The Learning Portal will be under maintenance Monday, 6 December between 6 AM and 5 PM EST. Portal functionality will be unavailable during this window.
We apologize for any inconvenience caused during this time.

Regulatory Focus™ > News Articles > 2019 > 3 > Industry Reiterates Concerns on FDA’s Software PreCert Working Model

Industry Reiterates Concerns on FDA’s Software PreCert Working Model

Posted 13 March 2019 | By Ana Mulero 

Industry Reiterates Concerns on FDA’s Software PreCert Working Model

The final round of comments on the development of the US Food and Drug Administration’s (FDA) digital health software precertification (PreCert) pilot are in, with some new submissions reiterating concerns facing industry.

FDA posted the latest batch of PreCert comments earlier this week and AdvaMed, the American Medical Association (AMA) and GE Healthcare are among those with new submissions. Comments have been submitted since 2017, showing general support for FDA’s Center for Devices and Radiological Health (CDRH) efforts for a new streamlined approach to the regulation of digital health.

But several commenters’ concerns remain, despite the working model now in its third version, relating to the disconnect between the risk classification framework established by the International Medical Device Regulators Forum (IMDRF) about five years ago and the current US regulatory paradigm. Other questions have centered on the application of FDA’s excellence appraisal at the manufacturer level and the need for additional legislative authority to fully implement PreCert.

The intended benefits touted by CDRH that have garnered industry support are centered on fast-track reviews of digital health tools, reducing burden, increasing access and fostering innovation.

AdvaMed vice president of technology and regulatory affairs Zachary Rothstein stressed “we strongly believe that the IMDRF framework must be adapted to the US regulatory paradigm.” This has been a concern for AdvaMed and member companies since 2017. But AMA executive vice president James Madara argued the IMDRF references are inappropriate due to a lack of consideration regarding machine learning and artificial intelligence (AI) and the unique risks that they pose compared to others.

“While a precertification program could be appropriate for AI-enabled systems at some point, the FDA has not established foundational definitions, risk frameworks, or excellence categories for AI-enabled systems, particularly those utilizing machine learning,” Madara said. “AMA urges FDA to establish accepted nomenclature and definitions and appropriate risk stratification for AI systems.” The group also “strongly urges” FDA to “strike all references to AI systems and machine learning” pending resolution of these issues.

AdvaMed, AMA and GE Healthcare also called for further clarity around how the agency intends to implement its excellence appraisal, including the scope and rigor of the initial appraisal.

There is also consensus around hindering the potential growth of the PreCert program post-pilot because of the currently limited scope. CDRH indicated in January that the PreCert test plan for 2019 will be limited to first-of-its-kind software as a medical device (SaMD). Rothstein noted that AdvaMed has supported this decision to limit PreCert to SaMD for the time being. But the finalized PreCert program “must also benefit software in a medical device,” Rothstein argued.

A long-standing issue in the PreCert journey relates to current statutory authorities. Now that the 2019 test plan is out, GE Healthcare digital health regulatory affairs director Camille Vidal called on the agency to clarify whether additional authority is needed for the “No Review” pathway and to offer 510(k) exemptions for changes to SaMD products approved via PreCert.

Some concerns emerged more recently with iterations of the working model over the past few years, following the 2017 announcement that kicked-off CDRH’s co-creation process in collaboration with the nine pilot participating companies. These relate to achieving streamlined review without early engagement with FDA for applications that contain complex clinical performance data, the two levels of certifications, descriptions around the real-world performance analytics (RWPA) and the current lack of review timeline. Commenters suggested potential solutions and clarifications to these concerns as well.

Madara reiterated that AMA is concerned with the creation of level 1 certification for entities lacking prior experience in SaMD development and deployment. “Beyond the clear risks posed to patients such an approach presents, it undermines FDA’s intention to promote this model to physicians as one that is based on excellence which suggests a track record of proven performance—not just policies and processes,” Madara argued. Vidal suggested FDA adopt a mechanism for entities to advance from level 1 up to level 2 as appropriate.

On RWPA, Rothstein argued that the working model’s descriptions “require greater clarity and transparency.” The Center for Digital Health Innovation at the University of California San Francisco pushed for adding four “especially important” subdomains to the current RWPA framework under PreCert. These relate to “real-world usage,” “workflow,” “interoperability” and “universal design.”

AdvaMed recommended FDA adopt a timeline for review of applications from PreCert participating manufacturers that can “fully articulate the benefits of the program.” This would involve a requirement on FDA staff to provide the manufacturer with a final decision summary and letter of substantial equivalence within five business days of confirming document receipt. Lack of clarity on how FDA will achieve the intended benefits is an ongoing concern facing the Medical Imaging & Technology Alliance.  

 

© 2021 Regulatory Affairs Professionals Society.

Regulatory Focus newsletters

All the biggest regulatory news and happenings.

Subscribe