This was a prospective study approved by the Ethics Committee of the Faculty of Medicine, Siriraj Hospital and informed consents were obtained prior to the study. The study was conducted at Siriraj Hospital, which is a tertiary care university hospital. We consecutively recruited high-risk newborn infants aged between 1 and 24 hours from the High-risk nursery and the Intermediate care unit. The patients included small for gestational age, large for gestational age, infant of diabetic mother, and low birth weight infants. Critically ill infants in the NICU were excluded due to possible multiple medication administration. The study was conducted only during office hours, in order to limit the POC glucose testing to that of trained operators, which were two pediatric residents and a technician.

Demographic data, hematocrit level on the same day of blood glucose testing, and time difference between blood sampling and blood test were recorded. Sample size calculation was based on sensitivity and specificity of at least 93% with acceptable error of 7%. According to the incidence of hypoglycemia in high-risk infants at our hospital (30%), we required at least 174 tests.

Sensitivity, specificity, positive and negative predictive values (PPV and NPV) for detection of hypoglycemia were determined. ROC curves of both glucometers were plotted. The pre-test probability (prevalence) of hypoglycemia in high-risk neonates at our hospital was 30%. We calculated PPV and NPV from sensitivity and specificity because the incidence of hypoglycemia in this study was 51%, which was higher than our population.

The study period was between December 1, 2008 and May 31, 2009. There were 166 infants (61% males) recruited with 180 sets of blood specimens. Mean (SD) gestational age and birthweight were 37.1 (2.8) weeks and 2,799.6 (837.3) grams, respectively. Mean (SD) hematocrit on the day of the study was 52.0% (7.2%). Range (minimum, maximum) and median age (P25, P75) at blood drawn were 23 (1, 24) and 1.0 (1.0, 1.0) hour, respectively. Approximately 31% and 91% of the blood specimens had glucose measurement performed at the laboratory within 30 minutes and 60 minutes of blood drawn, respectively. Mean (SD) time difference between blood drawn and laboratory plasma glucose measurement was 44.5 (24.7) minute.

Of 180 samples, 92 (51%) were diagnosed with hypoglycemia. Blood glucose levels measured by different methods are shown in Web Table I. Differences between blood glucose from each POC glucometer and from laboratory plasma glucose were normally distributed, thus 95% of the measurements had differences between mean plus and minus 1.96 SD. On average, most of the blood glucose read out from the Nova StatStrip and SureStep were higher than laboratory plasma glucose throughout the glucose range with mean differences (SD) of 11.2 (8.4) mg/dL and 13.7 (6.8) mg/dL, respectively (Fig. 1). Results presented following the standard DIN EN ISO 15197, NCCLS and ADA are in Web Table II.

Fig. 1 Scatter plots of differences between laboratory plasma glucose and (a) Nova StatStrip and (b) SureStep.

When using POC glucometer to diagnose hypoglycemia, the sensitivity of Nova StatStrip was higher than SureStep but both were less than 70%. The specificities and PPV were 100% for both POC glucometer testings. Negative predictive value of both POC glucometers were less than 90% (Table I).

TABLE I Sensitivity, Specificity, Positive and Negative Predictive Values for Detecting  Different Plasma Glucose Levels* 

From ROC curve (not shown), cut-off levels that would yield the best sensitivity and specificity for both POC glucometers were less than 52 mg/dL, with NPV of approximately 97% (Table I). To diagnose blood glucose of less than 45 mg/dL, the sensitivity of both glucometers were 72% and 61%, respectively (Table II). From ROC curve (not shown), cut-off levels that would yield the best sensitivity and specificity for Nova StatStrip and SureStep to diagnose blood glucose of less than 45 mg/dL were less than 60 mg/dL and less than 55 mg/dL, respectively (Table II).

TABLE II	Sensitivity, Specificity, Positive and Negative Predictive Values for Detecting Different Plasma 

Glucose Levels* (For Therapeutic Goal of Plasma Glucose At 45 mg/dL).

