Summary: In a small randomised trial in type 1 diabetes, a digital-twin decision-support system with a simulation-assisted bolus calculator increased time in range and reduced hypoglycaemia compared with traditional carbohydrate-counted dosing over four weeks.
PICO Summary
| Element | Detail |
|---|---|
| Population | 28 adults with type 1 diabetes using injections or sensor-augmented pumps; randomised, open-label, parallel-group study, Colombia. |
| Intervention | STUDIA digital-twin decision-support system with a simulation-assisted bolus calculator. |
| Comparison | Traditional carbohydrate counting for prandial insulin dosing. |
| Outcome | After 4 weeks, the simulation-assisted group had a 7% greater time in range (70–180 mg/dL; p<0.001) and a lower hypoglycaemia incidence rate (RR 0.31; p=0.022). The model’s mean absolute percentage error at 60 minutes was 19.2%, rising at longer horizons, with most simulation-versus-sensor discrepancies in no-risk or slight-risk zones. |
Digital-twin bolus calculator in type 1 diabetes
RCT · type 1 diabetes · 4 weeks
Over four weeks, the simulation-assisted calculator raised time in range by about 7 percentage points and cut hypoglycaemia, but the trial is small, short and open-label, so it is proof-of-concept rather than practice-changing.
Expert Commentary
This is a promising proof-of-concept for a pragmatic idea, that most people with type 1 diabetes worldwide remain on injections or sensor-augmented pumps rather than full automated delivery, and that a digital-twin simulator which previews the glucose consequences of a dose could sharpen everyday bolus decisions affordably. The result is attractive on both fronts that matter, more time in range and, importantly, less hypoglycaemia rather than the usual trade-off, suggesting the tool helped users dose more precisely rather than simply more aggressively. I would weigh the findings against clear limitations. The trial is small at 28 participants and only four weeks long, it is open-label so behaviour and engagement can be influenced by knowing one has a novel tool, and the model’s predictive accuracy degraded at longer time horizons, with a roughly 19% error even at one hour, which bounds how far ahead its forecasts should be trusted. Can I use this with my patients? Not yet as a specific product, since this system is investigational, but the direction is encouraging. It supports the broader value of decision-support and simulation tools for injection-based patients, and I would watch for larger, longer trials before recommending any particular calculator in routine care.
References
Builes-Montaño CE, Lema-Perez L, Ramírez-Rincón A, et al. A digital twin-enhanced decision support system improves time-in-range in type 1 diabetes: a randomized clinical trial. Sci Rep. 2025;15(1):39738. doi:10.1038/s41598-025-23165-x