Threshold and Predictive Alerts of Continuous Glucose Monitoring and Glycemic Control in Hospitalized Adults With Diabetes

Importance: Interest in use of continuous glucose monitoring (CGM) in the hospital is increasing, yet clear guidance on the optimal implementation of CGM glucose alerts is lacking. It also remains unclear if alerts themselves or the fact that sensor values are available in real time are the main reason for improved glucose outcomes observed with inpatient CGM use. Objective: To determine the incremental value of CGM threshold and predictive alerts for glucose management among hospitalized adults with diabetes. Design, Setting, and Participants: This single-center, open-label, randomized clinical trial was conducted between April 15 and December 9, 2024, at a tertiary hospital. Eligible participants were adults with type 1 or type 2 diabetes admitted to endocrinology wards. Interventions: Participants were randomly assigned (1:1:1) to receive 1 of the 3 CGM alert strategies: all alerts off (n = 178), threshold alerts only (n = 177), or threshold plus predictive alerts (n = 178). Main Outcomes and Measures: The primary outcome was the percentage of time within the target glucose range of 70 to 180 mg/dL (time in range [TIR]). The primary analysis was conducted on a modified intention-to-treat basis, including all randomized participants except those who were subsequently confirmed not to meet the eligibility criteria. Results: Among the 533 randomized participants, 331 (62.1%) were male, the mean (SD) age was 62 (12) years, and the mean (SD) hemoglobin A1c level was 9.2% (2.1%). The mean (SD) TIR during hospitalization was significantly higher among the group with threshold alerts on than the group with all alerts off (75.1% [16.6%] vs 71.2% [18.1%]; adjusted mean difference, 3.9% [95% CI, 0.1%-7.8%]; Holm-adjusted P = .04). Moreover, time above range (>180 mg/dL) was lower in the group with threshold alerts on than in the group with all alerts off (-3.9% [95% CI, -7.1% to -0.7%]), whereas time below range (<70 mg/dL) did not differ between groups (-0.04% [95% CI, -0.3% to 0.2%]). No differences in primary and secondary outcomes were observed between the group with threshold plus predictive alerts and either of the other 2 groups. Conclusions and Relevance: In this randomized clinical trial of hospitalized adults with diabetes, CGM threshold glucose alerts modestly improved glycemic control compared with CGM use without alerts. The addition of predictive alerts did not confer further glycemic benefit beyond threshold alerts alone. Future development of integrated clinical decision support systems or automated insulin delivery technologies may further enhance the effectiveness of CGM among hospitalized patients. Trial Registration: ClinicalTrials.gov Identifier: NCT05941286.