Digital Therapeutics for Better Health with Dr. Smit Patel: The Tech Between Us, Season 3, Episode 8, Part 2

Raymond Yin
Welcome back to the Tech Between us. We're continuing our conversation about Digital Therapeutics with Dr. Smit Patel, Associate Director at the Digital Medicine Society. To catch up on part one of our conversation, visit our Empowering Innovation Together page. "
Let's go back to diabetes for just a second. I've seen the Dexcom commercials. There are even a few people here at Mouser who have the little Dexcom sensor. Do all therapeutics have a hardware and a software component? Or is it primarily one or the other?

Smit Patel
That's a really great question. It depends if I'm talking from industry side versus a regulator side. So digital therapeutics as a term that has been coined with Digital Therapeutic Alliance, an organization, trade association that have brought together multiple stakeholders. And now ISO, which is the International Standard Organization, is also defining that is a software driven product. But also, software has to land somewhere.

It will have a hardware platform that it sits on. However, digital therapies are still combination of that. So let me walk you through how industry sees us, like hardware products and software products. And how regulators see more on the function language. It's important to see with new fields that we are building hardware components, and then software components. Hardware components being smart scales, wearables, smart products, any kind of smart products that goes on hardware mobile platforms versus software components being applications that sits on those hardware components. The question is, out of this solution, which is the digital health solution that may or may not have hardware and software components, or sometimes can have multiple of those. I'm thinking of a wearable sensor, if I have a wearable watch, and then there's a software component that connects all these three platforms that's like two hardware products, one software, two software products, which is a whole digital health therapy subset of a solution.

How FDA see is essentially based on functions. So, what is a device function and what is another function? Device function is the one that's the main component that does the actual job. Meaning it can meet the requirements of a medical device according to FDA versus an ancillary component is something that does not meet a medical device definition but may have some implications on the actual main component device function that can lead to a risk in a patient. So, say for example, and this might be a little complicated, but gives a clarity on what good those functions are. If I am building a transcutaneous electrical nerve stimulation device, it has a mobile platform, which is a variable for a user to treat pain. Electrical nerve stimulation, which has a variable component. There are two hardware device functions and there's also software for app that regulates level of stimulation. The device function or the main function becomes, that means electrical stimulation that's used for treatment because it's claiming that it's going to treat the pain. And second device function is this software app that controls the level of stimulation because based on the level of stimulation, it will treat a pain, or it will not treat pain. So those two becomes the main components or main functions, device functions of the product versus the other function is that mobile platform, which has the Bluetooth transceiver and connectivity, because Bluetooth platform, even though it connects all these solutions it is not the one that's treating, but the reliability and the security of that other function can have impact if, say, for example, the transmitter was compromised and the connectivity is compromised. The app can't do the level of stimulation what it was supposed to do. So that's the complicated version of it could be hardware, it could be a software. It depends on what that hardware and software is doing.

Raymond Yin
Got it.

Smit Patel
Does that clarify? It was a little complicated.

Raymond Yin
Yeah, so the actual hardware has to be the one doing whatever the therapy requires, like in the electro stimulation, it's got to be the wearable or the device that actually does the stimulation that provides the charge to the body. The software app actually controls that. But like you said, you've got different communication platforms that tie it all together, but those aren't necessarily directly related to the therapy itself.

Smit Patel
Correct. In this case, yes.

Raymond Yin
As an FAE, this is going back many, many lifetimes ago. I used to call on a company here in Dallas called Abbott Labs. One of the big medical companies out there and working with their engineering teams. One of their biggest headaches and their biggest hurdles was actually the FDA process - the approval process, the submittal process. Now, does digital therapeutics go through the exact same process or has that process been established for digital therapeutics?

Smit Patel
I think the process is a little bit simpler for digital therapeutics.

Raymond Yin
I think that would be good for a lot of engineers.

Smit Patel
A hundred percent because digital therapeutics are more software driven. And according to the FDA's infrastructure, based on risk, and we can dive into a little deeper if that would be of interest. Digital therapeutics falls under either class one or class two medical devices based on the risk that it has. And based on that, they will have either a 510K pathway or a de novo pathway, which is more of cleared or granted product other than the older generation hardware, which had to go through pre-market approval. Which took a long time, years, a lot of years of data, years and clinical studies, a lot of specific requirements to go through that because based on the risk. So digital therapeutic products have 510K pathway, which is the most common pathway for that meaning because it's a software base, because it's a risk is a lower compared to rest of the product, they will have a predicate that they can identify in the market.

