This article has been updated to include additional background details about Translational Software's regulatory experience and its current base of operations.
NEW YORK – Translational Software has failed to obtain 510(k) clearance from the US Food and Drug Administration for its pharmacogenetics decision support software and is shutting down its services by year-end, which will affect dozens of customers, GenomeWeb has learned.
In February 2020, the Mercer Island, Washington-based PGx software company made its first 510(k) submission for PGxPortal, a component of its Precision Health Nexus platform, and made the final submission in October 2022. After Translational Software's laboratory customers test patients' samples for genetic variations associated with drug response, they can run the detected variants through PGxPortal, which assigns a diplotype and generates a report flagging treatments that patients might respond well or poorly to. The firm provides this treatment-related information based on FDA-approved drug labels, the published literature, and recommendations from expert groups, such as the Clinical Pharmacogenetics Implementation Consortium (CPIC).
When seeking 510(k) clearance for a device, sponsors must demonstrate it is "substantially equivalent" to another legally marketed product. In its application, Translational Software tried to demonstrate that PGxPortal was as safe and effective as 23andMe's FDA-authorized Personal Genome Service (PGS). However, the FDA informed the company over the summer that PGxPortal was "not substantially equivalent" to the PGS.
"As a result of this determination, we have submitted a plan for withdrawing our services from the market" by year-end, Translational Software CEO Don Rule told its customers on Monday.
The FDA's decision to deny 510(k) clearance to PGxPortal comes as the agency is trying to oversee lab-developed tests and certain software used to guide treatment. Translational Software's regulatory experience provides a glimpse into the FDA's evidence expectations for such clinical decision support software when it uses data from lab-developed tests that also don't have FDA approval or clearance.
The FDA had originally approached Translational Software about garnering authorization for its platform as a De Novo device in 2016, and the firm submitted an application in December of that year. But around the same time, the 21st Century Cures Act passed, excluding certain software functions from the statutory definition of a medical device. The FDA told the company that its application was no longer of interest even though the agency deemed the firm's software a medical device.
Then, in 2019, around the time the agency was cracking down on PGx laboratory testing services that were on the market without its authorization, Translational Software reached out voluntarily to resume regulatory discussions. This was a year after the FDA had issued a safety alert cautioning that companies were making unsubstantiated claims about PGx tests and associated decision support software. The agency subsequently sent a warning letter to Inova Health System's genomics lab for conducting PGx tests without its approval and began calling up other companies with unapproved PGx products, asking them to scrub any mentions of drugs associated with PGx variants in their test reports.
At that point, the FDA told Translational Software that it had changed its position and that the firm's software and associated PGx content had to be submitted for review. The FDA also indicated then that while it is aware of CPIC's work, the consortium's recommendations "are not part of the safety [and] effectiveness evaluation process of PGx information that is or should be contained in the label," Rule recalled.
Because of the FDA's actions, Inova decided to stop providing PGx testing entirely. However, other firms pushed back, questioning the FDA's authority to regulate lab-developed tests. The FDA has traditionally not regulated tests developed and offered by a lab if that lab meets federal standards under the Clinical Laboratory Improvement Amendments (CLIA). However, in recent years it has tried to bring LDTs under its oversight, through guidance, legislation, and, as of last Friday, via rulemaking. The FDA also finalized guidance last year on clinical decision support software, in which it asserted its authority to regulate software used in therapy selection.
Even though many labs protested the FDA's efforts to regulate PGx products, Translational Software proactively engaged in the 510(k) process. The firm submitted more than 5,300 pages of information to the FDA, according to Rule, including around 2,300 pages on product validation and 2,400 pages on performance characteristics.
Despite these efforts, however, Translational Software and the FDA couldn't agree on a few key issues, he said. First, the FDA wanted Translational Software to "guarantee the analytical validity of the data [it] received from the labs." It was not enough for the FDA that Translational Software's lab customers had CLIA certification and accreditation through the College of American Pathologists, Rule said.
As such, Translational Software spent a lot of time trying to find ways of supplementing the analytical validation that partner labs had already done under CLIA and CAP. For example, the firm proposed using the quality scores in files that labs submitted, but this wouldn't account for sample contamination, Rule said. Ultimately, "we could not come up with a way that worked for all of the technologies" generating data that the software platform analyzes, he noted.
Very few PGx tests have garnered FDA authorization, and therefore, Translational Software's PGxPortal pulls in data from tests developed by CLIA-certified or CAP-accredited labs that don't have the agency's approval or clearance. The FDA's proposal now to regulate LDTs through rulemaking and reclassify them as medical devices will have profound implications, Rule said. Some labs may be able to expend the effort to convince the FDA about the analytical validity of their tests, he noted, but providing pharmacogenetic evidence will be more challenging.
The FDA also wouldn't accept that Translational Software's platform flagged treatments associated with PGx variants detected in patients based on CPIC and other expert groups' recommendations. The company spent "a lot of time and money" defining its procedures for evidence review and "going back to primary sources to justify our recommendations for 27 drugs, but still could not meet [FDA's] evidence bar," Rule said.
Translational Software hired another firm to review the evidence it had gathered on variants and adjudicate any differences in conclusions about their clinical importance. Although the FDA did not consider this sufficient, Rule pointed out that ironically, the US Centers for Medicare & Medicaid Services accepts CPIC evidence for reimbursing PGx tests, but "the same evidence is not considered sufficient [by FDA] for reporting" to patients.
Labs have criticized the FDA because its evidence bar for PGx claims differs from expert bodies, like CPIC. The consortium of PGx experts has issued dozens of guidelines on PGx variants since it started in 2009, and labs have increasingly applied them to design their tests.
Lastly, Rule acknowledged that Translational Software was in the midst of rebuilding its entire team after the COVID-19 pandemic when it made its final submission for the 510(k) application in October 2022. Because of this, the company failed to sufficiently explain the validation of its software, and Rule said the FDA raised "legitimate complaints" about this.
"In retrospect, we should have outsourced the validation process to a company that had successfully filed before," he said. "But given the time and money we spent on the other two areas, I am not sure that we could have done this." While Rule didn't have a reliable estimate for how much all the work the company put into the 510(k) process cost, he said consulting and legal fees amounted to $860,000 since Translational Software engaged in the regulatory process in 2019.
Given Translational Software's years-long history trying to meet FDA's regulations, Rule said the agency might take the position that the company just couldn't meet its evidence bar. Rule is of the view that Translational Software could have met FDA's bar, but the company simply could not keep up with the agency's expanding regulatory scope. Even if the company could have addressed the FDA's concerns about its software validation, there was no way to work around the other points of disagreement, he said, especially the agency's unwillingness to accept data from CLIA-certified labs and treatment information in test reports backed by CPIC and other expert groups.
Given this outcome, Translational Software has no option but to stop its services. Rule has proposed Dec. 31 as an end date, though he said the FDA may ask the firm to accelerate that schedule. Rule had hoped to merge with a larger company with more regulatory experience, but the lack of revenue and an adverse decision from the FDA made this impossible.
According to Rule, roughly 98 customers using Translational Software's PGxPortal will be affected. "The first priority is to get the message out to our customers, and the second is to try to find homes for employees," he said of the firm's 11 remaining staff members. "Then, we will try to figure something out" about the company's software assets.