| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Department of Clinical Biochemistry, The Royal London Hospital, Whitechapel, London E1 1BB, UK.
2
Biochemistry Laboratory, Henri Mondor Hospital, 94010
Creteil, France.
3
Central Laboratory, Ziekenhuis De Weezenlande, 8011 JW
Zwolle, The Netherlands.
4
Department of Hematology, Hospital de la Valle Hebron,
08035 Barcelona, Spain.
5
Anti Diabetic Centre, S. Michele Hospital, Via Peretti,
(09100) Cagliari, Italy.
a Author for correspondence. Fax +44 171 377 7777.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Key Words: indexing terms: diabetes mellitus • hemoglobin variants • HPLC • immunoassay.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Early methods for the measurement of GHb were based on charge separation, either involving minicolumn cation-exchange chromatography (1) or agar gel electrophoresis (3). Both methods had a number of interferences, most notably the adverse affect of abnormal Hbs. Since these early methods, techniques have been developed that overcome these problems. HPLC techniques have proved to be very successful with ion-exchange separation methods (4); additionally, methods based on affinity separation (5) and immunoassay (6) have also been described and are widely used in the clinical laboratory.
Increasing numbers of requests are being received by laboratories for GHb estimation, while these laboratories are being faced with increased staffing pressures. Therefore there is a need for highly automated GHb analyzers capable of a high sample throughput. The aim of this study was to fully evaluate a new Hb A1c analyzer, the Menarini–Arkray HA 8140, and compare results obtained with those from established methods in clinical use. The HA 8140 analyzer replaces the HA 8121 analyzer; the HA 8121 is also based on ion-exchange separation, but is incapable of accurately measuring Hb A1c in the presence of abnormal Hbs. The newer HA 8140 also incorporates a bar code reader and cap piercing capability.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The HA 8140 is at present unique among HPLC analyzers in that all operations are fully automated. The automatic feed to the analyzer holds up to a maximum of 100 samples at any one time. Bar coded samples are placed in racks that will hold up to 10 samples; additionally, the analyzer may be continuously fed with samples. After reading the bar code, the analyzer samples directly from the primary sample tube; if this tube incorporates a rubber stopper, the instrument will pierce the cap, thereby eliminating the need to remove it. The HA 8140 samples 3 µL of whole blood from the bottom of the sample tube; the blood is automatically hemolyzed with a tetrapolyphosphate (TPP) buffer at pH 6.0, and the hemolyzed sample is held in a thermal-jacketed loop for 2 min at 48 °C. Incubation in TPP has been shown to effectively remove the labile Schiff base (7). After this time the loop is switched and the hemolysate injected onto a methacrylic acid and methacrylate copolymer column (Micronex A1c-HSII column unit; Sekisui Chemical Co., Tokyo, Japan) thermostatically controlled at 40 °C. Separation is achieved with discrete addition of three phosphate buffers (pH 4.8) containing <6% inorganic phosphate; these are supplied by the manufacturer. Buffer is added in the order A, C, B; then A again. This will regenerate the column, switching is under computer control, and results are available in a further 2 min. A true analytical throughput of 15 samples per hour is achieved.
Results are presented as %Hb A1 and %Hb A1c; these are calculated from the peak areas of the different Hb fractions as a percentage of the total Hb. Extrapolation of the peaks to baseline is under the control of the analyzer's computer; the calculation of the percent Hb A1c may include a minor peak present as a shoulder eluting before the main Hb A1c fraction. This shoulder increases/decreases with the Hb A1c and can be seen with several HPLC methods that have a longer elution time.
If a small or a capillary sample is received, the HA 8140 analyzer will automatically switch to analyzing hemolysates that are manually diluted 1:150 in the buffer provided. This is achieved by leaving the first four positions in the rack empty; the analyzer will automatically treat the sample in position five as a hemolysate.
The analyzer is not calibrated; results are based on the natural color
of the Hb in the fractions measured at a wavelength of 415 nm (blanking
wavelength 500 nm) as they elute from the column. The analyzer does
have the capability of automatically "correcting" results after
standardization with a reference material when this becomes available.
