Off-Label Marketing of Artificial Intelligence Based Software Devices!
(Thursday, July 6, 2023)
In the last 5 years, the FDA has cleared about 120 software-enabled medical devices via the 510k pathway for various indications. A recent study of these discrepancies indicates that manufacturers of such devices might be exploiting a loophole in the current device regulation. A review of the 510k summaries and the subsequent marketing information of these devices indicates discrepancies between the claims included in the 510k application and that in the marketing material.
To be cleared for the market via the 510k regulatory pathway, a manufacturer needs to only demonstrate substantial equivalence between their new device and a device previously cleared by the FDA under the same regulatory pathway, the predicate device. This regulatory pathway is available for moderate risk devices such as artificial intelligence (AI)-enabled devices. The 510k application does not contain the final marketing material and the manufacturers are not required to get pre-approval of their marketing material from the FDA. Post-clearance, to differentiate their products from competition (mostly from the predicate devices), manufacturers frequently include additional information or claims about their products which may be considered off-label promotion. FDA usually does not have the resources to evaluate all marketing materials for all 510k devices.
The study published in the Journal of American Medical Association (JAMA) found disparities between statements made in the 510k summary for AI-enabled medical devices and subsequent marketing materials. Out of 119 devices analyzed for the study, about 1 in 5 exhibited different or ambiguous marketing claims compared to the representations made in their FDA applications. Many devices touted AI or machine learning (ML) capabilities that were conspicuously absent from the FDA applications. The authors did not speculate if these discrepancies were intentional or not and attributed these to deficiencies in the FDA requirements for the 510k applications for such devices and enforcement.
There is a valid concern about moderate risk devices cleared for market via the 510k pathway making off-label, dubious, ambiguous, false, or misleading claims. The FDA has so far applied traditional regulatory standards to such devices which do not adequately address the dynamic features of these devices. Although the study found that in 80% of such devices, the claims in the 510k summaries and the marketing materials matched, the fact that 20% seem to have discrepancies is concerning.
First, it highlights an obvious loophole that the manufacturers of such devices can exploit financially. Secondly, informed consent and proper utilization of medical devices can only be achieved if developers provide essential information about how these devices operate and their potential applications. Multiple surveys have shown concerns both from patients and physicians regarding the reliability of AI-enabled devices. While both believe such devices could markedly improve healthcare, they are concerned about adequate regulatory review, and transparency from the developers.
The JAMA study presents a high-level review from the physician’s perspective. It highlights the loophole in regulatory process. While most devices seem to be compliant, there is a significant number of devices with discrepancies. As the FDA's oversight of AI and ML devices intensifies, it is imperative to address these issues to safeguard patient safety and promote responsible innovation in the field of medical technology.
Dr. Mukesh Kumar
Founder & CEO, FDAMap
Linkedin: Mukesh Kumar, PhD, RAC