A newly formed FDA advisory committee recommended several approaches for how the agency should handle regulation of medical devices based on generative artificial intelligence (AI) during a two-day meeting that concluded THURSDAY.
The Digital Health Advisory Committee (DHAC) held its first meeting to provide advice to the FDA on a range of issues related to the development, evaluation, implementation, and ongoing monitoring of medical devices based on AI.
During the opening remarks, FDA Commissioner Robert Califf, MD, said DHAC will provide important guidance and recommendations on the benefits and risks associated with all digital health technologies, including medical devices AI-based generative tools.
“We created this committee because we see great potential in digital health technologies to help solve the critical health issues we face today, and we need these technologies to be developed, deployed and used widely. responsibly in the best interest of patients and consumers,” Califf said, adding that “artificial intelligence is changing the way we think about health and healthcare, and it is one of the most exciting areas of science. and the most promising because it is designed to transcend the borders.
In the Executive summary shared before the meeting, FDA staff asked the committee to consider planning and design, data collection and management, model construction and tuning, verification and validation, deployment, operating and monitoring models, as well as evaluating real-world performance.
The discussion produced a significant number of considerations and ideas for how the FDA should approach all phases of the development process, said committee chair Ami Bhatt, MD, chief innovation officer at the American College of Cardiology.
“There are many eyes, many opinions, many companies, research labs, friends of ours and, most importantly, patients who hope for the promise of generative AI, and that is why we consider our work here as developing an infrastructure for growth with guardrails,” she said. “This is not the end, but only the beginning of a process of continuous change.”
Although the committee did not vote on specific recommendations for the agency, Bhatt noted that they were able to create “an actionable framework” for how AI-based generative devices should be managed by the FDA in the future. Following the FDA’s list of discussion questions for the meeting, committee members proposed a framework based on three distinct areas: premarket performance evaluation, risk management, and postmarket performance monitoring .
Among pre-market performance considerations, committee members said the agency should develop customized multidimensional frameworks to evaluate the overall accuracy of AI models, the effectiveness of predicted generative AI outcomes based on user inputs or prompts, and the ability to identify potentials. the harms and risks introduced by AI-based generative devices.
In a similar vein, the committee noted that FDA should consider developing a small set of widely accepted measures and methods for use in evaluating such devices.
They also said the agency should develop standard definitions of terms and concepts to discuss generative AI, particularly for key limitations such as non-distribution datadata drift and hallucinations. Notably, the lack of consistent definitions for several terms related to generative AI presented challenges several times during the meeting.
The committee also highlighted the lack of sufficient study designs to test these devices for clinical use, and urged the FDA to explore the potential use of alternative study approaches, such as synthetic control trials, to improve the evaluation of the comparative effectiveness of these devices.
For postmarket performance evaluation, the committee said the agency should consider approaches to expand evaluations once a device is widely adopted by clinicians or consumers. They also highlighted the need to automate these monitoring and evaluation processes to avoid time-consuming and costly human review of these devices once they are used on a larger scale.
Additionally, they said the FDA should consider establishing new frameworks for understanding the impact of AI-based generative devices on society once they hit the market.
For this, the committee recommended that the agency consider establishing a centralized data repository and reporting mechanism that can be used to track errors and damage caused by these devices, and noted that these tools could also be used to continuously monitor device performance across various populations and settings.
The committee also emphasized that the agency must keep in mind the impact of these devices on health equity. They recommended that the FDA develop requirements that companies must implement and demonstrate how the safeguards protect against built-in or learned biases over time.
Finally, the committee said the agency should develop certification programs or other standards to ensure that companies developing these devices understand the risk of bias in their AI-enabled generative devices.
After two days of in-depth discussions on these issues, committee members noted that the development of this regulatory infrastructure would be an ongoing process. Bhatt acknowledged that this process would be gradual, but that establishing clear guidelines for the implementation of generative AI in the healthcare setting could help improve healthcare delivery across the country in the near future.
“One of the challenges we face, because generative AI is often tied to clinical guidelines or clinical decision support, is knowing what the gold standard is,” Bhatt said. “When we ask ourselves whether or not we are providing treatment derived from gold standard clinical guidelines throughout the United States across all different specialties, the answer is generally no.”
This technology could help improve the overall quality of care offered to patients, she added. So the question we should ask ourselves is: how close does generative AI bring us to the gold standard? »
