Endocrinol Nutr.

2012;59(1):28---34

ENDOCRINOLOGÍA Y NUTRICIÓN

www.elsevier.es/endo

ORIGINAL ARTICLE

Validation of a glucose meter at an intensive care unit夽

Miguel Ángel Castaño López ∗ , José Luís Fernández de Liger Serrano,
José Luís Robles Rodríguez, Tomas Márquez Márquez

Servicio de Análisis Clínicos, Hospital Juan Ramón Jiménez, Huelva, Spain

Received 6 May 2011; accepted 22 August 2011

Available online 16 February 2012

KEYWORDS Abstract
Glucose meter; Introduction: Hyperglycemic patients admitted to intensive care units (ICUs) have higher mor-
Glycemia; bidity and mortality than normoglycemic patients. Blood glucose levels of ICU patients are
Intensive care unit; usually measured with a glucose meter. The aim of this study was to evaluate a glucose meter
Evaluation (StatStrip, Nova Biomedical) to assess its agreement with the standard laboratory method for
testing glucose.
Material and methods: Eighty-nine different samples were collected from patients (76.4% men
and 23.6% women) admitted to an ICU from September to December 2010. Each blood sample
was collected into two tubes, a lithium heparin tube and an EDTA tube. The total blood aliquot
was used to measure glycemia using the glucose meter. The lithium heparin tube was processed
at the same time for measuring plasma glucose (Cobas 6000 Analyzer, Roche Diagnostics, SA).
Agreement between the two methods was assessed according to the EP-9-A2 Clinical Laboratory
Standards Institute guideline.
Results: Mean whole blood glucose level measured by the glucose meter was
126.53 + 49.28 mg/dL (range, 33.5---431 mg/dL), while mean plasma glucose value mea-
sured by the laboratory reference method was 138.13 + 78.6 mg/dL (range, 43---451 mg/dL).
Correlation coefficient was 0.99, with a 95% confidence interval of 0.98---0.99. Coefficient of
determination (R2 ) was 0.97, and intraclass correlation coefficient was 0.99 with a 95% CI of
0.98---0.99.
Conclusions: The tested glucose meter (StatStrip) shows a good linear association, precision,
and accuracy when compared to the laboratory reference method. This device is adequate for
glucose monitoring.
© 2011 SEEN. Published by Elsevier España, S.L. All rights reserved.

夽 Please cite this article as: Castaño López MÁ, et al. Validación de un glucómetro en una unidad de cuidados intensivos. Endocrinol Nutr.

2012;59(1):28---34.
∗ Corresponding author.

E-mail address: [email protected] (M.Á. Castaño López).

2173-5093/$ – see front matter © 2011 SEEN. Published by Elsevier España, S.L. All rights reserved.
Validation of a glucose meter at an intensive care unit 29

PALABRAS CLAVE Validación de un glucómetro en una unidad de cuidados intensivos

Glucómetro;
Resumen
Glucemia;
Introducción: Los pacientes ingresados en las unidades de cuidados intensivos (UCI) con
Unidad de cuidados
hiperglucemia presentan mayor morbi-mortalidad que los pacientes normoglucémicos.
intensivos;
Habitualmente, la monitorización de la glucemia de los pacientes en las unidades de cuida-
Evaluación
dos intensivos es realizado por medio de glucómetros. El objetivo del estudio fue evaluar un
glucómetro (StatStrip, Nova Biomedical) para determinar su grado de acuerdo con el método
habitual de determinación de la glucemia en el laboratorio.
Material y métodos: Se recogieron 89 muestras de diferentes pacientes (76,4% hombres y 23,6%
mujeres) ingresados en una UCI durante los meses de septiembre a diciembre del 2010. En cada
extracción, se recogió un tubo de heparina litio y otro tubo de EDTA. La alícuota de sangre total
era utilizada para la determinación de glucemia mediante el glucómetro. El tubo de heparina
litio era procesado a la misma vez para la determinación de la glucemia plasmática (Analizador
Cobas 6000, Roche Diagnostic, SA). Para evaluar el grado de acuerdo entre los dos métodos,
seguimos el procedimiento indicado en la guía EP-9-A2 del Clinical and Laboratory Standards
Institute (CLSI).
Resultado: La glucemia en sangre total medida por el glucómetro presentaba un valor medio de
126,53 ± 49,28 mg/dL con un rango de 33,5 a 431 mg/dL y la glucemia plasmática del método
de laboratorio reflejaba un valor medio de 138,13 ± 78,6 mg/dL con un rango de 43---451 mg/dL.
El coeficiente de correlación entre ambos métodos fue de 0,99 con un intervalo de confianza
al 95% (IC) de 0,98 a 0,99; el coeficiente de determinación (R2 ) fue de 0,97 y el coeficiente de
correlación intraclase fue 0,99 con un IC de 0,98 a 0,99.
Conclusiones: El glucosímetro evaluado (StatStrip) presenta una buena asociación lineal, pre-
cisión y exactitud, cuando es comparado con el método de referencia del laboratorio clínico.
Es un dispositivo adecuado para la monitorización de la glucosa.
© 2011 SEEN. Publicado por Elsevier España, S.L. Todos los derechos reservados.

