On March 5, ASA submitted comments expressing significant concerns with the Food and Drug Administration’s (FDA) request for information on the permanent removal of the 510(k) pre-market review process for certain class II medical devices. Artificial Intelligence (AI) and Machine Learning (ML) technologies, as well as non-AI driven device products that play a critical role in life support are among the devices the FDA proposed to be permanently removed from the premarket notification process.
Specifically, ASA’s comment letter requested that the FDA continue to require new device products to demonstrate “substantial equivalence” to those already approved, and explained how new market entry of devices, without any means to credibly verify performance, imposes risk to human health and safety. In addition, ASA described several features of current AI and ML that may have adverse effects on how medical care is delivered. For instance, an ML algorithm heavily weighted toward one gender or race may perform much worse when applied to a population heavily weighted toward an opposite gender or applied to a more heterogenous population.
In the comments, ASA stated that “[ASA] cannot support permanent removal of the 510(k) premarket notification requirements for AI-driven products,” especially in knowing “the complexities involved in assuring the proper training and validation of ML algorithms on target populations, the lack of easily identifiable failure modes, and the possibility that recommendations and inputs made by these algorithms will be given more weight than they deserve.”
During the pandemic, regulatory flexibilities removed the requirement for certain medical devices to be subject to the 510(k) premarket review. However, the ASA expressed that these flexibilities should not extend beyond the public health emergency or become part of the HHS/FDA’s permanent regulatory framework.
Click here to read the ASA’s letter to the HHS.