Reliability of point-of-care testing for glucose measurement in critically ill adults

1. Reliability of point-of-care testing for glucose measurement in critically ill adults

2. Abstract • Objective: To determine the accuracy and clinical impact of three common methods of bedside point-of-care testing for glucose measurements in critically ill patients receiving insulin infusions. • Design: Prospective observational study. • Setting: A 21-bed mixed medical/surgical intensive care unit of a tertiary care teaching hospital. • Patients: Thirty consecutive critically ill patients who were vasopressor-dependent (n = 10), had significant peripheral edema (n = 10), or were admitted following major surgery (n = 10).

3. Measurements: Findings from three different methods of glucose measurement were compared with central laboratory measurements: • (1) glucose meter analysis of capillary blood (fingerstick); • (2) glucose meter analysis of arterial blood • (3) blood gas/chemistry analysis of arterial blood.

4. Results: Clinical agreement with the central laboratory was significantly better with arterialblood analysis (69.9% and 76.5% for glucose meter and blood gas/chemistry analysis, respectively) than with capillary blood analysis (56.8%; p = .039 and .001, respectively). • During hypoglycemia, clinical agreement was only 26.3% with capillary blood analysis and 55.6% and 64.9% for glucose meter and blood gas/chemistry analysis of arterial blood (p = .010 and <.001, respectively). • Glucose meter analysis of both arterial and capillary blood tended to provide higher glucose values, whereas blood gas/chemistry analysis of arterial blood tended to yield lower glucose values.

6. Agreement vs Disagreement • Agreement: the blood glucose measurements under study generated an observed measurement that fell into the same category (i.e., resulting in the same intervention) as that of the hospital’s central laboratory. (fingerstick:4.1~ lab:4.3) • Disagreement: If the two results fell into different categories and would have resulted in different clinical interventions (fingerstick:4.1mmol/L~ lab:5.0mmol/L)

9. No statistically significant differences between measurement methods for vasopressor-dependent and edematous patients were observed, although there was a trend toward arterial blood providing more accurate results. • Agreement was significantly better for arterial blood measurements by blood gas/chemistry analysis (84.2%) than for capillary blood (fingersticks) (53.8%; p = .0045) among postsurgical patients.

10. When the three patient groups were combined ----- • hypoglycemic readings: arterial samples (55.6% agreement for glucose meter analysis; 64.9% agreement for blood gas/chemistry analysis) appeared significantly better than agreement for capillary samples (26.3%) • nonhypoglycemic readings:no significant differences between tests were observed • hypoglycemic and nonhypoglycemic readings were combined: arterial sample agreement (69.9% for glucose meter analysis; 76.5% for blood gas/chemistry analysis) was superior to capillary sample agreement (56.8%) for all patients studied.

11. For all patients, when compared with the reference standard : • capillary sample measurements were 8.8 ± 17.8% higher • glucose meter analysis measurements of arterial blood were 3.6 ± 15.1% higher • blood gas/chemistry analysis measurements were 2.5 ± 6.7% lower

12. Modified error-grid analysis • display the combined results of individual patient groups for each specimen studied. • blood gas/chemistry analysis of arterial blood: Ninety-nine percent of values fell within the target range • glucose meter analysis of arterial and capillary: 88% and 73% of values fell within the target range. • During hypoglycemia: • glucose meter analysis of arterial and capillary blood overestimated glucose values 5% and 9% • blood gas/chemistry analysis of arterial blood did not overestimate.

17. Compared with the suggested standard provided by the NCCLS • (NCCLS) National Committee for Clinical Laboratory • correlation coefficients measuring the associations between the reference method and each of the three test methods were calculated • The NCCLS suggests that a correlation above 0.9751 is indicative of a method equivalent to the laboratory standard. • blood gas/chemistry analysis met this standards overall (r = .9902), whereas glucose meter analysis of both capillary blood (r = .9516) and arterial blood (r = .9531) fell short of this guideline.

18. Compared with the suggested standard provided by the NCCLS • limited to hypoglycemic readings: none of the three methods reached this degree of association (blood gas/chemistry: r = .9251; capillary blood: r = .4942; arterial blood: r = .6153), • normoglycemic ranges: blood gas/chemistry: r = .9506; capillary blood: r = .7837; arterial blood: r = .8576)  none • hyperglycemic range: blood gas/chemistry: r = .9669; capillary blood: r = .8652; arterial blood: r = .8251) none

19. DISCUSSION • Overall, the performance was poor in all critically ill patients, as clinical agreement with the central laboratory was less than 80% for all tested methods of glucose measurement. • When blood sugars were in the hypoglycemic range, clinical agreement was <70% for all tested methods and as low as 26.3% with glucose meter analysis of capillary blood.

20. DISCUSSION • Blood gas/chemistry analysis appears to be the most accurate and reliable method. • The blood gas/chemistry analysis may also be the safer approach, since hypoglycemia appears to be overestimated, • whereas glucose meter analysis of capillary or arterial blood would underestimate hypoglycemia.

21. Thanks for your attention !