Blood glucose changes are among the most common emptying, decrease drug elimination (particularly nar-
metabolic changes in both diabetic and non-diabetic inpa- cotics), impair WBC function, increase frequency of
tients. Patients admitted to intensive care units (ICUs) bacteremia and fungemia, and decrease implantation of skin
have increased insulin requirements because of pain, grafts in patients with burns.5---8
trauma, surgery, sepsis, hypoxia, burns, cardiovascular Patients with abnormal blood glucose values admitted
changes, psychic stress, drugs, the administration of glu- to ICU have a higher mortality rate than normoglycemic
cose solutions, and so on.1 These patients experience patients.9 All studies recommend treatment of hyper-
a number of metabolic and circulatory changes defined glycemia (blood glucose > 140 mg/dL), but no agreement
as systemic inflammatory response syndrome (SIRS). SIRS exists as to the criteria and the type of monitoring required
induces the activation of a number of neuroendocrine in these patients. Today, most physicians consider hypo-
and inflammatory mediators (cortisol, glucagon, growth glycemia more serious than hyperglycemia.
hormone, catecholamines, glucocorticoids, and cytokines Van den Berghe et al. showed that strict monitoring
such as interleukin-1, inteleukin-6, and tumor necro- reduced morbidity and mortality rates in patients admitted
sis factor alpha) which increase hepatic gluconeogenesis to ICU.10,11 Blood glucose monitoring is a routine practice at
and peripheral insulin resistance, causing so-called stress these units. However, recently reported studies state that
hyperglycemia. monitoring causes a significant increase in the risk of hypo-
Until recently, stress hyperglycemia was thought to pro- glycemia with increased morbidity and mortality.12---14 Blood
vide adequate glucose levels for the brain, skeletal muscle, glucose levels are usually measured with glucose meters at
myocardium, and other vital organs in conditions where ICUs and other hospital units. Glucose meters are widely
glucose demand was increased, but it has been found to used because they provide rapid results, avoid repeated
increase infections and morbidity and mortality as the result blood sampling, and only very small sample volumes are
of increased oxidative damage and the enhancement of required.
proinflammatory response, amongst other actions.2 Glucose meters may give false readings due to various
Hyperglycemia can have severe consequences: it can reported interferences (including drugs, maltose, galactose,
increase cerebral ischemia, delay wound healing, increase hematocrit, ascorbic acid, xylose, uric acid, oxygen, and
the frequency of infections, and worsen the prognosis of bilirubin, amongst others),15---18 which often occur at ICUs.
primary disease.3,4 In addition, transient hyperglycemia An assessment of the analytical accuracy and precision of
may cause water and electrolyte changes, dehydration glucose meters used at the different hospital units is, there-
(osmotic diuresis), and lactic acidosis, decrease brain flow, fore, required. A comparison of their results to those of the
impair mental status, delay wound healing, delay gastric routinely used laboratory method is also recommended.
30 M.Á. Castaño López et al.

