Evaluation of a Prediction-Based Bedtime Intervention in Reducing Nocturnal Low Glucose in Adults with Type 1 Diabetes: The DailyDose Bedtime Smart Snack Cross-over Study

OBJECTIVE Nocturnal hypoglycemia is challenging for people with type 1 diabetes, particularly those physically active or using multiple daily injections (MDI). We hypothesized that adding bedtime decision-support to usual care with continuous glucose monitoring (CGM)-augmented MDI therapy could reduce nocturnal hypoglycemia. RESEARCH DESIGN AND METHODS We developed the DailyDose Smart Snack (DDSS) app, which features CGM and physical activity logging and delivers personalized bedtime snack recommendations to help prevent nocturnal hypoglycemia. Recommendations are based on the probability and timing of low glucose events forecasted by an evidential neural network. We conducted a randomized cross-over trial comparing DDSS to CGM-augmented MDI, for 4 weeks each under free-living conditions, to evaluate the feasibility and effect of DDSS in reducing nocturnal hypoglycemia. RESULTS Twenty participants completed the study (age 39±15 years; 10 Females; HbA1c 7.1±1.0% [54±11 mmol/mol]). There was no difference between arms in the proportion of nights with low glucose events <70 mg/dL lasting 10+ minutes (control: 26.0±14.1% vs. DDSS: 23.6±15.6%; OR=0.83, P=0.207) or in overnight CGM metrics. A post-hoc analysis showed that the proportion of nights with low glucose events <54 mg/dL lasting 10+ minutes was significantly reduced during the DDSS arm by 3.5% (11.9±12.9% vs. 8.3±8.8%; OR=0.64, P=0.040). CONCLUSION DDSS did not significantly reduce the proportion of nights with low glucose events <70 mg/dL lasting 10+ minutes but reduced nocturnal low glucose <54 mg/dL without compromising other glycemic metrics.