Control of nocturnal hypoglycemia by CGM-based AI-enabled nocturnal hypoglycemia prediction: A retrospective analysis of real-world data.

AIMS: To evaluate the effect of the Accu-Chek® SmartGuide Predict app Night Low Predict (NLP) feature use on nocturnal hypoglycemia in people with diabetes in a real-world setting.

METHODS

This retrospective analysis used deidentified, aggregated data from Predict app users. Outcomes included hypoglycemic events of any severity (glucose < 70 mg/dL), level 2 (L2) events (glucose < 54 mg/dL), hyperglycemic events (glucose > 180 mg/dL), time below 70 mg/dL (TB70), time below 54 mg/dL (TB54), and time above 180 mg/dL (TAR). Users without nocturnal hypoglycemic events were excluded. Nights when predictions were received were compared with nights when predictions were not generated.

RESULTS

Data from 249 people (aged 20-84 years) showed NLP use was associated with 20.0 % lower odds of hypoglycemic events of any severity occurring in a given night and 31.2 % lower odds of any TB70 occurring. Data from 159 people showed NLP use was associated with 31.0 % lower odds of L2 events occurring and 39.2 % lower odds of any TB54 occurring. Hyperglycemic events and TAR were not affected by NLP use.

CONCLUSION

NLP use is associated with lower odds of nocturnal hypoglycemia in people with diabetes without raising odds of hyperglycemia, offering valuable self-management support.