Digital therapeutics: The challenges and opportunities of software-based medical devices

Digital therapeutics (DTx) is a new category of medical devices that delivers treatment or therapy using clinically evaluated software. Like any new technology, DTx presents challenges in manufacturing, regulatory approval, and other areas that require creative thinking and new strategies. Dr. Acacia Parks, strategic advisor to clinical research organization (CRO) Lindus Health, discusses the benefits and challenges of DTx and how medical device developers can adapt to an increasingly digital industry.

What are some examples of commonly used software-based medical devices or digital therapeutics?
Digital therapeutics are relatively new, so I’m not sure any are “commonly used,” per se. But there are some that are commercially available to consumers. Among those with U.S. Food and Drug Administration (FDA) authorization, Endeavor by Akili for child ADHD, Rejoyn by Click and Otsuka for Major Depressive Disorder, and Sleepio for Insomnia and Daylight for Generalized Anxiety Disorder by Big Health are available by prescription. EndeavorOTC by Akili for adult ADHD is available over the counter. Freespira is a hardware/software combination treatment for PTSD and Panic. JOGO uses biofeedback to treat chronic pain.

How can digital therapeutics be integrated into physical medical devices?
One key way we see is biofeedback – collecting physiological data and then using software to tell the patient what’s happening, and what to do differently. An example is PeriFit – a pelvic floor device for urinary incontinence paired with a game on a smartphone. Pelvic contractions control the game.

Another is to administer treatment using a device and have software serve as a companion to that treatment. For example, Nerivio delivers electrical signals to the body to modulate migraine, and software is used to adjust intensity and record headache episodes.

Another is to provide precision medicine capabilities, i.e. to assess patients and determine what dose, or intervention, they should experience for optimal outcomes. Bullsai by Turing Medical is an example – it provides data to support decision making for neuromodulation therapies to identify patient-specific brain locations for treatment.

What unique regulatory challenges do digital therapeutics present, and why are they held to higher standards than other devices of a similar risk class?
Digital therapeutics are so new that there aren’t very many products on the market to serve as potential predicates. That means blazing new trails and forging new paths with the FDA and other regulatory bodies. Being first isn’t easy! Unlike a blood pressure cuff, which you can get cleared with nothing but bench testing to show it’s similar to another cleared blood pressure cuff, DTx have to do intensive clinical studies to demonstrate safety and effectiveness, even if the predicate provided similar data. I envision a future where digital therapeutics are so abundant, and their safety so thoroughly validated across a wide range of conditions, that creating a new one would require nothing more than bench testing data. We’re a long way away from that.

What types of software-based products require FDA regulation, and which do not?

© AdobeStock | 154814516-gguy

The FDA’s Policy for Device Software Function and Mobile Medical Applications (DSF-MMA) describes and provides examples of three types of health software. The first is general wellness (about which an entire guidance, called General Wellness: Policy for Low Risk Devices, has been published), which requires no regulation at all. These are products intended to 1) help encourage healthy activity and general health without making reference to any disease or condition, or 2) reduce the risk or impact of chronic diseases and conditions where healthy lifestyle is understood to play a role, by promoting healthy lifestyle.

The second type of software is regulated as a medical device, and subject to all the same general controls (including compliance requirements) as any medical device.

The third type is somewhere in the middle – it might qualify as a medical device, it might not, because it does something health-related, but it does it in a way that is so low risk that the FDA has chosen not to focus its oversight on it. We call these “enforcement discretion” products, because the FDA is exercising its discretion not to enforce its policies in these cases. An important distinction is it’s not that the product isn’t a device – in many other regions (say, the EU) the same product might be regulated as a Class I or Class II device under EU MDR.

Enforcement discretion products have been determined to be so low risk that the FDA has deprioritized them from a regulatory standpoint. This is a challenge to regulatory teams, because the products are not evaluated by the FDA, so there’s no clear feedback on whether they are compliant or not. If you see a product on the market, without FDA authorization, making health-related claims, it’s hard to judge whether it’s being legally marketed under enforcement discretion, or if it’s overstepping and at risk of FDA enforcement action. Commercializing in this ambiguous category requires a lot of careful planning and not a small amount of guesswork.

How can a developer of digital therapeutic applications determine whether to seek regulatory approval?
The FDA digital health policy navigator is a great first stop, but loosely speaking, if your software is claiming to treat or diagnose a disease, it probably needs market authorization. If technology supports a clinician’s decision making without making the decision itself – or alleviates symptoms in a patient without directly treating the disease – it may be considered an enforcement discretion device or even classified as non-device software. If it’s improving stress, quality of life, or any other non-clinical outcomes, it’s probably wellness.

What does the regulatory process look like for a software-based medical device, and how does it differ from other medical devices?
With some devices, it’s enough to look at what other similar devices have done and mimic that. In some cases, no clinical data is needed – just bench testing to show that the products are the same from a physical standpoint.

With software-based devices, this can be tougher to do. Every algorithm is different, even if its intended use is similar, so substantial equivalence often involves requirements around clinical data. Software-based devices are often novel, trying to do something with no clear predicate, and the FDA has to decide whether to let them use the 510(k) pathway with pretty loose connection between product and predicate (as we have seen a lot in DTx (i.e., Rejoyn, a depression product that modifies attention as its primary mechanism, using reSET, a behavior change product for addiction as a predicate) or to push products through the De Novo pathway, creating a new regulation.

The only way to be sure what the FDA will decide is to meet with them. Software-based products are well-advised to have presubmission meetings to address questions of pathway and clinical evidence requirements.

Are there any standards, either in place or in development, for digital therapeutics?
Yes! ISO 11147:2023 provides characteristics of a DTx, including product use cases, and discusses the various standards that apply to DTx products.

DTx have to follow ISO 13485 like any other device.

There are also a lot of useful guidance resources on the FDA website:

https://www.fda.gov/medical-devices/software-medical-device-samd/artificial-intelligence-software-medical-device

https://www.fda.gov/medical-devices/digital-health-center-excellence/regulatory-accelerator

What advice do you have for medical device developers who are working with both the physical and digital aspects of a product? How can they successfully navigate the different regulatory requirements for each of those aspects?
Don’t be afraid to talk to the FDA. They want to help! But sometimes, when dealing with new technology, not even the FDA knows what they want until you bring a product to them and ask them. The only way to do that is to do some kind of submission. Presubmissions are a good way to get input before you make a big commitment to run a study or propose substantial equivalence.  Submit early and often (but not too often!!).