So, say for example, I'm building a software-based app for ADHD. And there is some other, already cleared, in FDA database previously cleared another software app from a clinical standpoint, from a tech standpoint, which they call it comparing it to a predicate, meaning comparing it to something that's already in the market. If that is true, an innovator can submit for a substantial equivalence, which is saying essentially, I'm building this product, but it's similar to something that is in the market. I don't need to go through the whole level of testing. I don't need to go through the whole level of requirements for that new digital therapeutic product so they can potentially go through a 510 K pathway and get the clearance much economical and much faster way.

Raymond Yin
That's the key, the speed.

Smit Patel
I will preface this, yes, it is a faster pathway, but it depends what the intended use or depends on what the indication of use of the product is and what risk it poses to the patient.

Raymond Yin
So, an innovator in digital therapeutics can almost literally stand on the shoulders of those who came before them, even from an FDA perspective, as well as a technological perspective.

Smit Patel
Yes.

Raymond Yin
As long as they can prove that it's similar to something that's already been approved, they can go through an abbreviated process for FDA.

Smit Patel
Yeah. And 510K process have been in place for many, many years.

Raymond Yin
So, this is new for digital?

Smit Patel
So digital therapeutics, so how FDA has built this infrastructure is 1976 when there was the new Congress act that gave FDA the power to regulate medical devices. At that time in 1976, the regulation of medical devices was a little different. They were MRI machines, these large hardware components that were being regulated. So go 1990s and early 2000s, there was an advent of software-based products. The bubble of WWW came in. And since then, we have seen a lot more software-based product that came to the market. So, these digital therapeutics, very, very new, are still software driven. So, it falls under the infrastructure of that medical device regulations, how FDA does. The traditional route is 510K if they fall under class one and class two of the devices. And there's also a new pathway that FDA came out to, which is called de novo.

Say for example, a digital therapeutic, went through the database, looked at it couldn't find a predicate to compare to, can submit a de novo pathway, which is similar, but needs a little bit additional details for FDA because it has a unique intended purpose, which is a unique product compared to what's already in the market. So, FDA has been open to creating those, not faster or accelerated, but rather creating pathways that can help innovators go through the regulatory process more efficient way.  So, de novo Pathways another one that innovators can take, unless they are class three high risk products that has to go through the pre-market approvals, which takes some time and some labor of love and some money.

Raymond Yin
Like I said, the design of the product may have taken six months to a year, year and a half, but the approval process would take two or three times as long to get all the clinical data in. We've talked a lot about therapeutic software. Is there a special category, a special consideration for digital therapeutic hardware? Not the MRI, not the surgical equipment obviously, but things like wearables and sensor patches. Are those generally a faster path through FDA approval?

Smit Patel
That's a really good question, Raymond. I might change the narrative here because that's how the FDA's approval processes are. Digital therapeutics, again, falls under as a medical device categorization so be it hardware, be it software.

Raymond Yin
Okay. So, it's for both?

Smit Patel
It would depend on what it does. So, what is the intended use of that hardware? Or what is the indication of use for the software? For example, what specific condition that hardware product or software product is addressing, what age group it is addressing, what patient population addressing, what are some technological features it has that may or may not incorporate a risk of a product? Say smart mat, a hardware product. A smart mat that does remote temperature monitoring for patients for diabetic foot ulcers. There's a product out there in the market, that product has risk levels that are low risk levels.

It's external, it is non-invasive. One therapeutic in therapeutic intervention of prevention of diabetic foot ulcers, because five weeks in advance, it gives you that data, but at the same time, that product is class one. So that product, according to FDA, due to the low risk, is identified as class one. But then compared to a cardiac AI algorithm that does imaging for patients with heart failure, there's a little higher risk because if those analysis of specificity of those images are not what it claims for an adjunct assistance to a provider, then that leads to a little bit more risk. Imagine a heart failure patient has, by chance .001 chance, that it did not detect in a patient.
Can lead to death of an individual. So that's a little bit higher risk. That goes from class one to class two. Which is a moderate risk product. And almost 53% of all medical devices, digital health products fall under moderate risk.

