Digital twins: A new tool to better understand and manage type 1 diabetes

Managing type 1 diabetes (T1D) can feel like you’re juggling many balls at the same time, except instead of balls, you’re juggling with blood sugar monitoring, insulin dose and carbs calculations, the effects of physical activity and stress … constantly. And, hypoglycemia and hyperglycemia can still happen unexpectedly, even with help from new technologies (e.g., continuous glucose monitors [CGM], artificial pancreases) and the utmost caution.

Now, imagine a tool that could help you anticipate blood sugar variations and test different insulin doses and meals, and show you how your body may react. 

This is exactly what the digital twin is all about. It’s a virtual version of you that learns how your body reacts to different things and can simulate situations to help you make decisions and, as much as possible, improve your diabetes management. 

A digital twin … what is that?

A digital twin is a virtual version of a person based on their health information. For a person with T1D, the digital twin could use the following data. 

• CGM data
• Insulin doses;
• Meals;
• Physical activity;
• Data from a paired watch (e.g., heart rate, sleep schedule).

The digital twin is created by a software program based on this data, accumulated over time. Then, the digital twin can simulate different situations, for instance, insulin doses adjustments, pre- and post-meal boluses, a meal high in carbs, or exercising. It should be noted that this tool is currently used in research settings only. Their reliability and safety for everyday use have yet to be established.

How can this technology help T1D management?

Nowadays, even with the most advanced tools, diabetes management is often reactive, i.e., a situation is dealt with after it happens (e.g., hyperglycemia, hypoglycemia, significant blood sugar variations), rather than by anticipation.

A scientific journal recently suggested that technology such as digital twins could make diabetes management much more proactive and help anticipate and avoid problems.

Researchers are looking at various potential features.

1. Predict blood sugar: While CGMs display current blood sugar and short-term trend arrows, the digital twin could go even further, taking into account more factors (e.g., meals, insulin doses, physical activity, sleep, menstrual cycles) for more accurate predictions and earlier anticipation. This would allow for improved blood sugar predictions and earlier insulin adjustments.
2. Determine the insulin dose: Each person with T1D reacts differently to insulin, carbs and physical activity. With a digital twin, different insulin doses could be simulated in order to find out what works best for a person in a given context. The model could also learn how a person’s blood sugar behaves in different situations based on data accumulated over time (e.g., blood sugar, meals, physical activity), which would allow for increasingly more accurate predictions and better tailored adjustments. The development of this feature depends in part on the evolution of artificial intelligence, especially for the automatic calculation of carbs in a given meal.
3. Test different meals: Many people with T1D often wonder what impact a given meal will have on their blood sugar. There are many uncontrollable factors, besides the number of carbs, that also affect blood sugar. With simulated meals, carb counts and timings, a digital twin could help understand and anticipate the body’s reactions.
4. Improve automated insulin delivery systems: While existing closed loop hybrid systems (also called artificial pancreas) already adjust insulin using algorithms, digital twins could help train algorithms so they are better tailored to the user and more accurate.  

How advanced is research? What are the challenges?

Research evolves fast, but there is still plenty left to do before digital twins are available to people with T1D.  

A Swiss study published in 2025 featured a digital twin that could simulate real-time blood sugar taking meals, insulin doses and physiological factors specific to its user. The digital twin proved to be very accurate. The average gap between actual blood sugar levels and blood sugar levels determined by the digital twin was only 0.3 mmol/L (5 mg/dL), and the digital twin could indicate when its predictions were reliable, simulate new situations (e.g., a different meal, next-day blood sugar prediction) and make estimations very quickly (in a few seconds).

However, a scientific literature review published in 2026 highlights several hurdles to be cleared before this technology can truly improve diabetes management.

• User customization: Every body reacts differently to insulin, meals and physical activity. To make accurate predictions, the digital twin must be customized for every user. Relevant information must also be customized to avoid adding to the mental load with too much information.
• Privacy and data security: These tools use personal information (e.g., blood sugar, insulin doses, meals, physical activity, sleep). It is crucial that this information remain secure and confidential.
• Complexity of the human body: Blood sugar can be influenced by more than 40 factors; even the best digital twin can’t predict everything accurately.
• Data quality and integration: In order to work as intended, the digital twin uses information from various devices, involving different models and brands (e.g., CGM sensors, insulin pumps, paired watches, applications). This is an added technical difficulty that can affect accuracy. The different data sources also have to reliable (e.g, Was the physical activity accurately measured?)
• Trust and clinical integration: People with T1D and health professionals must understand and trust the digital twin’s predictions before it can be used in their practice.
• Equity and accessibility: The tools must work for everyone, regardless of technology access, income, or ethnicity. Solutions must include everyone and serve everyone equally. For instance, in Canada, intelligent insulin pens that automatically log insulin doses and injection timings are still unavailable.
• Cost and infrastructure: The digital twins are expensive to develop, maintain and host. Access to this technology will depend on insurance coverage and the availability of reliable Internet connections and servers.

Even though this technology is still under development, recent studies show that it in time, it could eventually complement the current tools and help achieve better tailored and more proactive diabetes management. From an ethical standpoint, it will be essential that each person be able to freely choose whether they want to use a digital twin.

Would you like to contribute to research on type 1 diabetes in Quebec?

If you or your child live with T1D (or LADA), you can help advance research on type 1 diabetes in Canada by joining the BETTER registry today.

The lived experience you share helps researchers better understand daily life with diabetes and develop new strategies to improve diabetes care and management.

Shaae your experience now!

References

• Cappon, Giacomo, and Andrea Facchinetti. “Digital Twins in Type 1 Diabetes: A Systematic Review.” Journal of diabetes science and technology vol. 19,6 (2025): 1641-1649. doi:10.1177/19322968241262112
• Hoang, TD. et al. (2026). A Real-Time Digital Twin for Type 1 Diabetes Using Simulation-Based Inference. In: Li, L., et al. Digital Twin for Healthcare. DT4H 2025. Lecture Notes in Computer Science, vol 16193. Springer, Cham. https://doi.org/10.1007/978-3-032-07694-6_4
• Olawade, David B et al. “Digital twin paradigm in diabetes prediction and management.” Diabetes research and clinical practice231 (2026): 113075. doi:10.1016/j.diabres.2025.113075