A typical chromatogram is shown in Fig. 1
.
|
Variant Hb A1c
analyzer.
After initial pretreatment, samples were loaded onto the Variant
automated Hb A1c analyzer (manufactured by Toya-Soda,
Tokyo, Japan), which is distributed by Bio-Rad Labs., Hemel Hempstead,
UK. All reagents used were supplied by Bio-Rad, and analyses performed
according to the manufacturer's instructions.
"In-house" cation-exchange chromatography.
Samples
were prepared by adding 45 µL of whole blood to 2.5 mL of hemolyzing
solution (acid potassium phthalate 25 mmol/L; KCN 8 mmol/L; Triton
X-100 10 mL/L, pH 5.5), mixed vigorously, sonicated (2 min), and then
incubated for 45 min at 25 °C to eliminate the labile fraction.
Chromatography was performed with a HPLC cation-exchange system with a
PolyCat A (diameter 4 x 50 mm) column. A two-buffer gradient was
used for separation. The starting buffer consisted of 0.04 mol/L Bis
Tris, 0.004 mol/L KCN, and 5 mL/L Triton X-100 adjusted to pH 6.65; the
second buffer was the same except for addition of 0.2 mol/L NaCl. This
was a modification of a method previously described
(8)(9). The precision (CV) of this method was
2.1% for Hb A1c of 10.8%, and the range of results found
in a nondiabetic population was 3.7–4.4% Hb A1c.
immunoturbidimetry
All reagents used in this latex particle immunoassay were supplied
by Boehringer Mannheim (Lewes, UK) and performed according to the
manufacturer's instructions. The method is calibrated with calibrators
whose values have been assigned by HPLC; the method was performed on an
Hitachi (Tokyo, Japan) 717 analyzer, which gave an imprecision of 4.6%
at 9.8% Hb A1c.
affinity chromatography
In addition to measuring Hb A1c, we measured
total GHb by affinity chromatography (Pierce and Warriner, Chester,
UK). This method has been previously described (10), and a
precision of 1.5% at a concentration of 18.7% GHb was reported.
variant hb identification
The different variant Hbs were identified either by isoelectric
focusing on thin polyacrylamide gel (pH 6–9) as previously described
(11), or by HPLC with cation-exchange chromatography, and
structurally studied by specific methods of protein biochemistry.
samples
Samples were collected into bottles containing EDTA; additionally,
samples were collected into lithium heparin and fluoride oxalate to
investigate the effect of different anticoagulants. Samples were
collected by venesection from healthy volunteers or from patients
having blood samples collected for medical reasons; permission was
sought before venesection and samples were analyzed blind in accordance
with local ethical recommendations.
effect of the labile schiff base
Two EDTA whole-blood samples with %Hb A1c values
within the nondiabetic and diabetic ranges were centrifuged, and the
plasmas removed and set aside. The red cells from both samples were
divided into three aliquots and suspended in phosphate-buffered saline
(PBS), pH 7.0, and in PBS containing 20 mmol/L glucose and 50 mmol/L
glucose. The suspensions were incubated at 37 °C for 4 h. After
this, the red cells were again separated by centrifugation and
resuspended in their own plasma; Hb A1c analysis was
performed in duplicate on these samples within 15 min of resuspension.
effect of hemoglobin f
Ten samples were analyzed for %Hb A1c. Two
500-µL aliquots of these samples were taken; one aliquot was
supplemented with 100 µL of cord blood, the other with 150 µL of
cord blood. These samples were again analyzed for %Hb A1c.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
.
|
Between-batch.
The above three samples were analyzed 20
times over 2 weeks; samples were stored at 4 °C during this period.
Only one batch of mobile phase was used during this study, but the
analyzer was switched off/on, and purged between each analysis.
