Digital Therapeutics and Type 2 Diabetes: A Review

Type 2 diabetes (T2D) is a serious disease affecting hundreds of millions of people around the world, causing serious microvascular and cardiovascular complications, including amputation, blindness and death.¹ A key tenet of T2D treatment is maintaining good glycaemic control, as measured via HbA_1c. Standard care for diabetes includes both lifestyle changes (increasing exercise and improving both diet and sleep) and – when necessary – pharmacological treatment.²

Mobile health interventions such as mobile phone apps that support self-management, have proven to be effective in increasing physical activity of T2D patients and improving glycaemic control.^3–6 There are a wide range of apps aimed at T2D patients, but most do not meet the standards of digital therapeutics (DTx). Definitions for DTx vary, but as applied to T2D, in this paper we use the following definition: a DTx is a patient-facing therapeutic to prevent, treat or manage T2D, with proven clinical effectiveness, implemented in software.^7,8 DTx is often viewed as a subset of Software as a Medical Device.

There are some subtleties here that warrant consideration (Table 1). One key concept is ‘patient-facing therapeutic’. This requirement excludes a wide range of software that helps healthcare professionals directly, as such software is not patient-facing. It can also be viewed to exclude therapeutics where the main effect is the result of something other than the digital intervention among patients. For example, if an app helps a patient to communicate with their physician, this is likely not DTx, as any improvement is primarily because of the physician rather than the method of communication. Perhaps more controversially, apps that help patients take medication optimally can have a great impact but are – in our view – not DTx because the benefit comes from the drug. These apps generally simply help patients take medications according to the instructions of physicians. Similarly, an app that merely communicates a new source of information, such as displaying the results of a continuous glucose monitor (CGM), may be very helpful, but that alone is not a full therapeutic and is not DTx, although a broader therapeutic incorporating such a device could be DTx.

A second area worth consideration is what it means to have ‘proven clinical effectiveness’. This is the key difference between ‘wellness’ apps and medical solutions, as the numerous wellness apps in the marketplace lack proof of clinical effectiveness. More broadly, the proven clinical effectiveness test excludes many T2D-oriented apps – including some whose creators view them as being a medical solution – from being labelled as DTx. The simplest criterion is to require regulatory approval for the prescription of the intervention to treat T2D; some practitioners view DTx as requiring this. There are some interventions that are approved but do not require a prescription, and there are other interventions that do not have approval but do have strong proof of clinical effectiveness. Perhaps the definition of DTx can be expanded to cover those cases as well, although some may feel that some form of regulatory approval is required.

Finally, there is the requirement that this be software. This is correct, but it somewhat misses the point. Yes, these interventions are implemented in software, but that is rather like emphasising that a treatment is via an injection instead of focusing on the pharmacological impact. The key idea here is not the software but the treatment that the software enables us to implement. DTx opens the possibility of meeting the patient where they live, largely free of constraints of space or time. Instead of infrequent scheduled visits to a clinic, treatment can be all day and every day, wherever the patient is located throughout the day. If we could achieve the same thing without software – via a complex analogue device or perhaps to some future technology – the benefit to the patient would be the same, but we are limited to just software. This removal of the constraints of time and space is the true power of DTx and is what makes it such a promising approach.

Digital Therapeutics and Diabetes: Currently Available Options

We conducted an exploratory search among English-language journal articles and supplier websites, with a cutoff date of July 2024. Given that this is a broad review paper, we did not use the rigorous search methodology used in formal reviews or meta-analyses. We examined a wide range of digital interventions that relate to T2D, including those focused on T2D directly, diabetes of all types (including T2D) and general health with applicability to T2D. We do not claim to have made a definitive list. We assessed these interventions using three frameworks: first, an initial filter eliminating candidates that are clearly not DTx because of the mechanism or a lack of proof of effectiveness; second, an exposition of each remaining candidate, capturing the proof of clinical effectiveness; and finally, a mapping of the remaining candidates in terms of the main therapeutic mechanism and the main behaviour change mechanism.

We examined 64 candidates. Applying our first framework, most do not meet even a loose definition of DTx (Table 2).^9–20 For many apps, we were not able to find proof of clinical effectiveness. This includes several apps using CGM. Some rely on healthcare professionals or human coaches for their effectiveness; these may well be worthy of use, but they are not DTx.

