Nearly Half of AI Medical Devices Lack Proper FDA Evaluation, New Report Reveals

The FDA is approving an average of 69 medical AI devices per year, but about half of those approvals lack clinical validation data according to an analysis published in Nature. The various uses for Artificial Intelligence in medical settings are rapidly growing, and the FDA only approved an average of two AI devices per year before 2016. The study authors wrote, “Devices that lack adequate clinical validation pose risks for patient care. A new validation standard is proposed to evaluate FDA authorization as an indication of clinical effectiveness in medical AI.”

Researchers Sammy Chouffani El Fassi and Gail E. Henderson led an analysis of over 500 AI medical devices to find that about half lacked clinical scientific data to prove the devices are effective for human use. The authors stated, “Although AI device manufacturers boast of the credibility of their technology with FDA authorization, clearance does not mean that the devices have been properly evaluated for clinical effectiveness using real patient data. With these findings, we hope to encourage the FDA and industry to boost the credibility of device authorization by conducting clinical validation studies on these technologies and making the results of such studies publicly available.”

144 of the devices were “retrospectively validated,” which means the AI is given image data from the past. A better scientific approach is prospective validation, which provides real-time patient data. 148 of the devices in the analysis were approved with prospective validation. 22 of the devices were approved with randomized control trials, which is the gold standard.

The study’s authors said there is a need for the FDA to clearly define and distinguish between the three types of validation. Chouffani El Fassi added, “We shared our findings with directors at the FDA who oversee medical device regulation, and we expect our work will inform their regulatory decision-making. We also hope that our publication will inspire researchers and universities globally to conduct clinical validation studies on medical AI to improve the safety and effectiveness of these technologies. We’re looking forward to the positive impact this project will have on patient care at a large scale.”

The AI medical device market will grow from $15 billion in 2023 to $22 billion in 2024. Estimations say it will reach $97 billion by the year 2028. An executive order from October 2023 lays out concerns and hopes for the development of AI technology for a wide variety of industries.

226 out of 521 authorizations by the FDA lacked validation data from real patients. Some of them contained created images rather than data from real patients.

That executive order states, “Artificial Intelligence must be safe and secure.  Meeting this goal requires robust, reliable, repeatable, and standardized evaluations of AI systems, as well as policies, institutions, and, as appropriate, other mechanisms to test, understand, and mitigate risks from these systems before they are put to use.”

In response to that executive order, the FDA Center for Drug Evaluation and Research (CDER) is forming a new council that will oversee all AI-related technologies within the FDA. This will replace the current AI-related steering committee and additional AI-related working groups.

The council is co-led by three individuals within the FDA. Tala Fakhour, the associate director for data sciences and AI policy, has previously worked for the CDC and ICF International, a consulting firm. Qi Liu, the associate director for innovation and partnership in the office of Clinical Pharmacology, formerly worked for Merck. Sri Mantha, the director of the Office of Strategic Programs, has previously worked for multiple pharmaceutical companies including Pfizer and AstraZeneca.

Chouffani El Fassi said, “We shared our findings with directors at the FDA who oversee medical device regulation, and we expect our work will inform their regulatory decision-making. We also hope that our publication will inspire researchers and universities globally to conduct clinical validation studies on medical AI to improve the safety and effectiveness of these technologies. We’re looking forward to the positive impact this project will have on patient care at a large scale.”

A 2023 research paper outlines several drawbacks for using AI in the medical field, including concerns related to ethics, society, and privacy. AI devices in healthcare require algorithmic models that use a significant amount of private patient data. Researchers have warned that this is a privacy risk because this data can potentially be hacked.

Last year, the World Health Organization (WHO) warned that the rapid adoption of AI technology for medical purposes can cause harm to patients. The WHO said, “Precipitous adoption of untested systems could lead to errors by healthcare workers, cause harm to patients, erode trust in AI and thereby undermine (or delay) the potential long-term benefits and uses of such technologies around the world.”

FDA Director Robert Califf warned last year about the healthcare industry being “swept up quickly by something that we hardly understand.” In March, Dr. Califf spoke at the Coalition for Health AI and raised similar concerns. Califf said, “My concern is that our health systems do not have the infrastructure and tools to make the most important determinations about whether an AI application is “effective” for health outcomes. The ability of algorithms to provide accurate assessments will drift if left untended, often in unpredictable, and sometimes dangerous ways.”

Califf added that it is the responsibility of the manufacturer to ensure products are safe with oversight from the FDA. He said it is unclear how this system can work effectively for AI models that change and adapt over time.

“Not surprisingly, I’m hearing that the “effectiveness” metric being used by health systems to make decisions about incorporating an AI implementation is a financial metric,” Califf said. “Will the algorithm improve the bottom line of the part of the health system making the purchase?  I worry that the main use of AI algorithms will be decisions that optimize the bottom line rather than optimizing the longevity and well-being of patients.  This is counter to the mission of the FDA, where effectiveness means an improvement in a health outcome.”

The HighWire has been reporting about the ongoing concerns with AI, including a conversation between Jeffery Jaxen and Host Del Bigtree about whether AI is an existential threat. In an earlier episode, the Jaxen Report included a warning by Professor Stephen Hawking that AI could destroy mankind. In June, Jaxen discussed AI technology that utilizes human brain tissue.