Additionally, control samples were analyzed twice daily during this
study. The results obtained are shown in Table 1
.
linearity
Linearity was investigated with a sample with a low Hb
A1c (2.4%) and one with a high Hb A1c
(14.2%). Hemolysates (1:150) were made of the samples in hemolyzing
buffer. These two hemolysates were then mixed with each other in
different proportions to produce a range of Hb A1c results;
expected results for the dilutions were calculated from the results
obtained on the high and low samples. The regression between the
measured %Hb A1c (y) and the calculated value
(x) was y = 1.03x -
0.12; r = 1.0.
method comparison
Samples collected from nondiabetic and diabetic subjects and
analyzed for Hb A1c on the HA 8140 were additionally
analyzed for Hb A1c with the Bio-Rad Variant HPLC analyzer,
a published "in-house" HPLC method, and a latex particle
immunoassay, and total GHb was measured by affinity chromatography. The
regressions obtained are shown in Fig. 2
. The methods were also compared by using Altman and Bland
plots; these produced mean bias (lower, upper 95% confidence limits)
of: Bio-Rad Variant minus HA 8140 = 0.99 (0.23, 1.74), Tinaquant
minus HA 8140 = 0.14 (-0.71, 0.98). Altman and Bland comparison
of the HA 8140 and the "in-house" method and affinity
chromatography were performed after log transformation of the data.
This was done because the bias was not constant (Fig. 2b
and 2d
); the
bias increased with increasing results: "in-house" HPLC minus HA
8140 = 0.91 (0.82, 1.01), affinity chromatography minus HA
8140 = 1.13 (0.90, 1.41).
|
anticoagulants
Ten subjects, 6 diabetics and 4 nondiabetics, had samples
collected into different anticoagulants, and each sample was analyzed
for Hb A1c. The mean Hb A1c result
obtained with EDTA was 6.88%, whereas the mean result (mean difference
from EDTA) for lithium heparin was 6.94% (0.06%) and for fluoride
oxalate was 6.97% (0.08%). The results obtained on the different
anticoagulants did not differ significantly (P <0.05) from
each other.
reference range
Blood was collected from 399 subjects into an EDTA tube and a
fluoride oxalate tube. Hb A1c was measured on the EDTA
tube. Subjects all denied being diabetic, and random blood glucose
measured on the fluoride oxalate sample was <6.5 mmol/L in all cases.
Results displayed a normal distribution, which produced a nondiabetic
reference range of 3.7–5.1% Hb A1c.
abnormal hbs
Hb F.
The Hb A1c results before and after
addition of cord blood to 10 samples containing normal adult Hb are
given in Table 2
. Addition of the Hb F did not affect the Hb A1c
result obtained, the increased Hb F peak being seen well in advance of
the Hb A1c peak (Fig. 3
a).
|
|
Variant Hbs.
The HA 8140 system incorporates separation
conditions that will detect many Hb variants; a number of the variants
were studied in this evaluation. With the common Hb variants such as Hb
S, Hb C, and Hb D Punjab, the abnormal Hb fraction eluted as a single
fraction after the Hb A0 peak; the variant was identified
by the analyzer by producing a "variant Hb" message, and the
abnormal fraction indicated by a hatched peak (Fig. 3b
and 3c
).
Analysis of samples that did not contain Hb A (e.g., Hb SS and Hb SC)
has shown that the glycated variant Hb peak is well separated from the
Hb A1c peak, and will not be included in the calculated
%Hb A1c. The %Hb A1c is automatically
calculated as a ratio of the Hb A1c peak compared with
total Hb minus the variant Hb fraction. When no Hb A is present in the
sample, as for example in the variant Hb SC, an "abnormal
separation" message is printed and no result displayed (Fig. 3e
).
One common Hb variant (Hb E) did not always separate. In this case a
shoulder that appears behind the Hb A0 peak can be
easily identified (Fig. 3d
); no error message is printed. The major
peak includes Hb A0 and Hb E0, and the glycated
Hb E fraction elutes separately from the Hb A1c fraction,
thereby producing an artificially low %Hb A1c result. The
presence of the Hb E variant is easily recognized from the chromatogram
produced by the analyzer; the chromatographs should always be examined
to eliminate a wrong result being generated.
