- βAI-guided bedtime snack recommendations cut severe nighttime lows (under 54 mg/dL) by about 30%
- βThe app worked without disrupting overall blood sugar control or causing overnight highs
- βThe approach is practical for everyday life, using CGM data and activity logs patients already track
| Journal | Diabetes care |
| Year | 2025 |
| Authors | Mosquera-Lopez, Roquemen-Echeverri, Jacobs |
| PMID | 40705054 |
What Was Studied and Why It Matters
Waking up with dangerously low blood sugar β or not waking up at all β is one of the most frightening risks for people living with Type 1 diabetes. Nighttime hypoglycemia is especially common in adults who use multiple daily insulin injections (MDIs) and those who are physically active. Researchers wanted to know: could a smartphone app that uses your continuous glucose monitor (CGM) data and exercise history predict overnight lows and suggest a personalized bedtime snack to help prevent them?
How the Study Worked
Researchers developed an app called DailyDose Smart Snack (DDSS). The app connects to a CGM, lets users log their physical activity, and then uses an advanced AI model β called an evidential neural network β to calculate the likelihood that blood sugar will drop dangerously low during the night. Based on that prediction, the app suggests whether and how much to eat before bed.
Twenty adults with Type 1 diabetes took part in a randomized crossover trial, meaning everyone tried both approaches: four weeks using the DDSS app, and four weeks using their CGM without the app's snack guidance. Participants went about their normal daily lives throughout the study.
What the Researchers Found
The main goal was to reduce nights where blood sugar dropped below 70 mg/dL for 10 minutes or more. The app did not show a statistically significant improvement for this threshold. However, a closer look at more severe lows β below 54 mg/dL β told a more encouraging story:
- Nights with severe lows (under 54 mg/dL) dropped by about 30% when participants used the app
- The proportion of nights with these dangerous lows fell from roughly 12% down to 8%
- Importantly, reducing these lows did not cause blood sugar to run too high overnight β overall glucose control stayed the same
What This Means for Real Patients
For people with Type 1 diabetes managing their condition with insulin injections and a CGM, this research suggests that a smart bedtime snack β guided by your own glucose trends and activity data β could meaningfully reduce the most dangerous overnight lows. It won't replace your diabetes care team's advice, but it adds a practical layer of personalized guidance right when you need it most.
Reliable CGM supplies are the foundation of tools like this. If you use a CGM as part of your daily management, keeping a consistent supply on hand is essential. MDS Diabetes makes it easy to stay stocked with the diabetes supplies you depend on, so technology like this can work as intended.
Bottom Line
An AI-powered app that recommends personalized bedtime snacks based on CGM data and physical activity significantly reduced the most severe nighttime blood sugar lows in adults with Type 1 diabetes β without negatively affecting overall glucose control. While larger studies are needed, this is a promising step toward smarter, safer nights for people managing Type 1 diabetes with insulin injections.