| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
Department of Clinical Chemistry, University Hospital, S-581 85 Linköping, Sweden.
a Author for correspondence. Fax + 46 13 22 32 40; e-mail ulf.hannestad{at}klk.us.lio.se
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResults and DiscussionReferences |
|---|
Key Words: indexing terms: gas chromatography–mass spectrometry • quality control • near-patient analysis • HemoCue glucose meter compared
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionReferences |
|---|
The technique of isotope dilution gas chromatography–mass spectrometry (ID GC-MS) has been used in many efforts to establish Reference Methods or Definitive Methods for glucose in serum (3)(4)(5)(6). This technique has also been used since the mid-1970s by the US National Institute of Standards and Technology (NIST) in their development of such methods for glucose and other endogenous analytes in serum (7)(8). Pioneers in the field of establishing Reference Methods and Definitive Methods were Björkhem et al. (9), who developed an ID GC-MS method for glucose in serum so as to evaluate routine methods for glucose determination.
All the currently published Reference Methods and Definitive Methods for glucose are applicable to measurement of glucose in serum and not in whole blood (10). The Reference Methods and Definitive Methods used by the NIST and others (3)(4)(5) for determination of glucose in serum all have very high accuracy and precision. This high accuracy and precision is in part achieved by the way the samples and standards are added in the methods. Serum and standards are measured gravimetrically, and the serum samples are bracketed by internal standard and standard samples having a concentration of glucose just above and below that of the serum sample. This requires preliminary quantification of glucose in the sample to find a proper concentration for the internal standard and the glucose standard.
We describe here an ID GC-MS method for glucose in whole blood. The derivatization of glucose is essentially performed as described by Stökl and Reinauer (5) and Magni et al. (6). The blood and standard solutions are sampled volumetrically. Instead of the bracketing technique presented by Cohen et al. (11), we use six standard solutions of glucose, ranging from 1.0 to 20.0 mmol/L, and a fixed concentration of the internal standard, [13C6]glucose.
We also present the results of an accuracy assessment of the HemoCue blood glucose analyzer in comparison with the ID GC-MS method as a Reference Method. The glucose was measured by split-sample technique in capillary blood of 67 diabetic patients and 62 nondiabetic persons.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionReferences |
|---|
HemoCue blood glucose analyzer.
The instrument and the
cuvettes came from HemoCue (Ängelholm, Sweden).
Sampling equipment.
Capillary blood from diabetics and
nondiabetics was collected with plain 50-µL Vitrex "microcap"
micropipettes (Modulohm I/S, Herlev, Denmark). The internal standard
solution, 200 µL, was manually delivered with a 40–200-µL
adjustable micropipette (Finnpipette, Labsystems OY, Helsinki,
Finland).
reagents
Standard solutions and controls.
We prepared standard
solutions from glucose (anhydrous, analytical reagent; Mallinckrodt
Chemicals, St. Louis, MO) or glucose Standard Reference Material (SRM
917a; NIST, Gaithersburg, MD). After drying the substance for 24 h
at 90 °C, we stored it over a drying agent. Deionized water was
added to 360.3 mg of glucose [as weighed with a semimicro balance
(Model MC210P; Sartorius, Göttingen, Germany), which has a
readability of 0.01 mg and a specified maximal inaccuracy of 0.03%
when weighing 360.3 mg], to give a final volume of 100 mL; the glucose
concentration of the solution was 20.0 mmol/L. Using volumetric
pipettes and volumetric flasks, we further diluted this solution to
10.0, 6.0, 4.0, 2.0, and 1.0 mmol/L, as follows. From the 20.0 mmol/L
standard we took 50.0 mL and diluted it to 100 mL with water, yielding
a glucose concentration of 10.0 mmol/L. From the 10.0 mmol/L solution
we took 30.0, 20.0, and 10.0 mL, diluting each to 50.0 mL to yield
glucose standard solutions of 6.0, 4.0, and 2.0 mmol/L, respectively.
The 1.0 mmol/L standard was prepared by taking 10.0 mL of the 10.0
mmol/L glucose solution and diluting it to 100 mL with water. These
standard solutions were divided into portions and stored at -20 °C;
repeated assays after various storage periods showed them to be stable
for at least 6 months under these conditions.
To determine the between-run CV, we analyzed Seronorm (Nycomed Pharma AS, Oslo, Norway; lot no. 182) as a control with each series of blood samples. For the low and high range of glucose concentration, Pathonorm L and Pathonorm H (Nycomed; lot nos. 010026 and 503418, respectively) were analyzed. To assess the accuracy of the method, we used SRM 909b Level I from NIST.
