- βDigital twin models can personalize mealtime insulin doses for children based on their changing metabolism
- βThe system showed steady improvement in time-in-range and reduced hyperglycemia over 24 weeks
- βCould reduce the burden of frequent manual dose adjustments for families and clinicians
| Journal | Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference |
| Year | 2025 |
| Authors | Pellizzari, Cappon, Sparacino |
| PMID | 41337180 |
What Was Studied and Why It Matters
Managing Type 1 diabetes (T1D) in children is especially challenging. Kids' bodies are constantly growing and changing, which means their insulin needs shift frequently β sometimes from week to week. Getting the mealtime insulin dose just right is critical: too little leads to high blood sugar (hyperglycemia), and too much can cause dangerous lows.
Researchers from a 2025 IEEE engineering and medicine conference explored whether a technology called a digital twin could help solve this problem by automatically personalizing insulin doses for pediatric patients over time.
What Is a Digital Twin?
A digital twin is a computer model built to act like a specific patient's body. Think of it as a virtual copy of a real child's metabolism. By feeding real glucose data into this model, researchers can simulate how that child responds to insulin β without any risk to the actual patient. The digital twin learns and updates itself as new data comes in, capturing how the child's needs change over time.
What the Researchers Did
The team tested their system using data from five real pediatric T1D patients, drawn from the Tidepool Big Data Donation project. The data included 724 days' worth of continuous glucose monitor (CGM) readings collected over 24 weeks.
Every two weeks, the algorithm used each child's digital twin to recalculate and fine-tune their mealtime insulin bolus β the dose given before eating. The goal was to minimize something called the Glycemia Risk Index (GRI), a score that reflects overall blood sugar danger from both highs and lows.
What They Found
The results were encouraging. Over the 24-week study period, the system showed:
- More time in the healthy blood sugar range β improving by about 0.63% every two weeks
- Less time with high blood sugar β decreasing by about 0.18% every two weeks
- A steadily falling risk score β the GRI dropped by roughly 0.64 points every two weeks
While these may seem like small numbers, consistent improvement over months adds up to meaningful real-world benefits β fewer dangerous highs, less risk of long-term complications, and better daily quality of life.
What This Means for Real Patients and Families
Right now, adjusting a child's insulin doses typically requires clinic visits, manual review of glucose logs, and a lot of guesswork. This technology could one day assist doctors in making faster, more precise, data-driven adjustments between appointments.
Accurate CGM data is the fuel that makes this kind of system work. Families who rely on consistent CGM supplies β available through trusted sources like MDS Diabetes β are already generating the kind of continuous data that makes personalized tools like digital twins possible.
It's important to note this was an in silico study, meaning it was tested entirely in computer simulations, not yet in real patients. Clinical trials will be needed before this becomes a real treatment tool.
Bottom Line
Digital twin technology shows real promise for helping children with Type 1 diabetes get more precisely tuned insulin doses without constant clinic visits. By learning from continuous glucose data every two weeks, the system gradually improved blood sugar control over a 6-month simulation. While still in early research stages, this approach could one day make personalized diabetes management smarter, safer, and less burdensome for kids and their families.