About the Author: Bradley Merrill Thompson is a member of the firm at Epstein Becker & Green, P.C. There, he counsels medical device, drug, and combination product companies on a wide range of FDA regulatory, reimbursement, and clinical trial issues. 

In January, FDA released version 1.0 of its Working Model for the agency’s Precertification Program for Software as a Medical Device (SaMD) (i.e. software which functions as a medical device without reliance on hardware). I actually think 1.0 is a bit of a misnomer. It ought to be labeled something like version 0.7, because it’s really about 70 percent of a complete proposal. It lacks many of the details necessary to give an understanding of how it will work.

FDA apparently intends to start filling in the remaining 30 percent during the course of 2019 as it proceeds with the pilot program, involving nine guinea pig companies selected back in 2017, and retrospectively reviews some previously submitted marketing applications for software products. So maybe by the end of 2019 we will see a true version 1.0. Companies beyond those nine will not be allowed to use this program until the pilot is over, presumably in 2020 or after.

While it is too early to fully evaluate the program, FDA has made enough progress since version 0.2 released in June 2018 that parts of it are starting to come into focus. This article summarizes the new developments, and assesses, at least at this juncture, what kinds of companies might find this program valuable in the future.  Spoiler alert: fewer than you might think.

New developments in FDA’s version 1.0

Probably the biggest piece of new information is that, at least for the foreseeable future, the agency has decided that the program will only be available for about one percent of new SaMD products. More specifically, the program will only be available to novel software products that have no forerunner in the marketplace, qualifying for the FDA’s so-called de novo review. This is the same regulatory process Apple used to get marketing authorization – just in time for its marketing event last September – for its two new apps, for use with the Apple Watch.

Limiting the scope of the program to only these novel products obviously diminishes the benefit of the program by about 99 percent. So why did FDA do it? Apparently the agency found that this was the best way to conduct the pilot program in the absence of new legislative authority. A group of senators raised concerns last fall that FDA does not have any authority from Congress to pursue this program, so for the pilot program during 2019 the agency will be reviewing the products two ways in parallel: the old de novo way and the new precert way. This has the benefit of producing some data to show how the two compare. FDA has suggested that after 2019, when the program is more fully baked, they are likely to seek statutory authority from Congress to implement it more broadly.

Beyond that, the other major new information we learned in January is the high-level contents of the streamlined review. We learned, for example, that the information reviewed will include information like the clinical algorithm, a clinical data analysis and interpretation, hazard analysis, instructions for use, labeling, product specific requirements, the software architecture and the validation performed in support of clinical claims. None of that is terribly surprising. What is less clear is the information that normally would be required for review that is no longer required under the precertification program.

Trying to understand the benefits and burdens of the program concretely, I went in search of answers to the following questions:

• What level of achievement does a company need to show in order to demonstrate in the excellence appraisal that it has a culture of quality and organizational excellence?
• How much faster will the precertification program be than the traditional 510(k)?
• Where does any enhanced speed come from?
• Specifically, which data collected as a part of the legally required 510(k) process will FDA not review at all? If FDA is reviewing all the same information that the agency otherwise would, but is simply doing it during a multistep process instead of a singular review, what benefit is there in terms of speed of review?
• Which products that would presently need to go through a 510(k) review would be exempt under the new precertification program?
• What information that FDA collects as a part of the precertification program will the agency in turn share with the public?
• What will happen as companies collect and evaluate real-world evidence, and a company reaches a different conclusion than FDA with regard to whether postmarket corrective action is required? 
• Since the excellence appraisal addresses many areas not covered by the GMPs, will FDA conclude that it has inspection authority to examine those new areas as well?
• Can companies with existing hardware products (e.g., medical imaging) create organizational boundaries between their SaMD divisions (e.g., AI/ML algorithms products) and their hardware divisions in order to allow the SaMD divisions to use different QMS processes and participate in the precert program? 

Unfortunately, I couldn’t find clear answers to any of those. If they exist, they’re buried well.

Recap of the benefits and burdens of participating in the program

Based on version 0.2, last September I analyzed the program to identify the major benefits and burdens. I’m not going to go through all of that again, but only summarize the issues at a high level.

Benefits

Remember the primary advertised benefits of the program include:

1. FDA has said that it will review fewer initial launches of new software products. Exactly which ones will be reviewed and which ones will not has not been shared.
    
2. FDA has said that for the products they will review, the review will cover fewer topics. While FDA has identified the topics they will review, they haven’t identified the topics that they will not review. It’s unclear from where the time savings will come.
    
3. FDA has said that they will review fewer changes to marketed software medical devices. They haven’t said which changes they will review and which ones they won’t.

One of the problems I have in assessing those benefits is the complex accounting that goes into FDA’s claims regarding the speed with which products are approved. For purposes of justifying the agency’s performance under the user fee agreements with Congress, FDA periodically releases data on how quickly its product reviews are conducted. But the problem is that many of those metrics focus on narrow activities like exactly how much time FDA is actively reviewing a 510(k) submission. Industry has a different interest, getting products on the market quickly.

So this can be a bit of a shell game. FDA might be saying that it can quicken the premarket review, but it accomplishes that by moving some of the activities from the “review process” to a pre-review process. FDA may be moving from a singular premarket review into a three-step process that requires (1) an initial excellence appraisal and then (2) a determination of the review path and then (3) a streamlined premarket review. Hypothetically, if previously companies got a particular type of software through an FDA premarket clearance in 150 days, and under the new precertification program the company spends 60 days getting through an excellence appraisal, then waiting 30 days for a pathway determination and then gets its streamlined premarket review process completed in 60 days, what has it really achieved in terms of speed to market? 

Burdens

From the prior article, recall that the burdens include:

1. The time and cost associated with the excellence appraisal.  Remember that FDA has said that the excellence appraisal is aspirational, and very few companies today would meet the new standard – whatever it is – for a culture of quality and organizational excellence.  Companies will have to invest a lot of time and money to satisfy this requirement.
    
2. Postmarket decision-making.  FDA has said that the price for faster approvals is more agency involvement postmarket. This means they want a constant stream of data to assure both that the culture of quality and organizational excellence is being maintained, and the performance of the medical devices in the marketplace is meeting expectations. If either of those two things disappoint FDA, the agency will have the power to demand changes to its satisfaction. Further, FDA says that this entire program depends on trust which requires transparency, so much of this information will be shared with the public as well.
    
3. Expanding scope of regulation. FDA says that this program will be so good, there will be no reason to exempt software products from FDA oversight. The agency plans to use this in lieu of exercising enforcement discretion for low risk products, contrary to the agency’s prior plans to exempt low risk clinical decision support software. FDA did an about-face on this issue when it started the precertification program in 2017.
    
4. Expanded FDA inspections. By expanding the scope of the agency’s oversight to address issues well outside of a quality system, in order to assure there is evidence of a culture of quality and organizational excellence, FDA will need broader inspectional oversight.
    
5. User fees. These excellence appraisals won’t conduct themselves. FDA will have to find resources, including potentially third parties, to do them. And that most likely means the companies will have to pay for this oversight.

With that broad overview of the benefits and burdens of participating in the program, the question is, what types of companies would benefit from participating?