Third is the high-risk product, usually Freestyle Libre that has been a variable sensor because there's a microneedle that goes into it that can be attached. There are multiple different indications that increases the risk. If their software stops functioning, does not provide good readings, can elevate an individual's blood sugar levels does not go detected. And if a physician changes their therapy, they can go hyperglycemic or high hypoglycemic, meaning their blood sugars can go really high or go really low.

That is an increased risk to the patient, means it's a high-risk patient, and that's how FDA categorizes in these risk panels. So, class one and class two, low risk and moderate risk goes through a more 510K, de novo, which is faster and a little cheaper pathway for regulatory. And then, high risk products go through those PMA pathways or pre-approval pathways that take some time for evaluation because FDA wants to make sure that those products are regulated and safe and effective.

Raymond Yin
It's really less about the fact that it's a digital therapeutic, but still falling back on the original FDA requirements of risk level to the patient, whether it's software or hardware and categorizing a device accordingly and then applying the right pathway through FDA approvals.

Smit Patel
A hundred percent. Goes back to, what is the purpose of that product? And we have, at DiMe, we did a pre-competitive project on US Digital Health Regulatory Pathways. We created over 50 resources, an interactive tool that individuals can click through. By answering a few questions, they can determine should they be regulator or not, be it hardware, be it software. Should they be regulator or not? What is their risk type? We walk them through a step-by-step process, and then at the end, it likely shows what's the right regulatory pathway for that product. Be it hardware components meet software components, but essentially walk them through the whole process because it's a little complicated and convoluted.

Raymond Yin
So, DiMe. FYI for everybody, DiMe is the Digital Medicine Society. I'm not sure if we actually called that out earlier. So, DiMe actually has created resources for entrepreneurs, for engineers, at new potential digital therapeutics companies to walk them through and help them through this complicated process.

Smit Patel
Yes. We have consolidated over 30 guidances, multiple policies, all brainchild into that one interactive tool and a flow chart that has results and walks through, and we did it together. We had med tech companies, tech companies, innovators, we had FDA at the table, as we were building this academia-like Harvard Medical Center for regulatory science experts, over 30 different experts’ brainchild into that one product. And alongside we have 50 different resources on various pathways and checklists.

Raymond Yin
That's fantastic! I know that is one of the most complicated issues for medical devices, and it sounds like you've been able to put a great tool together to help people navigate all that.

Smit Patel
Yes. So please, it's free. Use us as much as you want, and it's for all the innovators’ benefit as they're determining these really hard skating-the-line issues off non-regulated space and what to do with the product.

Raymond Yin
Smit, you had mentioned, as one of your examples, an AI based algorithm. And, in previous editions of our podcast, we've talked a lot about how artificial intelligence and machine learning, they're really touching almost everything that we do now from industrial to consumer products. And AI and machine learning are helping especially in radiology, identifying cancers. How are artificial and machine learning algorithms being used in digital therapeutics?

Smit Patel
That's a really good question. Across the board, AI is being used.

Raymond Yin
So, it's everywhere then?

Smit Patel
Yeah, it is everywhere. I'll go back to the notion that AI and algorithm is not going to treat by themselves. Again, if a software is preventing or managing a disease, then that becomes a digital therapeutic classification. So, developments are happening in the field of AI and ML, which are again, device software in the field. And it's very fun, interesting because FDA has been recognizing this across the board for the last few years, where, yes, we have seen development from the tech world. We have seen development from the healthcare world on utilization for various clinical applications, as you mentioned, cardiology imaging. For digital therapeutics, we have seen a lot of developments that are happening more on prevention and management.

And essentially our regulatory colleagues have come out with a very new guidance on AI ML enabled device that gives this new structure over the last few years that FDA has been seeing how AI is developing. So, this regulatory guidance goes through this concept of predetermined change control plan. This new idea, a next generation concept for today's concept for a next generation tool where what FDA's saying in this guidance in a very simplified format is individuals who are building AI software for medical purposes, individuals who are building machine learning algorithms that are going to be used in clinical care that impacts patients, essentially can go through FDA, submit their early plans through a PCCP. What does PCCP do or pred domain change control plan is manufacture or innovators has the ability to change and evolve in future without the new regulatory submission.

