Challenge Overview:

• FDA disagreement with predicate selection during Pre-Submission process
• Pre-Submission during COVID – limited FDA interaction opportunities available
• Navigating De Novo process for innovative AI/ML-based technology
• Designing clinical study to validate device and leverage ongoing clinical data efforts
• Complex analyses requested by FDA in information requests
• Use of data, including adjudication, from ongoing efforts to meet response timeline

Prenosis, Inc., an artificial intelligence (AI) company pioneering precision medicine in acute care, sought the support of a regulatory consultant to bring their AI-enabled sepsis diagnostic, the Sepsis ImmunoScore^™, to the US market. The device was developed through the innovative application of proprietary AI and machine learning (ML) techniques and Prenosis’ proprietary Immunix^™ platform to the complex process of sepsis diagnosis. Given the novelty of the device and the claims Prenosis intended to make regarding the device’s utility, including both diagnostic and prognostic functionalities, Proxima knew that strategic interaction and negotiation with FDA would be necessary to achieve marketing authorization.

However, Prenosis faced the added challenge of being ready to initially approach FDA in March 2020, when FDA significantly limited interactions as resources were allocated towards the pandemic response. Furthermore, given the novelty of the device and technological advancements associated with the product, FDA suggested Prenosis leverage the De Novo review pathway as there were no legally marketed devices suitable to serve as a predicate. To support this submission, Prenosis was also required to submit clinical validation data, gathered through a complex study design that included hundreds of patients and a physician adjudication process. While this burden was more substantial than the 510(k) pathway that is more commonly used to regulate Class II devices, the De Novo pathway would allow Prenosis to establish special controls that would be used to regulate and mitigate the risks of all future products with similar intended use and technological characteristics.

Given the challenges Prenosis’ regulatory pathway posed in getting the Sepsis ImmunoScore™ to market, Proxima recognized that a deep understanding of the sepsis diagnosis process as well as AI testing and characterization would be critical to an effective and comprehensive submission to FDA. Throughout the submission drafting process, Proxima took deep dives into the details of pertinent statistical analyses and how they were applied to the diagnosis paradigms. Additionally, Proxima provided thorough review of the clinical investigation report to confirm that this critical piece of the submission clearly communicated the design, conduct, outcomes, and interpretation of the clinical validation study that Prenosis conducted. This resulted in thorough and well-crafted documentation included within the De Novo submission, with the goal of minimizing potential confusion or questions during FDA’s review. Proxima also deployed an in-depth QC process, involving several levels of team members, to ensure that details were consistent across documents, putting forth the cleanest submission possible to facilitate a streamlined FDA review.

During the initial review stages, FDA issued a series of quick, interactive review requests to initially clarify the content of the submission. Proxima’s approach was nimble and flexible in receiving these requests, with the goal of efficiently relaying Prenosis’ responses while still maintaining high quality documentation. FDA then issued a formal request for additional information that included several highly complex analyses, including a request for additional clinical data, which were deemed necessary to help the review team understand and characterize both the risks and benefits of the novel product. An interaction of this type was expected given the use of the De Novo pathway as FDA had never reviewed a similar product prior to Prenosis’ submission, and receiving authorization for a device that included AI/ML technology and both predictive and diagnostic claims for a complex clinical condition, like sepsis, required specific details and performance characterization efforts that FDA could only request after initial review of the data. Upon receipt of FDA’s request for additional information, Proxima strategized responses to each item, brainstorming and vetting methods proposed by Prenosis to leverage easily accessible clinical data.

Given the complexities of FDA’s requests, Proxima recommended the use of Submission Issue Requests (SIRs) to gain FDA’s feedback on the proposed approaches to addressing each question. However, the nature of the requests required a level of urgency to gain feedback with sufficient time to implement it within the response timeline. As such, Proxima worked quickly to, again, pull together a comprehensive submission that clearly explained very complex analyses and provided sound justification for the proposed approaches, leveraging the in-depth technical expertise of the Prenosis team. Proxima worked with Prenosis to plan for any potential gaps that FDA may identify, so the team could plan appropriate responses ahead of time, fostering the most productive conversation possible. As a result of the strategic arguments put forth in the SIRs, FDA agreed to Prenosis’ plans, allowing them to leverage available clinical data and perform additional analyses to support FDA’s review of the device’s clinical validation.

Proxima worked tirelessly with Prenosis to prepare the final response for FDA within the required timeline, again bringing diligence and careful attention to detail to ensure the results were explained in a way that met FDA’s needs. During the final stages of the review, Proxima demonstrated agility and client prioritization in organizing and reviewing multiple overlapping responses requested by FDA within the allotted timeframes. Proxima facilitated communication with the client across multiple modalities to meet the rapid deadlines, ensuring the review team stayed on schedule and enabling a favorable review decision.

Prenosis executed highly complex analyses, supported the quick turnarounds set by FDA deadlines, and contributed detailed justifications to support their available information. The in-depth understanding of AI-enabled devices, knowledge of software requirements, and strategic approach to clinical validation data enabled Proxima to provide invaluable insights and support throughout the review process. As such, Proxima proved to be an asset to Prenosis throughout the regulatory process, particularly during the preparation and review, facilitating Prenosis in achieving the first marketing authorization from FDA for an AI diagnostic and predictive tool for sepsis. Since then, the Sepsis ImmunoScore™ received recognition from FDA, was named one of TIME’s Best Inventions of 2024, and was featured in a JAMA article written by 3 senior FDA officials, touting the device’s regulatory strategy as a practical application of a risk-based approach to regulating AI algorithms.