The EP-9-A2 consensus guideline of the Clinical and Lab- a. Between the mean values of both methods.
oratory Standards Institute (CLSI) allows for the assessment b. Between the individual Y values and the mean X val-
and comparison of two methods that measure the same ues.
analyte using patient samples, provided that one of the c. Between the differences between mean Y and mean
methods is the reference or standard laboratory method.19 X for each method versus the sum of mean Y and X
The purpose of this study was to assess the StatStrip glucose values divided by 2.
meter (Nova Biomedical, Boston, USA) in order to measure d. Between the difference between each individual Y
its agreement with the standard blood glucose measurement and X value versus the sum of mean Y and X values
method used at the laboratory of our hospital using the CLSI divided by 2.
EP-9-A2 consensus guideline. 4. These plots allowed for verifying whether a linear rela-
tionship existed, and whether systematic or random error
existed between both the methods studied.
Materials and methods 5. The correlation coefficient. If calculated r was 0.975
or greater or the coefficient of determination was
Eighty-nine samples were collected from different patients 0.95 or greater, it was considered that the variability of
(77.6% males and 22.4% females) admitted to the ICU of Juan X was acceptable, and that the potential measurement
Ramón Jiménez Hospital from September to December 2010. error of X was compensated for by the wide range of X
A lithium heparin and an EDTA tube were collected at each values.
sampling. Hematocrit was measured in the EDTA tube using 6. The performance of Bland---Altman plots. These allowed
a Sysmex blood cell counter (Roche Diagnostics, Mannheim, for determining whether significant differences existed
Germany). between the two measurement techniques studied and
The StatStrip glucose meter uses a procedure based on if constant scatter existed. If constant scatter existed, a
a strip consisting of four layers with different functions: regression analysis comparing both techniques was to be
a first isolating layer guarantees the preservation of the made.
lower layers; a second layer distributes the sample (50 ␮L) 7. The estimate of predicted systematic error and its confi-
to the third layer; a third layer contains four wells (12 ␮L in dence intervals (CI).
total) in which measurements are made of the total glucose 8. Intraclass correlation coefficient. Agreement was consid-
(first well), the amount attributable to interferences (sec- ered to be very good when this coefficient was greater
ond well), and the hematocrit (third well). Adequate sample than 0.90.20
volume is controlled in the fourth well. The last layer is
used for glucose meter control and calibration. This consists MedCalc 11.4 software was used for statistical analysis.
of a gold layer that confers stability in all environmental
conditions. Results
Blood glucose was measured in the lithium heparin tube A total of 89 patients with a mean age of 57.4 ± 14.8 years
according to the protocol below. Upon sample reception, (mean and standard deviation) were enrolled into the study.
a 1 mL aliquot was separated and labeled with the same Table 1 shows other characteristics of the study popula-
number as the sample. The tube was closed again and cen- tion. Most admissions were for major surgery (28.4%). Of
trifuged at 3500 rpm for 10 min. The whole blood aliquot these patents, 38.8% had diabetes mellitus and 89.5% were
was used to measure blood glucose (in duplicate) using the receiving multiple drugs.
glucose meter. Most patients had low hematocrit values
Blood glucose measurement in whole blood (aliquot) with (mean ± standard deviation, 33.13 ± 6.56%; range, 18---55).
the glucose meter and the measurement of plasma glucose Blood glucose (mean ± standard deviation) in whole blood
in a Cobas 6000 analyzer (Roche Diagnostics) by the hexoki- measured by the glucose meter was 126.5 ± 49.28 mg/dL
nase method (in duplicate) were done at the same time. The (range, 33.5---431 mg/dL), and laboratory blood glucose
reference method was the measurement of glucose using the (Cobas 6000) 138.13 ± 49.28 mg/dL (range, 43---451). Coef-
hexokinase method routinely used at our center for measur- ficients of variation (CVs) were 3.57% for the glucose meter
ing blood glucose levels. and 1.59% for the Cobas method. No outliers were found
The procedure indicated in the EP-9-A2 guideline was in the tested parameters of whole blood glucose measured
used for the statistical analysis of the results as summarized with the glucose meter and plasma glucose measured with
below: the reference method.
As regards linearity and constant scatter, both measure-
1. Verification of the normality of the tested parameters, ment methods followed a linear relationship, as they showed
for which a Kolmogorov---Smirnov statistical test was a Pearson’s correlation coefficient >0.975. This coefficient
used. of determination was greater than 0.95 with constant scat-
2. The detection of outliers. Absolute differences between ter, and linear regression could therefore be used to verify
duplicates of each method were compared. These differ- linearity of the measurement techniques tested (Fig. 1).
ences were to be not greater than 4 times the mean of The variables tested, whole blood glucose using a
the absolute differences. No outliers were found in this glucose meter and plasma glucose using Cobas 6000,
study. followed a normal distribution (p = 0.310 and p = 0.167
3. Charts for comparing the linear relationship between the respectively).
values of both methods. Four scatter plots were used The statistical parameters calculated included: (1) cor-
employing equal scales: relation coefficient: 0.99 (95% confidence interval (CI):
Validation of a glucose meter at an intensive care unit 31

Table 1 Characteristics of study patients.

