In conversations with Baku Patel, former Chief Digital Health Officer Global Strategy and Innovation at the FDA, he indicated that the FDA’s lighter touch for AI solutions is informed by the fact that these are not static technologies and by ensuring safety (as an assistive device) and acceptable efficacy, we can get these technologies in the hands of the users to assess how they perform in the real-world and receive the type of feedback that would help shape regulatory policy moving forward. He explained that thresholds used for clearance are matched to the claims and risks for those products.
The De Novo process (another pathway offered by the FDA for such solutions) provides a pathway to classify novel medical devices for which general controls alone, or general and special controls, provide reasonable assurance of safety and effectiveness for the intended use, but for which there is no legally marketed predicate device. De Novo classification is a risk-based classification process. Devices that are classified into class I or class II through a De Novo classification request (De Novo request) may be marketed and used as predicates for future premarket notification [510(k)] submissions.
All software regulated by FDA is considered a “medical device” and subject to the same general class of regulations as products such as pacemakers, stents or contact lenses. This was fine when the sum total of software regulated by FDA was limited to things like MRI firmware, but software has “eaten the world.” Not only can it be iterated upon much more quickly, the pace of innovation in new types of products such as digital therapeutics and AI-powered diagnostics depends on how quickly they can be iterated upon. Current FDA medical device approval timelines range from 90 days to 180 days, and delays can double that number; a timeline that is incompatible with modern software development.
Thus, Pre-Certification is yet another avenue for healthcare software products to get clearance. Instead of requiring a thorough review of the company and its product for each of its product iterations, FDA proposes to certify a company and its software development process for conformance to Principles of Excellence. Once certified, company products that fall below a certain risk threshold can undergo a “streamlined review” or no premarket review at all. (Figure 1)
FDA touts the future ‘Streamlined Review’ and ‘No Review’ benefits of the Pre-Cert Program, in particular that they will be more agile and result in patients getting access to new SaMD products and updates to SaMD products in a much faster and more efficient manner. The speed of updates, particularly those that add risk control measures, could be as fast as the company has them ready for deployment (e.g. weekly, monthly). This vision has always been in essence the promise of Pre-Cert, allowing software teams to release FDA-regulated software as fast as they develop it. Ensuring the ability of that timeline for low-risk modifications to products is a key tenet of the Pre-Cert Program
A framework for modifications to AI/ML-based SaMD that relies on the principle of a “Predetermined Change Control Plan.” The SaMD Pre-Specifications (SPS) describe “what” aspects the manufacturer intends to change through learning, and the Algorithm Change Protocol (ACP) explains “how” the algorithm will learn and change while remaining safe and effective