Several uncommon Hb variants were also investigated; Table 3
shows whether the HA 8140 detects the abnormal Hb. Separation
of the major fractions of Hb A and variant Hb will allow the correct
calculation of %Hb A1c; as in all cases the glycated
variant fraction elutes separately from the Hb A1c peak. If
no Hb A is present in the sample, an "abnormal separation" message
is printed and no result displayed.
|
effect of the labile schiff base
The results obtained by incubating in 20 mmol/L and 50 mmol/L
glucose are shown in Table 4
. There was a trend to increased results in the
glucose-incubated samples, but the largest increase was only 0.3% Hb
A1c in the nondiabetic sample; this increase was well
within the precision of the method.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The Hb A1c analyzer gave good within- and between-batch precision. Both comply with the within- and between-batch imprecision of 5% suggested by the National Diabetes Data Group (12), and a maximum CV of 5% with an optimum CV of ~2% published by Larsen et al. (13) as desirable for the assay.
The analyzer was found to produce good linearity through the operating range, and when compared with several GHb methods widely used in clinical laboratories showed a high degree of correlation. Results obtained on the HA 8140 analyzer were found to be negatively biased when compared with the Bio-Rad Variant analyzer, the Variant analyzer displaying results 1.0% Hb A1c higher than the HA 8140, this bias being constant throughout the range as shown by a slope of 0.95. The constant bias found suggests that this difference may be due to the calibration factor used by the Variant analyzer; the HA 8140 analyzer does not use standardization (although it has the ability to be calibrated). Good agreement (r = 0.99) was also found with an "in-house" HPLC method, the conditions of which had been optimized for Hb A1c separation. In this case the HA 8140 results were higher than those obtained with the optimized HPLC, the mean bias (after log transformation) being 0.91; this may reflect a "cleaner" separation obtained with a slower optimized system. The best regression was found between the HA 8140 and the latex particle immunoassay method (y = 0.99x - 0.06), even though these two methods involve different technologies. This good agreement is not totally surprising, as the immunoassay method is calibrated by using HPLC. When Hb A1c results obtained with the HA 8140 are compared with total GHb results measured by affinity chromatography, a difference in the results obtained is apparent; this reflects the fact that different glycated fractions are being measured. The difference in the results obtained with the two methods increases as the measured result increases. This relation has previously been identified (10), and may reflect increased glycation at sites on the Hb molecule other than the N-terminal valine; this occurs at high GHb concentrations. The differences in the slopes and intercepts of the comparison methods highlight the differences in calibration or in the type of fraction measured (i.e., total GHb or Hb A1c), supporting the need for standardization of methods. In the US, there has been a national attempt to overcome these result differences; this has been achieved by adopting a standardization protocol written by the AACC Subcommittee on Glycohemoglobin. The problem of result variability is currently a topic under investigation by the IFCC working party on GHb standardization; this group is attempting to produce a primary calibration material and a reference method for the measurement of Hb A1c.
The reference range for results in nondiabetics in this study was 3.7–5.1% Hb A1c, lower than that quoted for the Bio-Rad Diamat analyzer (4.3–6.1% Hb A1c) used in the DCCT (2). This difference is confirmed by the 1.0% bias found when the HA 8140 was compared with the Bio-Rad Variant analyzer. Additionally, the National Glycohemoglobin Standardization Program has recently certified one of the HA 8140 analyzers used in this study, and the conversion factor of yDCCT = 0.96xHA 8140 + 1.30 given is similar to that found in this study for the regression between the HA 8140 and the Bio-Rad Variant.
The effectiveness of the HA 8140 analyzer's ability to remove the labile fraction by incubating the sample in TPP (7) was investigated by incubating red cells in PBS containing 20 mmol/L and 50 mmol/L glucose. Previous investigations of incubation conditions by one of the investigators (W.G.J., unpublished data) and others (14) have confirmed that red cells suspended for this time in these glucose concentrations reproducibly produce an increase in the labile fraction by ~2% and 8% measured GHb, respectively. The samples incubated in glucose in this study showed only a small increase (0.3%) in measured Hb A1c; this increase may actually be due to the formation of the stable ketoamine. This increase is within the precision of the method, and in clinical terms is negligible considering the expected increase in labile fraction that will have occurred after the glucose incubation.
