMA) pathway). For changes in design that are specific to software that has been reviewed and cleared under a 510(k) notification, FDA’s Center for Devices and Radiological Health (CDRH) has published guidance (Deciding When to Submit a 510(k) for a Software Change to an Existing Device,4 also referred to herein as the software modifications guidance) that describes a risk-based approach to assist in determining when a premarket submission is required.5 1 Software as a Medical Device (SaMD): Key Definitions: http://www.imdrf.org/docs/imdrf/final/technical/imdrf-tech-131209-samd-key- definitions-140901.pdf. 2 Pre-Cert Program Version 1.0 Working Model: https://www.fda.gov/downloads/MedicalDevices/DigitalHealth/DigitalHealthPreCertProgram/UCM629276.pdf. 3 Software as a Medical Device (SaMD): Clinical Evaluation: https://www.fda.gov/downloads/medicaldevices/deviceregulationandguidance/guidancedocuments/ucm524904.pdf. 4 Deciding When to Submit a 510(k) for a Software Change to an Existing Device: https://www.fda.gov/downloads/medicaldevices/deviceregulationandguidance/guidancedocuments/ucm514737.pdf. 5 21 CFR 807.81(a)(3). Modifications to a device approved through a PMA are governed by the criteria in 21 CFR 814.39(a). Modifications to Devices Subject to Premarket Approval (PMA) - The PMA Supplement Decision-Making Process: https://www.fda.gov/downloads/MedicalDevices/DeviceRegulationandGuidance/GuidanceDocuments/UCM089360.pdf. The International Medical Device Regulators Forum (IMDRF) defines ‘Software as a Medical Device (SaMD)’ as software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device.1 FDA, under the Federal Food, Drug, and Cosmetic Act (FD&C Act) considers medical purpose as those purposes that are intended to treat, diagnose, cure, mitigate, or prevent disease or other conditions.

Page 3

pg. 3 www.fda.gov The 510(k) software modifications guidance focuses on the risk to users/patients resulting from the software change. Categories of software modifications that may require a premarket submission include: • A change that introduces a new risk or modifies an existing risk that could result in significant harm; • A change to risk controls to prevent significant harm; and • A change that significantly affects clinical functionality or performance specifications of the device. When applied to AI/ML-based SaMD, the above approach would require a premarket submission to the FDA when the AI/ML software modification significantly affects device performance, or safety and effectiveness6; the modification is to the device’s intended use; or the modification introduces a major change to the SaMD algorithm. For a PMA-approved SaMD, a supplemental application would be required for changes that affect safety or effectiveness, such as new indications for use, new clinical effects, or significant technology modifications that affect performance characteristics. To address the critical question of when a continuously learning AI/ML SaMD may require a premarket submission for an algorithm change, we were prompted to reimagine an approach to premarket review for AI/ML-driven software modifications. Such an approach would need to maintain reasonable assurance of safety and effectiveness of AI/ML-based SaMD, while allowing the software to continue to learn and evolve over time to improve patient care. To date, FDA has cleared or approved several AI/ML-based SaMD. Typically, these have only included algorithms that are “locked7” prior to marketing, where algorithm changes likely require FDA premarket review for changes beyond the original market authorization. However, not all AI/ML-based SaMD are locked; some algorithms can adapt over time. The power of these AI/ML-based SaMD lies within the ability to continuously learn, where the adaptation or change to the algorithm is realized after the SaMD is distributed for use and has “learned” from real-world experience. Following distribution, these types of continuously learning and adaptive AI/ML algorithms may provide a different output in comparison to the output initially cleared for a given set of inputs. The traditional paradigm of medical device regulation was not designed for adaptive AI/ML technologies, which have the potential to adapt and optimize device performance in real-time to continuously improve healthcare for patients. The highly iterative, autonomous, and adaptive nature of these tools requires a new, total product lifecycle (TPLC) regulatory approach that facilitates a rapid cycle of product improvement and allows these devices to continually improve while providing effective safeguards. This discussion paper proposes a framework for modifications to AI/ML-based SaMD that is based on the internationally harmonized International Medical Device Regulators Forum (IMDRF) risk categorization principles, FDA’s benefit-risk framework, risk management principles in the software 6 21 CFR 807.81(a)(3). 7 We define a “locked” algorithm as an algorithm that provides the same result each time the same input is applied to it and does not change with use. Examples of locked algorithms are static look-up tables, decision trees, and complex classifiers.