So, once they do it with the PCCP submission, it has descriptions of what will be modified in future, what are the modification protocols, what are the impact assessment of, say, AI tool once it learns more data, once it learns from information, once it changes what impact it will have for those changes. What FDA's saying with this new guidance and will start asking innovators is, you can submit it now with that plan, so that in future, you don't have to do the whole regulatory resubmission again, but rather make your product safe and transparent from now. And FDA is willing to work with that. And company called Caption Health's cardiac AI software was the first that was granted early this year with the right PCCP modules, which gives innovators the autonomy to change and do modifications to the product without worrying that I'll have to resubmit entire things what I build originally, like a year ago or five years ago.

Raymond Yin
As their models evolve and grow, based on data and improvements. They don't have to completely start over again. They can just slot that in based on this PCCP program.

Smit Patel
Exactly. And I will say it would depend again on modification. An innovator has to submit what kind of changes are anticipated. It ties back to if those risks in future would be very high risk. And if you change from that plan, if you deviate from that plan five years from now, yes, FDA can come back to you and ask you to reassess, reevaluate, or resubmit. However, if you stick to that plan that was discussed with FDA, and this is the more collaborative version or way where we are going, industry and agencies are working together to make sure that we are building these products out in the market, simplifying the processes for innovators and developers, and sharing those plans early on, on what that future state looks like. So that once regulated and gone through that regulatory process, innovators may not have to do it again and again and again every time they modify a product or device.

Raymond Yin
It sounds like machine learning will eventually be an integral part of digital therapeutics in the future.

Smit Patel
Yes. I can foresee a lot more than digital therapeutics and digital medicine approach as incorporating AI.

Raymond Yin
Let's shift our focus towards the future. Our next question is brought to you by our sponsor, partner Bourns, a leading manufacturer in semiconductor and system level solutions. Explore more from them by visiting mouser.com/bourns.

Speaking of looking forward, I'm going to ask you to look into your crystal ball. What areas of digital therapeutics do you think are going to be innovative and revolutionary in terms of either applications or methodology? What do you see for the future of therapeutics in the overall scheme of digital medicine and digital healthcare?

Smit Patel
Ooh, a lot of them. I'll try to break it down into my top three that I see.

One, definitely is measurements. So digital medicine measurement products, which digitally measures information about our body, physiological functions. We have tapped into information that we didn't have before. Think about driving a car with no sensors of like oil or brake or how fast you're going. You don't know where you're going to end up. That was health until we had sensor products, until we had that, we were actually measuring your CO2 or oxygen saturation levels, your accelerometer levels that now you can measure for specific disease or disorder. So, I think digital measurement products will help us define our health more effectively. Second is around AI ML. The potential of AI and ML algorithms is integral in terms of efficiencies across the board, clinical applications.

There's a lot that doctors cannot do. AI will be a good supportive tool to help address some of those inefficiencies that exist in the system. I'm really excited that we can start using now in US. Prior authorization takes forever to write down, to submit to a payer, to get a reimbursement for a patient's drug when it's legitimate. AI can help, instead of typing down all the notes that doctor has to go through every 30 minutes, they can use the time to essentially build the relationship with the patients and be that friendly counterpart to the patients that they need. And the third, Raymond, do you know how long it takes to get a doctor's appointment in Boston? If it's a specialist appointment of cardiologist? It is 72 days you had to wait In Boston.

Raymond Yin
It is? Oh my gosh!

Smit Patel
Even longer in rural communities where there are no specialized clinicians available. These digital therapies, one that can be scaled across massively through just our handheld device has a huge impact on what and how we can do it. So essentially digital therapeutics can be provided to a lot more patients a lot faster, in a lot of efficient way. I foresee being utilized in multiple different therapeutic areas from mental health to even wellness and prevent prevention benefits.

Raymond Yin
Thank you for listening to this episode of The Tech Between Us. This podcast is part of Mouser’s in-depth look at Digital Therapeutics, explore our entire Empowering Innovation Together content series for videos, technical articles, and more at mouser.com/empowering-innovation.

Sponsor
Serving a broad range of markets including telecommunications, automotive, computer, and more, Bourns leads the way in circuit protection, inductors, trimmers, and resistor components and system level solutions. To learn more, visit mouser.com/bourns.