Sex

Male Female
No. No.
Associated conditions
Diabetes mellitus 39.33% 23 12
Multiple drugs 89.9% 62 18
Reason for admission
Traffic accident 23.6% 15 3
Major surgery 28.1% 17 7
Stroke 9% 11 1
Severe respiratory insufficiency 21.4% 13 4
Other 17.9% 9 6
SD: standard deviation.

0.98---0.99); (2) regression line: Cobas 6000 blood glu- Using the criterion of standard 15197 of the International
cose (y) = 1.67 + 1.03 × Cobas b221 blood glucose (X); (3) Organization for Standardization (ISO), 100% of the values
coefficient of determination (R2 ): 0.97; and (3) intraclass obtained with the glucose meter were within ±15% for val-
correlation coefficient: 0.99 (CI: 0.98---0.99). ues less than 75 mg/dL, and 100% were within ±20% for
An analysis was also made using Bland---Altman plots to values of 75 mg/dL or more. For the criteria of the Amer-
verify that there were no significant differences between ican Diabetes Association (ADA), only 60.6% of values found
the pairs of points between the two types of equipment with the glucose meter met the goal of ±5% as compared
being compared (Fig. 2). It could be shown that 5.2% of data to the values obtained with the reference method. Finally,
exceeded two standard deviations. using the criterion of the Clinical Laboratory Improvement
Table 2 shows the quality specifications of the differ- Amendments (CLIA), 91.01% of the values met the goal of
ent international organizations. Table 3 gives the results ±5.4 mg/dL or 10% as compared to the values obtained with
achieved with the glucose meter based on the quality objec- the reference method.
tives established by these international organizations.
The most interesting results given in Tables 2 and 3 are
discussed below. Using the CLSI criterion, 100% of values Discussion
obtained with the glucose meter were within ±14.7 mg/dL
of the values obtained with the reference method for blood The use of glucose meters at hospitals for measuring blood
glucose values less than 75 mg/dL, and 96.01% were within glucose in diabetics is very common. The main problems of
15% of values of the reference method for concentrations
of 75 mg/dL or more. A single sample was 21.2% higher in
the glucose meter as compared to the reference method.
Cobas 6000-MEDIA MEDIA GLYCEMIC BLOOD SUGAR Nova StatStrip

50
Blood glucose by cobas 6000 (mg/dL)

500 40
450
30 +1.96 SD
400
27.5
350 20
300
10 MEDIA
250
200 0 5.9
150
−10
100 −1.96 SD
150 −20 −15.7
0
−30
0 100 200 300 400 500
Blood glucose by nova statstrip (mg/dL) 0 100 200 300 400 500 600
(MEDIA MEDIA 6000 + BLOOD SUGAR GLUCOSE COBAS StatStrip Nova) / 2
Correlation coefficient: 0.99. 95% confidence interval (CI): 0.98‐0.99)
SD: Standard Deviation ng
Regression line: y = 1.67 + 1.03 x Coefficient of determination
(R2): 0.98 Intraclass correlation coefficient: 0.99. CI: 0.98‐0.99
Figure 2 Bland---Altman plot between Cobas 6000 and Stat-
Figure 1 Regression line between Cobas 6000 and StatStrip. Strip.
32 M.Á. Castaño López et al.

Table 2 Quality objectives according to different international organizations.