Page 4

pg. 4 www.fda.gov modifications guidance8, and the organization-based TPLC approach as envisioned in the Digital Health Software Precertification (Pre-Cert) Program.9 It also leverages practices from our current premarket programs, including the 510(k), De Novo, and PMA pathways. This discussion paper describes an innovative approach that may require additional statutory authority to implement fully. The proposed framework is being issued for discussion purposes only and is not a draft guidance. This document is not intended to communicate FDA's proposed (or final) regulatory expectations but is instead meant to seek early input from groups and individuals outside the Agency prior to development of a draft guidance. This proposed TPLC approach allows FDA’s regulatory oversight to embrace the iterative improvement power of AI/ML SaMD while assuring that patient safety is maintained. It also assures that ongoing algorithm changes are implemented according to pre-specified performance objectives, follow defined algorithm change protocols, utilize a validation process that is committed to improving the performance, safety, and effectiveness of AI/ML software, and include real-world monitoring of performance. This proposed TPLC regulatory framework aims to promote a mechanism for manufacturers to be continually vigilant in maintaining the safety and effectiveness of their SaMD, that ultimately, supports both FDA and manufacturers in providing increased benefits to patients and providers. II. Background: AI/ML-Based Software as a Medical Device In this paper, we use John McCarthy’s definition of AI as the science and engineering of making intelligent machines, especially intelligent computer programs.10 AI can use different techniques, such as ML, to produce intelligent behavior, including models based on statistical analysis of data, and expert systems that primarily rely on if-then statements. In this paper, we refer to an ML system as a system that has the capacity to learn based on training on a specific task by tracking performance measure(s). AI, and specifically ML, are techniques used to design and train software algorithms to learn from and act on data. These AI/ML-based software, when intended to treat, diagnose, cure, mitigate, or prevent disease or other conditions, are medical devices under the FD&C Act, and called “Software as a Medical Device” (SaMD) by FDA and IMDRF. The intended use of AI/ML-based SaMD, similar to other SaMDs, may exist on a spectrum of impact to patients as categorized by IMDRF SaMD risk categorization framework.11 8 Deciding When to Submit a 510(k) for a Software Change to an Existing Device: https://www.fda.gov/downloads/medicaldevices/deviceregulationandguidance/guidancedocuments/ucm514737.pdf. 9 Developing a Software Precertification Program: A Working Model; v1.0 – January 2019: https://www.fda.gov/downloads/MedicalDevices/DigitalHealth/DigitalHealthPreCertProgram/UCM629276.pdf. 10 http://jmc.stanford.edu/articles/whatisai/whatisai.pdf. 11 Software as a Medical Device (SaMD): Possible Framework for Risk Categorization and Corresponding Considerations: http://www.imdrf.org/docs/imdrf/final/technical/imdrf-tech-140918-samd-framework-risk-categorization-141013.pdf. Non-device software functions are not subject to FDA device regulation and are not within the scope of this paper. In addition, as detailed in section 502(o) of the FD&C Act, software functions intended (1) for administrative support of a health care facility, (2) for maintaining or encouraging a healthy lifestyle, (3) to serve as electronic patient records, (4) for transferring, storing, converting formats, or displaying data, or (5) to provide certain, limited clinical decision support are not medical devices and are not subject to FDA regulation.