The predecessor to the HA 8140 analyzer (the HA 8121) was adversely affected by abnormal Hbs; it was therefore important to identify if, as claimed, the HA 8140 has overcome this problem. In this study a number of samples containing common and uncommon Hb variants were analyzed for Hb A1c. The results obtained indicate that with the most commonly encountered abnormal Hbs, i.e., Hb S and Hb C, the HA 8140 can be confidently used to measure Hb A1c. Another commonly encountered Hb variant, Hb E, is not always separated completely from Hb A0. When separation is not achieved, this variant can easily be recognized by inspecting the chromatogram produced. Hb F, a variant that commonly interferes with GHb measurement, did not interfere with this analyzer, which can be used to measure Hb A1c even in the presence of very high concentrations of Hb F. Of course, the presence of Hb variants may cause a decreased red cell survival, thereby increasing the Hb turnover, which leads to a decreased exposure time of this protein to glucose, resulting in a decreased percentage being glycated (15).
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  L. Monnier, C. Colette, G. J. Dunseath, and D. R. Owens The Loss of Postprandial Glycemic Control Precedes Stepwise Deterioration of Fasting With Worsening Diabetes Diabetes Care, February 1, 2007; 30(2): 263 - 269. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Sakurabayashi, T. Watano, S. Yonehara, K. Ishimaru, K. Hirai, T. Komori, and M. Yagi New Enzymatic Assay for Glycohemoglobin Clin. Chem., February 1, 2003; 49(2): 269 - 274. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Luraschi, S. Brambilla, R. Mozzi, G. Cattozzo, and C. Franzini Monitoring Analytical Quality in Routine Glycohemoglobin Measurements Clin. Chem., September 1, 2002; 48(9): 1594 - 1597. [Full Text] [PDF]
|
|
|
|
 |
|
 T Lahousen, R E Roller, R W Lipp, and W J Schnedl Silent haemoglobin variants and determination of HbA1c with the HPLC Bio-Rad Variant II J. Clin. Pathol., September 1, 2002; 55(9): 699 - 703. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. L. Roberts, B. K. De, D. Brown, C. M. Hanbury, J. D. Hoyer, W. G. John, T. L. Lambert, R. B. Lundell, C. Rohlfing, and R. R. Little Effects of Hemoglobin C and S Traits on Eight Glycohemoglobin Methods Clin. Chem., February 1, 2002; 48(2): 383 - 385. [Full Text] [PDF]
|
|
|
|
 |
|
 R J Jarrett, E. Lamb, K.-T. Khaw, N. Wareham, and V. Vlassov Glycated haemoglobin, diabetes, and mortality in men BMJ, April 21, 2001; 322(7292): 996 - 996. [Full Text]
|
|
|
|
|
|
L. Bry, P. C. Chen, and D. B. Sacks Effects of Hemoglobin Variants and Chemically Modified Derivatives on Assays for Glycohemoglobin Clin. Chem., February 1, 2001; 47(2): 153 - 163. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Gibb, A. Parnham, M. Fonfrede, and F. Lecock Multicenter Evaluation of Tosoh Glycohemoglobin Analyzer Clin. Chem., October 1, 1999; 45(10): 1833 - 1841. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. A. Nisbet, F. G. Boa, and W. G. John Correction: Evaluation of the Menarini-Arkray HA 8140 Hemoglobin A1c Analyzer • Dr. John responds: Clin. Chem., January 1, 1999; 45(1): 154 - 154. [Full Text] [PDF]
|
|
|
|
|
|
S. Blincko and R. Edwards Non-separation assay for glycohemoglobin Clin. Chem., June 1, 1998; 44(6): 1302 - 1308. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