CLIA Values obtained ±10% or 5.4 mg/dL as compared to values with standard method
CLSI 100% values ± 14.7 mg/dL for values less than 75 mg/dL
100% values ± 15% for values ≥75 mg/dL
ISO 15197 95% values ± 15% for values less than 75 mg/dL
100% values ± 20% for values ≥75 mg/dL
ADA 100% values ± 5% from the reference method value
ADA: American Diabetes Association; CLIA: Clinical Laboratory Improvement Amendments; CLSI: Clinical and Laboratory Standards
Institute; ISO: International Organization for Standardization.

such measurements include their accuracy and interference glucose meter assessed therefore met both the CLSI and the
by hematocrit, very high protein levels, and the use of ISO criteria.
certain drugs. The use of glucose meters at the different Another criterion for assessing the precision of evaluation
units requires a verification of agreement with the routine was proposed by the ADA, which recommended that meter
laboratory method. Different authors have assessed readings should be within ±5% of the reference method.16
the reliability of different glucose meters and However, no currently available glucose meter has
have compared these to the standard laboratory achieved this goal.16,25,26 The proportion of measurements
methods.16,21 meeting this requirement was 60.5%, and was lower in some
The EP-9-A2 was used for method validation. This reports and higher in others.16,21,25,26 This suggests that Stat-
guideline may appear complex, but is easy to standard- Strip is an improved method but that it does not meet the
ize when used. The procedures used by this guideline ADA requirements.
do not correspond to the classical statistical methods for According to the CLIA, the results should be within 10% of
method validation, but because of its effectiveness as the results of the reference method or ±5.4 mg/dL. In our
a validation method, it has been proposed as a valida- study, 91.0% of the results were within that limit.
tion tool of both accredited laboratories with a flexible Poirier et al.27 classified glucose meters as good, accept-
scope and for method validation of laboratories aiming at able, and unacceptable for clinical use. A glucose meter
accreditation.22 is defined as good when 60% of its results are within
In this study, the Nova StatStrip glucose meter showed ±10% of the reference method values. The glucose meter
a good linearity within the studied range and a coeffi- assessed in this study may, therefore, be classified as
cient of variation of 3.57%. Skeie et al. recommended a good.
maximum coefficient of variation of 5%.16 The compar- A variety of substances cause interference with
ison of the StatStrip glucose meter and the reference glucose meters.16,28---31 The main such substances are
laboratory method (Cobas 6000) showed a very high cor- acetaminophen, ascorbic acid, and maltose. Other authors
relation, with a slope close to 1 and an intercept close found that variations in hematocrit levels may affect glu-
to 0. cose meter measurements.30,32 This interference occurs with
The best parameter for assessing the agreement of a some meters, even when the manufacturer’s recommenda-
quantitative variable using two different instruments is tions are followed.33 Different studies have reported that
intraclass correlation coefficient (ICC). Agreement is con- the above substances do not interfere with the StatStrip glu-
sidered to be very good when ICC is >0.90.20 In this cose meter, but most such studies were conducted in vitro.
study, the glucose meter showed a very good agreement Although an interference study was not made in our study,
with the reference method (ICC = 0.99). The glucose meter the occurrence of interferences, particularly the most com-
provided glucose values slightly lower than those found mon ones,16,28,30 may be shown, because the characteristics
with the reference laboratory method, with a mean dif- of the patients admitted to ICUs (multiple drug treatment,
ference of 5.9 mg/dL, similar to that reported by other low hematocrit, severe diseases, various treatments, etc.)
authors.16,24,25 This difference may be explained by the may have caused the appearance of such interferences. In
samples used, because glucose concentration in whole this study, no significant interferences or accuracy losses
blood was approximately 10---15% lower than plasma glucose were found.
levels.24 The glucose meter assessed (Nova StatStrip) showed a
In any case, the compliance of glucose meters with good linearity, precision, and correlation as compared to
the established quality recommendations is required. With the reference method of the clinical laboratory. The Stat-
regard to CLSI and ISO 15197:2003 criteria,22,23 the results of Strip meets the quality requirements of all the different
this study showed that 100% of glucose meter readings were international organizations, except for those of the ADA.
within ±14.7 mg/dL of those obtained with the reference However, there is currently no glucose meter that meets
method for values less than 75 mg/dL. 96.01% of readings the requirements set by this international organization. It
were within 15% of Cobas 6000 values for concentrations may, therefore, be concluded that this glucose meter is an
higher than 75 mg/dL. A 21.2% higher value was found with adequate device for blood glucose monitoring in patients
StatStrip as compared to Cobas 6000 in a single sample. The admitted to ICUs.
 Validation of a glucose meter at an intensive care unit
Table 3 Values obtained with the glucose meter assessed as a function of the quality objectives of the different international organizations.
<75 mg/dL (21 samples) ≥75 mg/dL (68 samples)