Our results agree with previous studies which concluded that glucose reagent strips should be considered only as a screening test, not as a diagnostic test, due to their questionable reliability [2,6,7]. Though, the difference of glucose level between both POC glucometers and laboratory is normally distributed, 95% of the differences lie in a wide range (between -5.3 and 27.7 mg/dL for Nova StatStrip and between 0.4 and 27.0 mg/dL for SureStep), and are not acceptable for clinical purposes.

Validity of POC glucometer also depends on pre-analytical processes. Using sample tubes containing sodium fluoride (NaF) has been recommended in order to minimize ex-vivo glycolysis [13, 14]. Such a method was adopted in our study. Though fluoride is the best available preservative for blood glucose measurement, the antiglycolytic action may delay for up to 4 hours [15]. The mean plasma glucose concentration could decrease by 4.3% at 1 hour and by 4.6% at 2 hours in blood kept in tube containing NaF [15,16]. Our results are not applicable for different glucose preservative used or different glucose measurement process.

Mann, et al. [17] found that error rates for low hematocrit between 25% and 34% ranged from 16.4% to 18.4%. Hematocrit of the infants recruited in our study ranges from 27% to 65% which is within the operational range of accuracy (25-65%) for both POC glucometers. There were only three specimens with hematocrit less than 34% in our study. Hence, glucometer error secondary to low hematocrit should not be a major problem in this study.

NCCLS recommended that discrepancies in blood glucose measurements should be less than 15 mg/dL when actual glucose concentrations were less than 99 mg/dL. Subsequently, 75.3% of Nova StatStrip fulfilled such standard, compared to 61.8% of SureStep [11]. More-over, using the ADA standard, none of the POC glucometers tested demonstrated satisfactory accuracy. Nova StatStrip and SureStep had only 22.6% and 9.6% of their tests achieved a total error of less than 10% at glucose concentration ranging from 30 to 400 mg/dL [12].

In clinical practice, sending for confirmation plasma glucose in all cases would add to the expense, considering how often glucose levels are assessed in newborns. Using POC glucometer alone, we may miss 38% and 47% of hypoglycemia (POC glucose level <40 mg/dL) by Nova StatStrip and SureStep, respectively. With ROC analysis, the cut-off level with the best sensitivity and specificity of both POC glucometers are 52 mg/dL. At this cut-off level, we will still miss approximately 6-8% of hypoglycemia. For POC glucose reading of less than 40 mg/dL, we could be confident of the diagnosis of hypoglycemia without sending confirmatory plasma glucose. Thus, confirmatory plasma glucose for diagnosis of hypoglycemia would be required when POC glucose readings are between 39 and 63 mg/dL for Nova StatStrip and between 39 and 62 mg/dL for SureStep.

Practically, different blood glucose levels are used as the "operational threshold" and the "therapeutic objective" [2]. Mostly accepted therapeutic goal is to keep blood glucose higher than 45 mg/dL [2,13]. For this purpose of defining blood glucose less than 45 mg/dL, the cut-off levels that confirmatory blood glucose is not necessary would be different. Confirmatory plasma glucose for diagnosis of blood glucose less than 45 mg/dL is needed when POC glucose readings are between 44 and 63 mg/dL for Nova StatStrip and between 44 and 69 mg/dL for SureStep.

The clinical application of our study is limited to clinically stable infants without medications that might interfere with glucose measurement. The participants in this study were infants aged between 1 and 24 hours. If the POC glucose measurement is done in older infants, the clinicians must be aware of the possibility of hyperbilirubinemia interfering with glucose measurement.

Both POC glucometers in this study did not meet the minimum requirement of DIN EN ISO 15197, NCCLS, and ADA. A degree of caution should be exercised in the interpretation of POC glucose measurements as they may not possess sufficient accuracy to replace laboratory plasma glucose results. Individual devices may need their own operational cut-off values for sending plasma glucose for confirmation.

Acknowledgements: Dr Chulaluk Komoltri for helping in statistical analysis.

