FDA's Plan for AI and Machine Learning

The Food and Drug Association, more commonly known as the FDA, is the oldest consumer protection agency and is part of the U.S Department of Health and Human Services. On 12 January 2021, it released a document titled Artificial Intelligence and Machine Learning (AI/ML) Software as a Medical Device Action Plan to propose how AI/ML software should be regulated when used as a medical device.

Software as a medical device (SaMD) is software intended to be used for one or more medical purposes that perform these purposes without being part of a hardware medical device [1].

The use of SaMDs is ever-increasing, as technological capabilities increase. While you may not be aware of it, SaMDs are becoming ubiquitous. For example, smart-watches that monitor physiological parameters such as glucose levels and EEG are SaMD. Given this increase, the FDA decided to publish 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. The aim was to initiate a discussion in order to define a framework for the use of AI/ML as SaMD. The plan of action mentioned previously is the FDA's next step towards regulating AI/MD as SaMD.

The Plan of Action

The plan of action consists of 5 items that were identified and formulated in direct response to the stakeholder feedback (that is, feedback from the discussion paper mentioned previously).

1. Tailored Regulatory Framework for AI/ML-based SaMD

One of the complexities of regulating AI/ML is that there is a "learning" aspect to the algorithms. This means that the algorithms are bound to change. A principle of Predetermined Change Control should be used. This Change Control should specify what the manufacturer intends to change, and how the algorithm will learn and change while remaining safe and effective. The FDA has committed to issuing a guidance document, open for public feedback, for the Predetermined Change Control Plan.

2. Good Machine Learning Practice (GMLP)

According to the feedback from the discussion paper, stakeholders were in support of the harmonization of the development of GMLP through consensus standards efforts and other community initiatives. In this regard, the FDA will collaborate with Medical Device Cybersecurity Program and put in place GLMP.

3. Patient-Centered Approach Incorporating Transparency to Users

Stakeholders called for further discussion with FDA on how AI/ML-based technologies

interact with people, including their transparency to users and to patients more broadly. The FDA's response to this is to hold a public workshop in the future to discuss device labeling transparency from the manufacturer's perspective, and how user trust can be enhanced.

4. Regulatory Science Methods Related to Algorithm Bias & Robustness

The FDA plans to support regulatory science efforts to develop a methodology for the evaluation and improvement of machine learning algorithms, including for the identification and elimination of bias, and for the evaluation and promotion of algorithm robustness.

5. Real-World Performance (RWP)

There are many questions around the measurement of performance of AI/ML SaMD. For example, what reference data should be used in the field? How much data should be provided to the FDA and how often? The list goes on. There is an important expectation that the FDA should provide clarity in the RWP area. Therefore, the FDA is committed to supporting the piloting of real-world performance monitoring by working with stakeholders on a voluntary basis. This will be accomplished in coordination with other ongoing FDA programs focused on the use of real-world data.

My Takeaway

Given how rapidly technology is growing, it is not surprising that SaMDs are on the rise. I appreciate the FDA's proactiveness towards AI/ML regulation, although I am not holding my breath. This is nothing against the FDA or any regulatory authority that will find itself in a similar position. My concern lies mainly in the implementation of such plans. Unfortunately, we live in a world where regulation and bureaucracy are unable to keep up with innovation. We often hear plans of action, strategies, and the like, but if experience serves us well, we'll realize that talk is cheap. Hats of to the FDA for anticipating the changes in the medical device industry. I guess time will reveal the effectiveness of this plan. May we not be disappointed.

References

[1] http://www.imdrf.org/docs/imdrf/final/technical/imdrf-tech-131209-samd-key-definitions-140901.pdf

[2] https://www.fda.gov/media/145022/download

[3] https://www.fda.gov/media/122535/download