0---5.0% 5.1---8.0% 8.1---10.0% >10.0% 0---5% 5---10% 10.1---15% >15%

No. % No. % No. % No. % No. % No. % No. % No. %

ADA 15 71.43 4 19.04 2 9.52 0 0 51 73.9 9 13 7 10 1 1

<5.40 mg/dL 5.40---10.85 10.86---14.7 >14.7 mg/dL <5% 5.1---10% 10.1---15% >15%

No. % No. % No. % No. % No. % No. % No. % No. %

CLSI 15 71.43 4 19.04 2 9.52 0 0 51 73.9 9 13 7 10 1 1

0---5% 5.1---15% 15.1---20% >20% 0---10% 10.1---20% 20---30% >30%

No. % No. % No. % No. % No. % No. % No. % No. %

ISO 15 71.43 6 28.57 0 0 0 0 60 88.2 7 10 1 1.5 0 0

0---5.4 mg/dL >5.4 mg/dL 0---5% 5.1---10% >10%

No. % No. % No. % No. % No. %

CLIA 63 70.79 26 29.21 66 74.2 15 16.9 8 8.99
ADA: American Diabetes Association; CLIA: Clinical Laboratory Improvement Amendments; CLSI: Clinical and Laboratory Standards Institute; ISO International Organization for Standard-
ization; No.: number of samples.

33
34 M.Á. Castaño López et al.

