Prognostic accuracy of continuous glucose monitoring in the prediction of diabetes mellitus in children with incidental hyperglycemia: receiver operating characteristic analysis.

ABSTRACT

AIM: The purpose of this study is to evaluate continuous glucose monitoring (CGM) in predicting diabetes mellitus in children with incident hyperglycemia (IH) and negativity for some insular autoantibodies. SUBJECTS AND METHODS: Thirty-one autoantibody-negative children who presented at our center with IH underwent a baseline assessment and were followed up for 23.8 months (range, 6-48 months). At the end of the follow-up, we compared the receiver operating characteristic (ROC) areas under the curve (AUCs) of metabolic markers from 17 children who developed diabetes (Group A; n=17) and 14 children who did not develop diabetes (Group B; n=14). RESULTS: Only two oral glucose tolerance test (OGTT)-derived markers and three CGM-derived markers showed a good prognostic performance, with ROC AUCs indicating significant results (P<0.0001) for the following markers 2-h glucose, OGTT (0.813; 95% confidence interval [CI] 0.621-0.954); AUC glucose, OGTT (0.832; 95% CI 0.611-0.950); CGM glucose measurement peak (0.803; 95% CI 0.621-0.923); percentage of CGM glucose measurements inside the range 70-125 mg/dL (0.866; 95% CI 0.695-0.961); and percentage of CGM measurements ≥126 mg/dL (0.889; 95% CI 0.724-0.973). The combination of the OGTT-derived markers did not increase the predictive value, but the combination of CGM markers with each other or with the OGTT markers yielded higher ROC AUCs (ranging from 0.828 to 0.945). CONCLUSIONS: This is the first study showing that CGM is useful in predicting diabetes mellitus in children with IH.