Page 5

pg. 5 www.fda.gov The IMDRF SaMD risk categorization framework takes a risk-based approach to categorize SaMD based on intended use, similar to traditional risk-based approaches used by the FDA. The IMDRF risk framework identifies the following two major factors as providing a description of the intended use12 of the SaMD: 1) Significance of information provided by the SaMD to the healthcare decision, which identifies the intended use of the information provided by the SaMD – i.e., to treat or diagnose; to drive clinical management; or to inform clinical management; and 2) State of healthcare situation or condition, which identifies the intended user, disease or condition, and the population for the SaMD – i.e., critical; serious; or non-serious healthcare situations or conditions. Taken together, these factors describing the intended use can be used to place the AI/ML-based SaMD into one of four categories, from lowest (I) to highest risk (IV) to reflect the risk associated with the clinical situation and device use. While AI/ML-based SaMD exist on a spectrum categorized by risk to patients, they also exist on a spectrum from locked to continuously learning. “Locked” algorithms are those that provide the same result each time the same input is provided. As such, a locked algorithm applies a fixed function (e.g., a static look-up table, decision tree, or complex classifier) to a given set of inputs. These algorithms may use manual processes for updates and validation. In contrast to a locked algorithm, an adaptive algorithm (e.g., a continuous learning algorithm) changes its behavior using a defined learning process. The algorithm adaptation or changes are implemented such that for a given set of inputs, the output may be different before and after the changes are implemented. These algorithm changes are typically implemented and validated through a well-defined and possibly fully automated process that aims at improving performance based on analysis of new or additional data. The adaptation process can be intended to address several different clinical aspects, such as optimizing performance within a specific environment (e.g., based on the local patient population), optimizing performance based on how the device is being used (e.g., based on preferences of a specific physician), improving performance as more data are collected, and/or changing the intended use of the device. The adaptation process follows two stages: learning and updating. The algorithm “learns” how to change its behavior, for example, from the addition of new input types or adding new cases to an already existing training database. The “update” then occurs when the new version of the algorithm is deployed. As a 12 Information that may be used to describe intended use for FDA purposes is set forth in 21 CFR 807.92(a)(5), 814.20(b)(3), and 860.7(b), and could be written using terminology as described in the IMDRF risk categorization framework. State of healthcare situation or condition Significance of information provided by SaMD to healthcare decision Treat or diagnose Drive clinical management Inform clinical management Critical IV III II Serious III II I Non-serious II I I Figure 1: SaMD IMDRF risk categorization

Tags

Doxxly editorial guide

Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback: summary, contents and background

This guide helps you understand the scope of Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback before or alongside reading the original document. It is based on the catalogue record, credited creator, source, classification and topics shown on this page. It does not replace the original work and does not invent claims about passages that Doxxly cannot verify.

Summary and overview

Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback is a 20-page documentcredited to Cleo D. Jarrow. FDA discussion paper proposing a total product lifecycle regulatory framework for modifications to AI/ML-based software as a medical device. It addresses changes to performance, inputs, and intended use; good machine learning practices; and predetermined change control plans comprising SaMD Pre-Specifications and an Algorithm Change Protocol. The proposal also covers premarket review, transparency, and real-world performance monitoring. Appendices provide hypothetical device modification scenarios and proposed protocol content, followed by questions for stakeholder feedback. The paper is for discussion only and is not draft or final guidance. It appears in the Law & Legal Studies collection, which groups it with other material that readers may use for comparison and further study.

The item can be approached in several ways. A reader looking for a quick orientation can begin with the free five-page preview, inspect the page structure and use the metadata to judge relevance. Someone undertaking deeper research can work through the complete file, search its indexed text, record useful page references and compare its treatment of FDA, Artificial intelligence, Machine learning, Software as a Medical Device, Medical device regulation, Total product lifecycle, Algorithm change control, Good machine learning practices, Real-world performance monitoring, Discussion paper with the related resources linked below.

Rather than reducing the work to a single claim, this page treats it as a source that should be read in context. The title, creator, publication source, licence and format all help explain what kind of evidence it may offer. Readers should distinguish the original author's statements from Doxxly's catalogue description, and should cite the original edition when using the material in academic, professional or published work.

Key topics and reader takeaways

The catalogue connects this document with FDA, Artificial intelligence, Machine learning, Software as a Medical Device, Medical device regulation, Total product lifecycle, Algorithm change control, Good machine learning practices, Real-world performance monitoring, Discussion paper. These terms offer practical entry points rather than a substitute for close reading. Start by identifying the definitions, examples, arguments, tables or visual explanations attached to each topic. Then note where the creator draws connections, makes distinctions or moves from background material to a conclusion, recommendation, calculation or narrative development.

A useful reading method is to separate description from interpretation. First record what the item actually presents: names, headings, figures, formulas, quotations or stated propositions. Next consider why those details matter within the whole work. Finally compare them with another source from the same subject area. This process helps turn a long digital file into evidence that can be evaluated rather than merely skimmed.