Internal standard solution.
[13C6]Glucose (Cambridge Isotope Labs.,
Andover, MA) was used as an internal standard. The solution was
prepared by weighing 9.31 mg of [13C6]glucose
and adding deionized water to a final volume of 50 mL, giving a
concentration of 1.0 mmol/L. The solution was stored at -20 °C in
5-mL portions.
Chemicals.
All chemicals were of analytical grade.
Pyridine was from May & Baker (Dagenham, UK) and acetic anhydride,
methanol, and hydroxylamine hydrochloride were purchased from Fluka
Chemie (Buchs, Switzerland). The pyridine was distilled before use and
stored over KOH pellets.
analytical procedure
Blood samples were collected from patients with diabetes mellitus
and from apparently healthy nondiabetic individuals. The sampling was
approved by the Committee for Medical Ethics at the University Hospital
of Linköping and was conducted in accordance with the rules of
the Committee. Capillary blood (50 µL) was taken with a microcap and
immediately transferred to a tube containing 200 µL of
[13C6]glucose, 1.0 mmol/L. The microcap
was washed five times with the internal standard solution and the tube
was thoroughly shaken. The sample was left on the bench for 1–2 h, and
1 mL of methanol was added for deproteinization. After mixing the
sample, we let it stand for 30–45 min and then centrifuged it for 10
min at 1000g. A portion (0.5 mL) of the methanol/water phase
was taken to dryness under a stream of nitrogen on a heating block at
50 °C. The glucose and [13C6]glucose in
the sample were then converted to aldononitrile derivatives by adding
150 µL of 0.2 mol/L hydroxylamine hydrochloride solution in pyridine
and heating at 90 °C for 40 min. After cooling, the derivatives were
acetylated by addition of 200 µL of acetic anhydride, followed by
heating at 90 °C for 60 min. The samples were then evaporated just
to dryness under nitrogen (at 50 °C), and the residues were
dissolved in 0.5 mL of chloroform. Before the GC-MS analysis, 15 µL
of the solution was diluted with 135 µL of chloroform in small glass
vials that were then capped with polytetrafluoroethylene septum caps.
By dissolving the samples in 0.5 mL of chloroform and diluting the
samples an additional 10-fold in chloroform, we could detect no
deterioration of peak shape from eventual residual hydroxylamine, not
even after several hundred injections. The vials were loaded into the
autosampler, which was programed to inject 1 µL of each sample into
the gas chromatograph. The standards and the controls were treated the
same as the blood samples. A calibration curve was produced and a
control sample was included in every batch of blood samples measured.
All samples, standards, and controls were measured singly, and results
were quantified by measuring the peak areas for m/z 314.0
and 242.1 for glucose and those for m/z 319.0 and 246.1 for
[13C6]glucose. The m/z 242.1 and
246.1 ions were also used as "qualifier ions" in the ChemStation
software. The qualifier ions served to confirm the identity of the
glucose and [13C6]glucose derivatives. The
concentration of glucose in an unknown sample was calculated from a
linear regression fit of the peak area ratios (m/z 314.0 +
242.1)/(m/z 319.0 + 246.1).
other studies
Influence of glycolysis.
The effect of the glycolysis in
the analysis procedure was tested on three different occasions by
collecting into EDTA K3-containing tubes a blood sample
from nondiabetic individuals. Blood aliquots were taken from the tubes
with 50-µL microcaps and transferred to seven small plastic tubes
containing 200 µL of internal standard solution. After the microcaps
were washed five times in the internal standard solution, each sample
in the plastic tubes was mixed. Methanol was then added to the tubes
after 5, 15, 30, 60, 120, 180, or 240 min. The samples were then
analyzed as described in the preceding section.
Estimation of accuracy, precision, and specificity.
The
accuracy of the method was assessed by measuring glucose in SRM 909b
Level I with five analyses at two different occasions with use of both
glucose from Mallinckrodt and the NIST SRM 917b as standard material.
Within-run imprecision was determined by multiple analyses of whole
blood, Seronorm, Pathonorm L, and Pathonorm H in the same series. The
between-run imprecision was calculated from independent single analyses
of the control samples on different days. Seronorm was analyzed as a
single sample on 36 different days within a period of 10 months;
Pathonorm L and Pathonorm H were measured as single samples on 19 and
21 different days, respectively. The within-run and between-run
imprecision was determined so as to include all parts of the analysis,
i.e., sample collection, sample preparation, and quantification of
glucose with the GC-MS instrument. Assay specificity was demonstrated
by analyzing a sample of EDTA K3-treated blood that had
been gently mixed for a little more than 3 days. During this time, all
the glucose was consumed by glycolysis, so that any interfering
substances present should be disclosed in the chromatogram. The
imprecision in the two pipetting steps in the method was determined by
performing multiple weighings of the 50-µL microcaps filled with
water and by weighing the amount of water delivered by the 40–200-µL
adjustable micropipette.