Contributors: NS, PS, KT, WP, KR and VP: Conception and design, analysis and interpretation of data, drafting the manuscript, critical revision of the manuscript for important intellectual content and final approval of the version to be published; IL: Acquisition of data, critical revision of the manuscript for important intellectual content, final approval of the version to be published.

Funding: This study was supported by Siriraj Routine to Research Management Fund. The strips of Nova StatSripTM were provided by Meditop Co, Ltd, Bangkok, Thailand;
Competing interests: None stated.

1. Burns CM, Rutherford MA, Boardman JP, Cowan FM. Patterns of cerebral injury and neurodevelopmental outcomes after symptomatic neonatal hypoglycemia. Pediatrics. 2008;122:65-74.

2. Cornblath M, Hawdon JM, Williams AF, Aynsley-Green A, Ward-Platt MP, Schwartz R, et al. Controversies regarding definition of neonatal hypoglycemia: suggested operational thresholds. Pediatrics. 2000;105:1141-5.

3. Innanen VT, DeLand ME, deCampos FM, Dunn MS. Point-of-care glucose testing in the neonatal intensive care unit is facilitated by the use of the Ames Glucometer Elite electrochemical glucose meter. J Pediatr. 1997;130:151-5.

4. Leonard M, Chessall M, Manning D. The use of a Hemocue blood glucose analyser in a neonatal unit. Ann Clin Biochem. 1997;34:287-90.

5. Girouard J, Forest JC, Masse J, Leroux M, Bradburn NC, Noblet TC, et al. Multicenter evaluation of the Glucometer Elite XL meter, an instrument specifically designed for use with neonates. Diabetes Care. 2000;23:1149-53.

6. Ho HT, Yeung WK, Young BW. Evaluation of "point of care" devices in the measurement of low blood glucose in neonatal practice. Arch Dis Child Fetal Neonatal Ed. 2004;89:F356-9.

7. Michel A, Kuster H, Krebs A, Kadow I, Paul W, Nauck M, et al. Evaluation of the Glucometer Elite XL device for screening for neonatal hypoglycaemia. Eur J Pediatr. 2005;164:660-4.

8. Holtzinger C, Szelag E, Dubois JA, Shirey TL, Presti S. Evaluation of a new POCT bedside glucose meter and strip with hematocrit and interference corrections. Point of Care. 2008;7:1-6.

9. Scher MS. Seizures in neonates. In: Martin RJ, Fanaroff AA, Walsh MC, editors. Fanaroff and Martin’s neonatal-perinatal medicine: diseases of the fetus and infant. Philadelphia: Mosby Elsevier; 2006. p. 956-76.

10. Mahoney JJ, Ellison JM. Assessing glucose monitor performance—a standardized approach. Diabetes Technol Ther. 2007;9:545-52.

11. Sacks DB, Bernhardt P, Dunka LJ, Goldstein DE, Hortin GL, Mueller P. Point-of-Care Blood Glucose Testing in Acute and Chronic Care Facilities; Approved Guideline. 2002. Available from: URL: http://www.clsi.org/source/orders/free/c30-a2.pdf. Accessed on March 24, 2011.

12. American Diabetes Association. Self-monitoring of blood glucose. Diabetes Care. 1994;17:81-6.

13. Adamkin DH. Postnatal glucose homeostasis in late-preterm and term infants. Pediatrics. 2011;127:575-9.

14. World Health Organization. Use of anticoagulants in diagnostic laboratory investigations 2002; Available from: URL: http://whqlibdoc.who.int/hq/2002/WHO _DIL_LAB_99.1_Rev.2.pdf. Accessed on March 24, 2011.

15. Chan AY, Swaminathan R, Cockram CS. Effectiveness of sodium fluoride as a preservative of glucose in blood. Clin Chem. 1989;35:315-7.

16. Gambino R, Piscitelli J, Ackattupathil TA, Theriault JL, Andrin RD, Sanfilippo ML, et al. Acidification of blood is superior to sodium fluoride alone as an inhibitor of glycolysis. Clin Chem. 2009;55:1019-21.