Table of contents and reading guide

The following is an editorial content guide based on the verified catalogue topics. It is not represented as the original author's formal table of contents. Use it to organise notes while moving through the 20 pages in the player.

  1. Part 1

    FDA

    Use this section of the reading guide to identify how FDA appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  2. Part 2

    Artificial intelligence

    Use this section of the reading guide to identify how Artificial intelligence appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  3. Part 3

    Machine learning

    Use this section of the reading guide to identify how Machine learning appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  4. Part 4

    Software as a Medical Device

    Use this section of the reading guide to identify how Software as a Medical Device appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  5. Part 5

    Medical device regulation

    Use this section of the reading guide to identify how Medical device regulation appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  6. Part 6

    Total product lifecycle

    Use this section of the reading guide to identify how Total product lifecycle appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  7. Part 7

    Algorithm change control

    Use this section of the reading guide to identify how Algorithm change control appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

  8. Part 8

    Good machine learning practices

    Use this section of the reading guide to identify how Good machine learning practices appears in Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, note important terms, and connect the subject to the wider purpose of the work.

When a heading in the original differs from this guide, follow the original heading in your citation. Page numbers can also vary between editions. Doxxly's direct page field is useful for navigation, but formal references should include the title, creator, source and edition details available in the file itself.

About Cleo D. Jarrow

Doxxly credits Cleo D. Jarrow as the author or creator of Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback. This page does not add an unverified biography, job title or institutional affiliation. The attribution is preserved so readers can distinguish the creator from the hosting library and follow the original record when an external author link is available.

Author context matters because a work reflects a particular purpose, audience and point of view. Before relying on the material, consider what the creator set out to explain, when the source was produced and whether later research or revised guidance should also be consulted. For literature and philosophy, compare translations and editions; for technical material, check whether a newer version has superseded the one shown here.

About the source organisation

The recorded source for this item is fda.gov. Doxxly provides access and discovery tools but does not claim to be the original publisher unless explicitly stated. Source links, creator attribution and the recorded Publicly available file collected for editorial review licence are displayed to help readers assess provenance and permitted use.

Provenance is especially important for reports, standards, presentations and working spreadsheets. Check the original source for amendments, supporting files and current policy. A trustworthy research workflow keeps the source URL with any notes or downloaded copy, verifies important figures against the original, and avoids treating an archived document as automatically current.

Visit fda.gov

How to read and use this document

Begin with the title and opening pages, then use the guide above to decide whether to read sequentially or move to a particular topic. The Doxxly player provides previous and next navigation, a direct page field, zoom steps, fit-to-width and fit-to-page modes, rotation and fullscreen. These controls are useful when a page contains small type, a wide table, an image, a scanned layout or a dense sequence of notes.

The first five pages are open to every visitor. A plan unlocks the complete item and full-document search. Search is best used for distinctive names, technical terms or short phrases; broad words can produce too many possible contexts. When you find a useful passage, read the surrounding pages so that a sentence, chart or formula is not separated from its explanation.

If downloading is enabled, members can obtain the original PDF file. Keep the original format when formulas, slide sequencing, pagination or layout carries meaning. Respect the Publicly available file collected for editorial review terms displayed on this page, and use the rights-report link if you believe the attribution or availability needs review.

Research questions to consider

As you read Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback, ask which audience the creator addresses, what problem or idea organises the work, what evidence is presented, and which assumptions remain unstated. Note whether important terms are defined consistently and whether examples represent a broader pattern or a narrow case. For quantitative material, record units, time periods, formulas and data sources. For interpretive material, separate the creator's language from your own summary.

A second set of questions concerns relevance today. Which parts remain durable, which depend on their historical or institutional setting, and what would need confirmation from a newer source? Following the tags and the Law & Legal Studies collection can reveal useful comparisons. Agreement between sources can strengthen confidence, while disagreement can point to changing definitions, methods, evidence or priorities.

Finally, decide what you need from the item: background knowledge, a quotation, a worked example, a model, a visual explanation or a starting bibliography. That purpose should guide the depth of reading and the notes you keep. Doxxly helps locate and navigate the source; careful interpretation remains the reader's responsibility.

Explore more Law & Legal Studies resources

Frequently asked questions about Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) - Discussion Paper and Request for Feedback

Practical answers about reading, downloading and understanding this document on Doxxly.

More in Law & Legal Studies