Linearity.
We used a standard curve over the 1.0–20.0
mmol/L concentration range and checked the linearity by linear
regression fit.
HemoCue/ID GC-MS comparison.
The HemoCue blood glucose
analyzer is a bedside or near-patient analyzer with dedicated cuvettes;
glucose dehydrogenase, mutarotase, and diaphorase act on the sample in
connection with a color-generating reagent. The glucose is measured
after the erythrocytes have been lysed in the cuvette. The analytical
principle was described by Banauch et al. (12).
The recommended measurement range for the HemoCue blood glucose analyzer is 0–22.2 mmol/L. The HemoCue analyzer is calibrated by the manufacturer against an automated glucose dehydrogenase method and a Yellow Springs Instruments glucose analyzer (YSI Ltd., Hampshire, UK). Each individual analyzer is equipped with a control cuvette to assure the stability of the calibration. For an analysis to be acceptable, the result for the control cuvette may not deviate >± 0.3 mmol/L from the target value. The effect of hematocrit on the HemoCue blood glucose analyses has been studied by Wiener (13), who saw no appreciable effect from variations in the hematocrit.
In the comparison study, the HemoCue instrument with disposable cuvettes was carried to the patients, as were small plastic tubes containing 200 µL of [13C6]glucose, 1.0 mmol/L (used as sampling tubes for the ID GC-MS method). Capillary blood was collected into the HemoCue cuvette, which was inserted into the photometer to read the glucose concentration. Without delay after the cuvette sampling, capillary blood was also taken into a 50-µL microcup and transferred at once into the 200-µL solution of internal standard. The 50-µL microcup was washed five times with the [13C6]glucose solution, and the sample was thoroughly mixed. These tubes (containing the mixture of blood and internal standard solution) were brought to the laboratory and analyzed by ID GC-MS as described above.
For accuracy control in the comparison study, we measured a control sample (Seronorm) with every batch of blood samples analyzed in parallel by both the HemoCue method and the ID GC-MS method. The Seronorm was analyzed by both methods. The accuracy of the HemoCue analyzer was also checked for every sample batch by testing of the control cuvette.
|
Results and Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionReferences |
|---|
). We saw no decrease at all in glucose concentration for 240
min in blood samples so collected; a stable glucose concentration was
reached within a few minutes. The standard protocol we propose calls
for the samples to be left for 1–2 h after collection of blood to
ensure equilibrium between blood glucose and
[13C6]glucose before the addition of
methanol.
|
When performing isotope dilution analysis, one must make sure that equilibrium is reached between the internal standard and the substance to be measured. This is especially important for whole blood, which is not homogeneous and differs in glucose concentration between plasma and the erythrocytes. In the method we present, problems with this concentration gradient are overcome by adding 50 µL of blood to 200 µL of an aqueous solution of the internal standard and mixing thoroughly. The erythrocytes are lysed in this solution and the intracellular glucose is released. We found that deproteinization could be performed 1–2 h after mixing the blood and the internal standard solution because full equilibrium between glucose and [13C6]glucose was achieved in <1 h.
In the glycolysis test, all of the blood samples had normal hematocrit values. Sidebottom et al. (14) showed that, in a blood sample with a hematocrit of 0.43, the glucose concentration decreased ~25% in 4 h; in a sample with a hematocrit of 0.75, the glucose concentration decreased ~50%. In this test, where we mixed 50 µL of whole blood with 200 µL of internal standard solution, no decrease in glucose concentration was apparent after 4 h. In view of the dramatic decrease of glucose concentration in nontreated blood samples (14), we conclude that the inhibition of the glycolysis in the method described is independent of the hematocrit of the blood.
Precision and accuracy.
Within-run imprecision was
checked by multiple analyses of glucose in whole blood, Seronorm,
Pathonorm L, and Pathonorm H (Table 2
). To check the between-run imprecision, we analyzed Seronorm as
a control in different series of blood samples on 36 different days
over 10 months. Calculations from these analyses showed that the
between-run CV was 1.44% (n = 36). Pathonorm L and Pathonorm H
were analyzed similarly, on 19 and 21 different days, respectively. The
method showed good precision in all materials and at the three
concentration ranges tested.
|
To check the contribution of the pipetting steps to the imprecision of the method, we weighed 20 empty microcaps and then reweighed them after filling each with 50 µL of water at 22 °C, calculating from the density of water the volumes delivered by the microcaps. The imprecision (CV) of this pipetting step was 0.44%. The exact volume of this step in the analysis procedure is not crucial because we used the same type of microcap to deliver the standard solutions. The imprecision (CV) of pipetting the internal standard was 0.11%, determined by weighing 20 times 200 µL of water at 22 °C as delivered by the Finnpipette used for the internal standard solutions.