Of the 64 digital interventions related to T2D, we identified 12 that are either clearly DTx, arguably DTx, or close enough to be worth some discussion. These 12, examined using our second framework, vary widely in approach (Table 3). Some are approved and seem clearly to be DTx; there is variability by country and agency in the level of rigour required for approval, but all require proof of clinical effectiveness for solutions such as these. Others are not approved and may be viewed by some as not fully DTx, and the level of proof of effectiveness varies. We were unable to find any approved T2D DTx in east Asia.

We applied our third framework to analyse the 12 candidate T2D DTx in terms of therapeutic mechanism (what directly causes the medical impact) and behaviour change mechanism (how the therapeutic behaviour is induced) (Figure 1). We used publicly available information to infer the main mechanisms and noted secondary mechanisms used. This is not intended to be a definitive view and some may argue with some of these details; instead it is intended to apply a framework to understand this confusing space.

In terms of therapeutic mechanisms, we found three broad categories. At one extreme are two apps (mySugr and DarioHealth) focused on providing information. At the other extreme are five apps (KYT Adhere, Insulia, d-Nav, iSage Rx and My Dose Coach) focused on helping patients take their medication. Between these are five apps (Aspyre, DialBetesPlus, Klinio, BlueStar and StepAdd) that focus on inducing specific lifestyle behaviour changes.

Information-focused Digital Therapeutics

Two programmes (mySugr and DarioHealth) focus on providing information to the patient, with the concept being that the patient will use this information to make better lifestyle choices to manage their health.^9,12 They focus on providing information rather than specific direction about what behaviour to change. Both have regulatory approval with proven benefits – albeit not via the gold standard of a randomised control trial – and meet the strictest definition of DTx. Dario Health, for example, focuses on making blood glucose measurements convenient. Common information includes outcome measures (weight and blood pressure) as well as direct in-process measures (blood glucose). Both Dario Health and mySugr use direct measurement of blood glucose via a finger stick and have added measurements via CGM.

Medication-focused Digital Therapeutics

Five apps focused on helping patients take their medication, with four (Insulia, d-Nav, iSage and My Dose Coach) focused on the timing or dosage of insulin and one (KYT-Adhere) focused on more general medication use. We argue that these five are not, in fact, DTx, despite four of the five having regulatory approval and large proven benefits, as their therapeutic effectiveness is primarily because of the pharmacological treatment rather than the app itself. Some are included in other lists of DTx (such as Insulia in the Digital Therapeutics Alliance list), and they are certainly worthy of discussion.²¹

Behaviour Intervention Digital Therapeutics

Five apps (Aspyre, DialBetesPlus, Klinio, BlueStar and StepAdd) focus on inducing specific lifestyle behaviour changes. Just two of the five (Aspyre, BlueStar) have regulatory approval and meet the strictest definition of DTx. Aspyre focuses on managing stress.¹ The others focus on exercise and diet behaviour. BlueStar is a key programme in this space, and it was the first DTx solution approved for T2D.^10,22 BlueStar intervenes primarily on both diet and exercise. DialBetesPlus also addresses diet and exercise (specifically walking).¹⁸ Although DialBetesPlus has a proven effect, it has not been submitted for regulatory approval, and we would argue that it is DTx-adjacent rather than DTx. Klinio is focused on diet and body weight and does not have regulatory approval. StepAdd focuses on increasing step counts.^17,19 Like DialBetesPlus, StepAdd is not approved and thus is not yet true DTx, but a large trial is underway that may lead to regulatory approval. We found no DTx focused on sleep, with only one (BlueStar) addressing sleep in any way.²³

Behaviour Change

The programmes differ in how they aim to induce behaviour change. Many provide information or direct specific actions, but without a strong base on a theory of behaviour change. There is evidence that behaviour change interventions have better results when they are based on a theory of behaviour change.²⁴ DialBetesPlus is described using the theory of planned behaviour (TPB), although it was not built based on it.²⁵ We found just two DTx candidates that seem to be built on a theory of behaviour change. BlueStar uses habit theory, as it aims to help patients form new healthy habits. StepAdd applies social cognitive theory (SCT).^26,27 Because DialBetesPlus, BlueStar and StepAdd use different models for the same underlying thing – human behaviour – there is considerable overlap in their approaches along with differences in scope and specificity. SCT focuses primarily on self-efficacy, the belief that one can perform the desired behaviour (e.g. walking more). TPB uses the same or a very similar concept – perceived behavioural control – but in a larger theory that aims to explain the key drivers of behaviour. Habit theory focuses on a key technique – forming habitual behaviour to reduce or eliminate the cognitive burden of making a decision – rather than striving to explain all of human behaviour.