Conflicts of interest 18. Karon BS, Griesmann L, Scott R, Bryant SC, Dubois JA,
Shirey TL, et al. Evaluation of the impact of hemat-
ocrit and other interference on the accuracy of hospital-
The authors state that they have no conflicts of interest.
based glucose meters. Diabetes Technol Ther. 2008;10:
111---20.
References 19. NCCLS. Method comparison and bias estimation using patient
samples: approved guideline. 2nd ed; 2002 [NCCLS document
1. Surwit RS, Schneider MS, Feinglos MN. Stress and diabetes mel- EP9-A2 NCCLS].
litus. Diabetes Care. 1992;15:1413---22. 20. Prieto L, Lamarca R, Casado A. La evaluación de la fiabilidad en
2. Kost GJ, Vu HT, Lee JH, Bourgeois P, Kiechle Fl, Martin C, et al. las observaciones clínicas: el coeficiente de correlación intra-
Multicentre study of oxygen-insensitive handheld glucose point- clase. Med Clin (Bar). 1998;110:142---5.
of-care testing in critical care/hospital/ambulatory patients in 21. Soldin SJ, Devairakkam PD, Agarwalla PK. Evaluation of the
the United States and Canada. Crit Care Med. 1998;26:581---90. Abbott PCx-point of care glucose analyzer in a pediatric hos-
3. Pulsinelli WA, Levy DE, Sigsbee B, Scherer P, Plum F. Increased pital. Clin Biochem. 2000;33:319---21.
damage after ischemic stroke in patients with hyperglycemia 22. Bernabeu FA, Corcho MA, Redondo M, Sivera L, Coca MC, Arribas
with or without established diabetes mellitus. Am J Med. I. Procedimiento de validación de magnitudes en un gasómetro.
1983;74:540---4. Aplicación al alcance flexible en la Norma ISO 15189. Rev Lab
4. McMurry Jr JF. Wound healing with diabetes mellitus. Better Clin. 2010;3:58---62.
glucose control for better wound healing in diabetes. Surg Clin 23. National Committee for Clinical Laboratory Standards. Point-of-
North Am. 1984;64:769---78; care blood-glucose testing in acute and chronic care facilities:
Rayfield EJ, Ault MJ, Keusch GT, Brothers MJ, Nechemias C, approved guideline. 2nd ed. Wayne, PA: NCCLS; 2002 [NCCLS
Smith H. Infection and diabetes: the case for glucose control. document C30-A2].
Am J Med. 1982;72:439---50. 24. International Organization for Standardization. In vitro diag-
5. Queale WS, Seidler AJ, Brancati FL. Glycemic control and sliding nostic test systems----requirements for blood glucose monitoring
scale insulin use in medical inpatients with diabetes mellitus. systems for self-testing in managing diabetes mellitus. Geneva,
Arch Intern Med. 1997;157:545---52. Switzerland: ISO; 2003 [ISO 15197:2003].
6. Pretto Z, Koproski J, Poretsky L. Hospital management of dia- 25. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, McDonald JM,
betes in the post-DCCT era. Practical Diabetol. 1995;14:2---7. Parrott M. Guidelines and recommendations for laboratory anal-
7. Hirsch IB, McGill JB. Role of insulin in management of ysis in the diagnosis and management of diabetes mellitus. Clin
surgical patients with diabetes mellitus. Diabetes Care. Chem. 2002;48:436---72.
1990;13:980---91. 26. Lippi G, Salvagno GL, Guidi GC, Negri M, Rizzotti P. Evaluation
8. Rossetti L, Giaccari A, DeFronzo RA. Glucose toxicity. Diabetes of four portable self-monitoring blood glucose meters. Ann Clin
Care. 1990;13:610---30. Biochem. 2006;43:408---13.
9. Krinsley J. Association between hyperglycemia and increased 27. Poirier JY, Le Prieur N, Campion L, Guilhem I, Allannic
hospital mortality in a heterogeneous population of critically ill H, Maugendre D. Clinical and statistical evaluation of self-
patients. Mayo Clin Proc. 2003;78:1471---8. monitoring blood glucose meters. Diabetes Care. 1998;21:
10. Van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx 1919---24.
F, Schetz M, et al. Intensive insulin therapy in the critically ill 28. Holtzinger C, Szelag E, Dubois JA, Shierey TL, Presti S.
patients. N Engl J Med. 2001;345:1359---67. Evaluation of a new POCT bedside glucose meter and strip
11. Van den Berghe G, Wilmer A, Hermans G, Meersseman W, with hematocrit and interference corrections. Point Care.
Wouters PJ, Milants I, et al. Intensive insulin therapy in the 2008;7:1---6.
medical ICU. N Engl J Med. 2006;354:449---61. 29. Louie RF, Tang Z, Sutton DV, Lee JH, Kost GJ. Point-of-care
12. Finfer S, Chittock DR, Su SY, Blair D, Foster D, Dhingra V, et al. glucose testing: effects of critical care variables, influence of
Intensive versus conventional glucose control in critically ill reference instruments, and a modular glucose meter design.
patients. N Engl J Med. 2009;360:1283---97. Arch Pathol Lab Med. 2000;124:257---66.
13. Chen L. A literature review of intensive insulin therapy and mor- 30. Tang Z, Du X, Louie RF, Kost GJ. Effects of drugs on glucose
tality in critically ill patients. Clin Nurse Spec. 2010;24:80---6. measurements with handheld glucose meters and a portable
14. Bilotta F, Rosa G. Glucose management in the neurosurgi- glucose analyzer. Am J Clin Pathol. 2000;113:75---86.
cal patient: are we yet any closer? Curr Opin Anaesthesiol. 31. Karon BS, Griesmann L, Scott R, Bryant SC, Dubois JA, Shirey
2010;23:539---43. TL, et al. Evaluation of the impact of hematocrit and other
15. Skeie S, Thue G, Sandberg S. Patient-derived quality specifica- interference on the accuracy of Hospital-Based Glucose Meters.
tions for instruments used in self-monitoring of blood glucose. Diabetes Technol Ther. 2008;10:111---20.
Clin Chem. 2001;47:67---73. 32. Tang Z, Lee JH, Louie RF, Kost GJ. Effects of different hema-
16. Vanavanan S, Santanirand P, Chaichanajarernkul U, Chittamma tocrit levels on glucose measurements with handheld meters
A, Dubois JA, Shirey T, et al. Performance of a new interference- for point of care testing. Arch Pathol Lab Med. 2000;124:
resistant glucose meters. Clin Biochem. 2010;43:186---92. 1135---40.
17. Goudable J, Pollet J, DuBois JA. Inaccuracy of glucose meters. 33. Louis PS, Ethier J. An evaluation of three glucose meter
Automatic correction for hematocrit variations and the pres- systems and their performance in relation to criteria of accept-
ence of exogenous interfering components. Ann Biol Clin. ability for neonatal specimens. Clin Chim Acta. 2002;322:
2008;66:647---55. 139---48.