By Kevin Truong, MedCity News | April 24, 2019
At the MedCity INVEST Conference in Chicago, the FDA’s Bakul Patel spoke about the recent developments in the agency’s thinking about regulating the emerging Software as a Medical Device space.
The emergence of new data sources and the ubiquity of cloud computing has meant that the speed and rate of healthcare technology development has exponentially increased. Developers also have the opportunity to proactively modify or update technology that’s already on the market.
Those new realities pose a challenge for regulators looking to balance new innovative solutions that can improve health outcomes, with concerns about patient safety.
At the MedCity INVEST Conference in Chicago, Patel spoke about the recent developments in the agency’s thinking about regulating the emerging Software as a Medical Device (SaMD) space.
Patel has been one of the key architects of the FDA’s Digital Health Software Pre-Certification program. FDA launched a pilot of its pre-cert program in 2017 with nine companies including Apple, Fitbit, Pear Therapeutics and Samsung to model a different approach to regulating digital health tech with leading healthcare software developers.
The idea is that instead of approving products in a piecemeal manner, pre-certification program participants would be given a “Excellence Appraisal” which could streamline much of the quality requirements necessary under the traditional De Novo pathway.
Additionally they would be required to include a plan to monitor and evaluate real-world performance of the technology.
One of the figures on the vanguard of building the framework of those determinations has been Bakul Patel, the director of the FDA’s digital health division, who has been instrumental with helping to develop an alternative regulatory pathway for software-based health technologies.
“No regulator in the world could keep with with the volume of new software being created,” Patel said. “The pre-certification is really taking the oversight in two different directions. One is upstream and understanding where the company is and how excellent they are. Two is how do you move from this concept of quality systems and compliance to excellence?”
Earlier this year the FDA released more details on the model and how the agency plans to test the program. One area mentioned by Patel as a particular area of focus is continuous learning software and how FDA thinks about issues like data set size and accuracy, the explainability of algorithms and issues of potential bias.
In a recently released discussion paper, the agency proposed a framework which would require premarket review for software updates, including in cases where the modification introduces “a major change to the SaMD algorithm.”
This comes along with increased transparency into the real-world performance of these technologies to ensure continuing safety and effectiveness.
According to Patel, the idea is to shift to a complete lifecycle approach for regulation that moves the system from “episodic to more continuous, which relies on getting the right information at the right time to make sure we are promoting public health.”
For digital health companies looking to potentially advance products from the unregulated wellness space to the regulated healthcare space, Patel had the advice to be proactive in data collection and evidence gathering.
“If you’re moving in the direction of trying to be a valuable tool in healthcare, I would collect the information so I don’t have to go back and re-do the clinical trials and spend millions of dollars after you realize you’ve got something on your hands,” Patel said.
When it comes to the digital therapeutics space, Patel said what’s necessary is building a stronger base of data and evidence so that products and applications actually have a comparison point with which to judge effectivness.
In response to a question about putting something comparable to an airplane’s “black box” into machine learning systems, Patel said oversight for this type of technology needs to go past just logging events.
“In the machine learning world we not only want what’s happening, but we also want to know what the next outcome of the software will be in the bounds of what we’re thinking,” Patel said. “It’s analogous to the power-on self test that your computers have today.”
Even with the recent departure of Scott Gottlieb from the FDA’s top job, Patel said he was confident in the organization’s support of the pre-cert program and called for industry to be stronger advocates as well.
“If you guys think this is valuable you guys should support it too, don’t wait for somebody else. From my perspective we’re not building something that the agency wants, we’re building something that the industry wants,” Patel said.
Ultimately, Patel sold a solid and consistent regulatory framework as key to keep patients safe as well as keeping industry moving steadily in the right direction.
“Speeding up regulations doesn’t really mean anything,” Patel said. “If we don’t normalize the process then we have chaos and we don’t have good products out in the marketplace. Speed is an outcome of efficiency in the regulatory system.”