Gaps and Future Directions

There are relatively few DTx or DTx-adjacent treatments for T2D. Our framework shows some interesting concentrations in similar solutions for data management and managing insulin use and more diversity among lifestyle change solutions. There are gaps.

We have found that patients dislike measuring blood glucose via a finger stick, leading to low rates of usage.¹⁸ Some DTx have added measurement via CGM. The use of CGM is growing, and it seems likely that many future DTx will be built around the detailed information that CGM provides. Of course, some patients dislike CGM, and rates of sustained usage are relatively low.²⁸ More generally, combining the direct feedback of blood glucose, perhaps via CGM, with a broader, theory-based intervention on diet would seem likely to be effective.

Although exercise is of proven benefit for T2D patients, relatively few programmes addressed it. Furthermore, they generally focus on walking. Walking is certainly beneficial and easy to measure via pedometer, but other modes of exercise and factors such as intensity are also important and there seems to be a lack of programmes fully capturing the totality of the human physical activity experience.⁶

Diet modification is a key part of diabetes treatment, yet diet behaviour is quite complex. Non-DTx programmes have had success through use of human coaches. There are numerous ‘diet apps’ aimed at the general population, but evidence of effective usage in this T2D population is slight. Calorie counting is difficult to implement, as errors or omissions in data logging can have major implications for the assessment of the patient’s energy balance. There is a clear need for better T2D-focused DTx solutions that help patients make better dietary choices.

Sleep is addressed by only one T2D DTx, BlueStar, and sleep is not BlueStar’s focus. There are numerous wellness apps assessing sleep quality and quantity, but there is a lack of evidence that improving sleep in T2D patients improves glycaemic control to a level that would meet the DTx requirements. Specific recommendations for sleep for T2D patients are relatively new, and we expect interventions based on sleep to be an active area of research that may, eventually, lead to a sleep-focused T2D DTx programme.²

As with all DTx, T2D DTx solutions must overcome issues of usability in the target population. T2D patients are generally older than the general population. Patients may lack digital literacy, basic digital access or even the cognitive ability to engage via a DTx mechanism. More generally, even patients with high digital literacy may be discouraged by a burdensome interface and patient engagement is an on-going challenge.

A possible solution is to use generative artificial intelligence (AI) to make a more streamlined interface between the software and the human. Numerous apps claim to use AI, but one suspects that many of these claims are mostly marketing. However, current generations of language-model AI seem finally to be suitable for use in a flexible interface for a vulnerable patient population. We expect to see AI used to greatly improve patient engagement. More generally, there is a need for improvements to DTx to make them more broadly accessible.

We were surprised that nearly all T2D DTx were not built on a theory of behaviour change. A decade ago, digital interventions were perhaps not viewed as serious by much of the research community, and any intervention that achieved a positive result was significant, even if the mechanisms underlying the change were unclear. We believe that the situation is now fundamentally different, as numerous interventions have shown their benefit.

As scientists, we need to understand how an intervention functions, not just that it does function, so that we can increase the focus on effective parts, trim out ineffective or even deleterious elements, and discover new avenues for interventions. Applied to interventions that seek to change patient behaviour, this means that we need to understand how specific elements are changing human behaviour. It seems difficult to do this without any underlying theory of behaviour, and it seems critical to measure specific elements and their associated mechanisms. This requires measuring psychological elements, such as the constructs in the TPB theory of human behaviour. There is a clear need for T2D DTx-related interventions to explore exactly how they are inducing changes in behaviour.

There are no approved T2D DTx in east Asia. Although Asia is lagging the US and Europe in this area, we expect T2D DTx to grow in Asia. Most parts of Asia face ageing populations. By using DTx, we can focus our human medical resources on medical procedures that can only be performed by humans. DTx offers a means of effectively using limited medical resources in ageing societies.

Conclusion

There are effective T2D DTx options but they are relatively scarce. We found just four that met the strictest definition of DTx. DTx offers the capability to meet the patient where they are throughout the day. For a disease such as T2D that has a major lifestyle component, the ubiquity of DTx holds great promise, which is currently only partially fulfilled. Human behaviour is complex and getting patients to change their lifestyles to exercise more, eat better or sleep more is difficult. Future efforts will likely focus on the underlying aspects of human behaviour rather than solely focusing on displays and software, likely using AI to improve patient engagement. Finally, there is great need for T2D DTx solutions tailored to Asian populations and regulatory environments.