The largest contribution to the total imprecision of the method was from sample delivery with the microcap. This sampling technique is nevertheless unavoidable if capillary blood is to be taken. In all concentration ranges (low, normal, and high), the between-run imprecision is low, with CVs of 1.79%, 1.44%, and 1.57%, respectively, and should be compared with the CV limit of 2.0% for a glucose Reference Method proposed in the German program for Quality Assurance in Medical Laboratories (16).
The accuracy of the method was assessed by analysis of SRM 909b Level I. The accuracy test was performed with standard solutions of glucose prepared from both Mallinckrodt and SRM 907a (SRM 907a was not available at the start of this study, hence the use of the Mallinckrodt glucose). The mean ± SD concentrations of glucose in SRM 909b Level I were 5.41 ± 0.093 mmol/L (n = 10) with Mallinckrodt standard solutions and 5.38 ± 0.047 mmol/L (n = 10) with SRM 917a standard solutions. The results are well within the concentration range certified by NIST for glucose in reconstituted SRM 909b Level I (i.e., 5.40 ± 0.28 mmol/L).
Linearity.
The mass spectra of the aldononitrile
pentaacetate derivatives of glucose and
[13C6]glucose are shown in Fig. 1
. The ratio of m/z (314.0 + 242.1) for glucose to
m/z (319.0 + 246.1) for
[13C6]glucose was used for linear regression
fit. To study the accuracy of the HemoCue blood glucose analyzer, we
used six standard solutions with glucose concentrations of 1.0–20.0
mmol/L, a range within which the standard curve was found to be linear.
The linear regression data for a representative standard curve had a
slope of 0.228 and an intercept of 0.026, with a correlation
coefficient of 1.000 and dispersion around the regression line
(Sy|x) of 0.0178 mmol/L.
|
Specificity.
The specificity of the method was checked
by analyzing blood from an EDTA K3-containing tube that had
been gently mixed for a little more than 3 days. The blood was analyzed
by the routine protocol described above, and a peak eluting at the same
place as glucose, i.e., at 11.41 min, was hardly detectable (Fig. 2
B). When the analysis was carried out without the addition of
[13C6]glucose, no peak at all could be seen
at the elution position of glucose (Fig. 2A
). The small glucose peak
seen in Fig. 2B
, which probably originates from trace amounts of
glucose in the [13C6]glucose batch used, was
estimated as 0.2% of the amount of
[13C6]glucose added. The peak corresponds to
a glucose concentration of <0.01 mmol/L and its contribution to
quantification of glucose in blood can be neglected. We conclude that
the method is highly specific for analyzing glucose in whole blood.
|
Method comparison.
In the accuracy study of the HemoCue
blood glucose analyzer, capillary blood was taken from 67 patients with
diabetes mellitus and 62 nondiabetic individuals. Samples were taken on
two different occasions from three of the diabetic patients and from
eight of the nondiabetic individuals. Single measurements were
performed on all samples by both methods. Collecting single capillary
blood samples for assay by both methods minimized any eventual
alterations of blood glucose concentration during sampling. The two
methods performed well enough to justify single measurements with the
ID GC-MS and HemoCue method during the method comparison.
In all, 140 blood samples were measured in the comparison study: an average of 4–5 samples per day on 32 days. Moreover, in each batch of samples, we assayed Seronorm (lot no. 182) by both methods to check the accuracy and precision of the methods. At the same time, we also used the control cuvette to control the precision of the HemoCue instrument. The mean glucose concentration of Seronorm measured with HemoCue was 6.33 mmol/L (CV = 2.0%, n = 32). The HemoCue blood glucose analyzer is dedicated for measurements of glucose in whole blood and does not give correct concentration values in serum. The stability of the manufacturer's calibration of the instrument was assured by repeatedly measuring the control cuvette. The result for this cuvette was 15.3 mmol/L at every measurement (n = 32) during the comparison study—the same as the target value for the instrument used.
Measurements of Seronorm by the ID GC-MS method during the comparison
study are shown in Table 2
. The mean concentration of glucose measured
in Seronorm was 5.00 mmol/L (CV = 1.44%, n = 36).
