Continuous Glucose Monitors Have Acceptable Accuracy but High Discordance at High and Low Glucose Concentrations in Pediatric Hospitalized Patients.

Abstract

BACKGROUND Continuous glucose monitors (CGMs) assess interstitial glucose concentrations and improve diabetes management in outpatient settings. However, limited studies have assessed CGM performance in hospitalized pediatric patients, especially at high and low glucose thresholds near clinical decision limits. METHODS This retrospective study included 72 hospitalized pediatric patients with dysglycemia who used CGMs during hospitalization. Paired CGM results within 10 min of laboratory (Lab) and point-of-care (POC) glucose results were retrieved. The mean absolute relative difference (MARD) and percentage of glucose values in acceptable Parkes error zones were assessed. Concordance of CGM and POC results and the frequency that CGM results <70 mg/dL and >180 mg/dL were confirmed (-15 min to +30 min) using POC was assessed. RESULTS There were 2228 paired CGM and Lab or POC glucose results with a MARD of 14.8%, and 99.2% of results were in Parkes zones A and B. The MARD was 20.2% and 13.6% in the hypoglycemic and hyperglycemic ranges. The MARD for POC glucose meters was 15.6% and 8.2% for the hypoglycemic and hyperglycemic ranges. The Cohen kappa between CGM and POC was 0.66 (95% CI, 0.63-0.69). CGM results in the hypoglycemic and hyperglycemic ranges were repeated 80.2% and 16.5% of the time, respectively, with POC methods. CONCLUSION The MARD of CGM in hospitalized pediatric patients is clinically acceptable but there is high discordance between CGM and POC. This implies a clinical need to confirm high and low glucose concentrations with Lab or POC methods but confirmatory testing is commonly not performed.