What is the FDA’s AI/ML “Action Plan”?

What is the FDA’s AI/ML “Action Plan”?

This report from the FDA is a high-level (low on details) action plan that just describes how they are trying to address stakeholder concerns with their proposed regulatory framework for utilizing AI/ML as a medical device.

While artificial intelligence (AI) and machine learning (ML) continue to be all the rage, there is a question as to how you actually get AI (or a device that uses AI) into the clinic. As clinical or translational AI is relatively new and most anything that touches a patient has to go through the FDA, everybody is still learning how these algorithms get approved for use with patients. The FDA produced a report earlier in the year that describes their action plan regarding the incorporation of AI- and ML-based methods in medical software (“Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) Action Plan”).[1] I was curious about what was in this “action plan” so had a look.

Three software categories for the FDA

Before describing this plan, note that there are three types of software categories related to medical devices that are recognized by the FDA. The first is software that is embedded within a medical device and is integral to its function. As with many consumer devices, the vast majority of medical devices today have some kind of software in them that makes them tick. If you are putting software in a medical device, the FDA has to approve that it is safe and works as advertised.

The second category deals with software that will be used either in the manufacture or maintenance of a medical device. This category isn't very sexy and gets the least publicity, but it too has its own rules and regulations that manufacturers of medical devices have to follow.

Finally, the third and last category is that of "software as a medical device" (SaMD)[2]. In this case, the software acts independently of a device, making clinical classifications, decisions or other operations largely all on its own. More formally, it is defined as "software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device." This category is where you expect to see many of the AI/ML applications land.

The FDA’s “Action Plan”

Having a look at this short report (six pages of text) you'll find that it largely focuses on describing pretty high-level concerns from outside stakeholders regarding an earlier publication — the "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,"[3] along with comments on what the FDA is thinking of doing to address those concerns.

Taken straight from the report are the FDA's five major (intended) actions and goals:

  • Develop an update to the proposed regulatory framework presented in the AI/ML-based SaMD discussion paper, including through the issuance of a Draft Guidance on the Predetermined Change Control Plan.
  • Strengthen FDA’s encouragement of the harmonized development of Good Machine Learning Practice (GMLP) through additional FDA participation in collaborative communities and consensus standards development efforts.
  • Support a patient-centered approach by continuing to host discussions on the role of transparency to users of AI/ML-based devices. Building upon the October 2020 Patient Engagement Advisory Committee (PEAC) Meeting focused on patient trust in AI/ML technologies, hold a public workshop on medical device labeling to support transparency to users of AI/ML-based devices.
  • Support regulatory science efforts on the development of methodology for the evaluation and improvement of machine learning algorithms, including for the identification and elimination of bias, and on the robustness and resilience of these algorithms to withstand changing clinical inputs and conditions.
  • Advance real-world performance pilots in coordination with stakeholders and other FDA programs, to provide additional clarity on what a real-world evidence generation program could look like for AI/ML-based SaMD.

So what can you learn from this?

While each of these actions were fleshed out with a bit more detail in the report, the main outcome of this report is largely just an acknowledgement by the FDA that they heard concerns from relevant stakeholders and are working to address them.

For example, the first point above is focused on developing an updated regulatory framework for AI/ML-based SaMD. Feedback from stakeholders was broadly supportive of the FDA's proposal for how to describe which aspects of an AI application will change through learning while remaining safe and effective. There were, however, suggestions for additional types of modifications that should be addressed in this framework. In this report, the FDA simply acknowledged that they recognize these suggestions and will update this proposed regulatory framework to address these sorts of concerns.

While this "action plan" wasn't nearly as interesting as I thought it would be, it is valuable to understand what it is that the FDA and other parties are focusing in on as likely regulatory obstacles for translational AI.

There are other reports from the FDA that are certainly of more interest to translational AI folks, such as the discussion paper on which this action plan is based. I'll describe this one and potentially some others in future posts.

I'll be honest, reading about rules and regulations is incredibly boring stuff (at least for me). But you need at least a basic understanding of the hurdles in front of you if you want to make a contribution in this space. Thankfully, there are people out there who love this stuff. You will want to work with them if you want to really figure out how to get an AI/ML tool into the clinic and helping patients.

References


  1. The action plan: (https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-and-machine-learning-software-medical-device) ↩︎

  2. (https://www.fda.gov/medical-devices/digital-health-center-excellence/software-medical-device-samd) ↩︎

  3. The discussion paper on which this action plan is based:(https://www.fda.gov/media/122535/download) ↩︎