For measurements of the 140 blood samples, the results from the HemoCue
instrument were compared with those of the ID GC-MS method by a bias
plot (Fig. 3
) and linear regression analysis. The bias plot showed very good
accuracy for the HemoCue instrument, especially for concentrations <10
mmol/L, where the mean HemoCue deviation from the ID GC-MS results was
0.04 mmol/L (or 0.9%), the maximal deviation being 0.95 mmol/L
(9.8%). In the glucose range of 10–22.2 mmol/L, a positive bias was
seen: mean 0.55 mmol/L (3.6%). The maximal deviation in this
concentration range was 2.27 mmol/L (13.0%). The mean deviation over
the whole concentration range was 0.24 mmol/L (2.0%). Note that the
results from all samples taken in the method comparison are included in
the bias plot and in the regression analysis; no outliers have been
excluded. The only samples not included were those from six diabetic
patients, who gave a result of HHH on the HemoCue analyzer, indicating
that the glucose concentration exceeded the recommended measuring limit
of the instrument, 22.2 mmol/L.
|
Regression analysis between the HemoCue instrument and the ID GC-MS method showed a slope of 1.051 (95% confidence interval: 1.031–1.071), an intercept of -0.222 (95% confidence interval: -0.016 to -0.428), and r = 0.994. Dispersion around the regression line (Sy|x) was 0.59 mmol/L. The HemoCue instrument performed well, especially in the concentration range <10 mmol/L. The positive bias found at concentrations >10 mmol/L have small clinical significance. In any case, the accuracy of the HemoCue glucose analyzer is well within the acceptable deviation limit set by an American Diabetes Association consensus conference (17). The recommendation from this conference is that assaying with glucose monitoring instruments should give values that deviate <15% from those of a Reference Method. Of all 140 results from the HemoCue instrument, 69% were within 5% of the ID GC-MS method, 91% were within 10%, and all results were within 15% (actually, within 14.3%) of the ID GC-MS method. The need for accuracy of blood glucose monitoring devices is obvious from the comparison study by Havlin et al. (18), who compared the results of six different glucose meters with those of a Beckman glucose analyzer (comparison method) for a total of 496 blood samples. Of the results from all six instruments, 23% to 59% deviated by >15% from the comparison method.
In conclusion, the method described is to our knowledge the first ID GC-MS technique adopted for analysis of glucose in capillary whole blood. We used it to assess the accuracy of a bedside whole-blood glucose analyzer. To be able to use the ID GC-MS method as a Reference Method for bedside glucose determination, we adjusted the analysis procedure to better fit the clinical circumstances. The ID GC-MS protocol is easy to follow, and both precision and accuracy are good enough to regard the method as a Reference Method for whole-blood glucose determinations.
|
Acknowledgments |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResults and DiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  A. Hussain, S. Vaaler, M. A. Sayeed, H. Mahtab, S. M. K. Ali, and A. K. A. Khan Type 2 diabetes and impaired fasting blood glucose in rural Bangladesh: a population-based study Eur J Public Health, June 1, 2007; 17(3): 291 - 296. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. W. El-Kadi, R. L. Baldwin, N. E. Sunny, S. L. Owens, and B. J. Bequette Intestinal Protein Supply Alters Amino Acid, but Not Glucose, Metabolism by the Sheep Gastrointestinal Tract J. Nutr., May 1, 2006; 136(5): 1261 - 1269. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. O. Silva Microwave-Assisted Derivatization of Glucose and Galactose for Gas Chromatographic Determination in Human Plasma Clin. Chem., February 1, 2006; 52(2): 334 - 335. [Full Text] [PDF]
|
|
|
|
 |
|
 A. D M Stork, H. Kemperman, D W. Erkelens, and T. F Veneman Comparison of the accuracy of the hemocue glucose analyzer with the Yellow Springs Instrument glucose oxidase analyzer, particularly in hypoglycemia Eur. J. Endocrinol., August 1, 2005; 153(2): 275 - 281. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. W. Vesper, E. Archibold, K. H. Porter, and G. L. Myers Assessment of a Reference Procedure to Collect and Analyze Glucose in Capillary Whole Blood Clin. Chem., May 1, 2005; 51(5): 901 - 903. [Full Text] [PDF]
|
|
|
|
|
|
C. Petibois, V. Rigalleau, A.-M. Melin, A. Perromat, G. Cazorla, H. Gin, and G. Deleris Determination of Glucose in Dried Serum Samples by Fourier-Transform Infrared Spectroscopy Clin. Chem., September 1, 1999; 45(9): 1530 - 1535. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
