Reliability of continuous glucose monitoring system in the inpatient setting

AIMS/HYPOTHESIS: Hyperglycemia and hypoglycemia are associated with increased morbidity and mortality in the inpatient setting. Standard point of care capillary glucose testing (POCT) is commonly used in hospitalized patients to monitor their glucose levels. The goal of this study was to examine the relationships between the glucose readings obtained by a continuous glucose monitoring system (CGMS) (Freestyle Libre) and the capillary blood glucose results obtained by the inpatient glucose POCT meter (Accuchek Inform II) as well as between CGMS readings and the serum glucose values obtained by the hospital laboratory. Study participants had either primary or secondary diagnosis of diabetes mellitus and were admitted to non-critical units. We hypothesized that there exists an acceptable agreement between the capillary blood glucose results obtained by the inpatient glucose POCT meter (Accuchek Inform II) and the readings obtained by the CGMS (Freestyle Libre); and that there exists an acceptable agreement between the serum glucose levels and the glucose values obtained by the CGMS. METHODS: This was an Institutional Review Board approved prospective cohort study for the non– critical inpatient setting. Fifty-two hospitalized patients with diabetes were recruited. After informed consent was obtained, patients were instructed on the application and use of the CGMS. The data were assessed using a standard regression analysis and modified Bland Altman analysis. All analyses were conducted using SAS, release 3.8 Enterprise Edition (SAS Institute Inc., Cary, NC). RESULTS: Fifty-two subjects recruited into the study represented a sample of convenience. There were a total of 467 AccuChek-Libre pairs, The regression analysis showed a negative bias between. Libre and AccuChek, R(2) = 0.83, with Libre glucose readings on average being lower than those of AccuChek. Using Bland-Altman analysis, 42% of the 467 Libre-AccuChek pairs had a difference in glucose reading more than 15%. Mean absolute relative difference (MARD) between Libre and AccuChek was 15.6%; mean relative difference (MRD) between Libre and AccuChek was −11.4%. The regression analysis showed a negative bias between Libre and serum glucose, R(2) = 0.89. Using Bland Altman analysis, 36% of the 44 Libre-serum pairs had a difference in glucose reading more than 15%. Mean absolute relative difference (MARD) between Libre and serum glucose was 13.2%; mean relative difference (MRD) between Libre and serum glucose was −12.5%. A review of the data pairs showed that 71/467 Accuchek-Libre pairs had one result that was either below 70 mg/dl or above 200 mg/dl (combined American Diabetes Association-ADA-, American College of Physicians-ACP- and American College of Endocrinology-AACE- goals). Thus 85%, of these pairs would have yielded results that engendered the same intervention (e.g. treatment for hypoglycemia or hyperglycemia). Likewise 5/45 Serum-Libre pairs had one result that was either below 70 mg/dl or above 200 mg/dl; thus 89% of these pairs would have yielded results requiring the same intervention. CONCLUSION/INTERPRETATION: These findings confirm the existent literature and indicate acceptable agreement between the standard POCT and the CGMS as well as between serum glucose and the CGMS values. Because of the advantages of the CGMS over capillary blood glucose testing (reduced patient discomfort and reduced staff exposure to patients in isolation) CGMS use may be preferable to the current bedside capillary blood glucose testing in hospitalized patients with diabetes mellitus. As with other laboratory measures, clinical judgement needs to be exercised when the laboratory values are used to guide patient care.