A discussion paper was posted this month to request feedback on a proposed regulatory framework for AI/ML SaMD. The current regulatory framework is expressed in the guidance “Deciding When to Submit a 510(k) for a Software Change to an Existing Device”; the purpose of the paper is to propose a more flexible approach and solicit feedback on a number of questions. The proposed new framework relies on the total product lifecycle and the agency’s expectations for a quality system to enable certain types of software modifications driven by learning and adaptation from real-world data acquired through clinical use of the approved product. The basic framework would apply to SaMD requiring some type of pre-market review process. During the initial pre-market review, the manufacturer would propose a list of anticipated modifications to performance or inputs or changes related to the intended use of the AI/ML SaMD. This list is called the SaMD Pre-Specifications (SPS). The manufacturer would also submit the Algorithm Change Protocol (ACP), or the list of methods to be used to achieve the type of modifications listed in the SPS. The ACP will use a combination of Good Machine Learning Practices (GMLP) and the operation of the manufacturer’s Quality Management System to verify, validate, and implement the proposed software changes. Following pre-market clearance or approval, modifications that fall within the approved SPS+ACP could be completed and reported to the agency without a formal review process. 

This proposal is at an early stage, and the report cautions that additional statutory authority may be necessary before this type of process may be implemented. Feedback from interested parties (including patients and providers) may be submitted at docket FDA-2019-N-1185 until 03JUN2019. More information may be found at the Artificial Intelligence and Machine Learning in Software as a Medical Device web page.

Text Copyright © 2019 Katrina Rogers