Capillary electrophoresis system for hemoglobin A determinations evaluated 1c

Clinical Chemistry 43:4
644 – 648 (1997)
Automation and
Analytical Techniques
Capillary electrophoresis system for hemoglobin
A1c determinations evaluated
Cees J.A. Doelman,1* Carla W.M. Siebelder,1 Wim A. Nijhof,1 Cas W. Weykamp,1
Jacques Janssens,2 and Theo J. Penders1
chromatography) for measuring glycohemoglobin, or immunological techniques [3–5]. All methods have their
own advantages and well-known limitations [3–5], e.g.,
interferences from fetal hemoglobin (Hb F), carbamylated
and acetylated hemoglobins, labile Hb A1c fractions, or
hemoglobin variants.
Capillary electrophoresis (CE) is a modern analytical
technique that separates molecules on the basis of their
charge and their hydrodynamic volume [6]. The CE
method described here for separating hemoglobin derivatives and hemoglobin variants makes use of a dynamic
coating technique that allows rapid separation (#4 min)
of hemoglobin variants and derivatives at pH 4.5. This
proprietary coating principle was developed by Analis
(Namur, Belgium) [7]. Here we describe our evaluation of
this new method, especially as developed for Hb A1c
measurement. We examined some potential interferents
and compared the results with those by cation-exchange
Hb A1c is the analyte of choice for monitoring metabolic
control in patients with diabetes mellitus. Here we
present a new analytical technique for measuring Hb
A1c, capillary electrophoresis. The Hb A1c determination
is not influenced by the labile Hb A1c fraction or by
carbamylated or acetylated hemoglobin derivatives.
Also, hemoglobin variants (Hb F, Hb S, and Hb C) do
not interfere. This new application of capillary electrophoresis seems to be a valuable analytical tool for
measuring Hb A1c in the clinical laboratory.
diabetes mellitus
hemoglobin variants
Diabetes mellitus is one of the most prevalent chronic
diseases of humans. Not only are affected patients predisposed to late complications (e.g., retinopathy, neuropathy,
and nephropathy), but also they are at notable risk for
cardiovascular disease. All of these complications are
hypothesized to be the effect of chronic glycosylation of
proteins and cellular structures [1].
Measurement of Hb A1c (the reaction product of glucose and the N-terminal valine of the b-chain of hemoglobin) has been used to monitor the metabolic control of
patients with diabetes mellitus. The amount of Hb A1c
present is related to the risk of long-term diabetic complications, as clearly shown in the Diabetes Control and
Complications Trial (DCCT) [2]. Accordingly, Hb A1c has
become a generally accepted marker for follow-up of
diabetic therapy. Several analytical methods currently
available measure either Hb A1c or glycohemoglobins
(;60% of which is Hb A1c). These methods are based on
either differences in electrical charge (HPLC, electrophoresis) for measuring Hb A1c, specific binding (affinity
Materials and Methods
Blood samples (anticoagulated with EDTA) were obtained from diabetic and nondiabetic patients who entered our outpatient clinic. All procedures involving patients were in accordance with the Helsinki Declaration of
1975, as revised in 1983.
Blood samples with a high Hb F content were derived
from cord blood. Carbamylated and acetylated hemoglobin were synthesized in vitro as described by Weykamp et
al. [3]. Blood with a high percentage of labile Hb A1c
fraction (Schiff base) was made by incubating washed
erythrocytes in a 100 mmol/L glucose solution for 24 h.
To prepare the samples for assay, we used 20 mL of
EDTA-anticoagulated blood mixed with 100 mL of saponin-containing “hemolyzer” reagent (provided by Analis).
Department of Clinical Chemistry and Hematology, Queen Beatrix
Hospital, Beatrixpark 1, 7101 BN Winterswijk, The Netherlands.
Analis S.A., 14 Rue Dewez, B 5000 Namur, Belgium.
*Author for correspondence. Fax 131543524265.
Received May 29, 1996; revised October 29, 1996; accepted November 12,
Capillary zone electrophoresis was performed on a Beckman P/ACE System 5000 (Beckman, Brea, CA) with a 25
Clinical Chemistry 43, No. 4, 1997
mm (i.d.) 3 24 cm fused-silica capillary at 25 °C. Proprietary patented reagents (malic acid buffers, pH 4.5) [7]
were obtained from Analis. Before sample injection, the
capillary was first rinsed with initiator solution (containing a polycation, albumin, pH 4.5) for 0.3 min under 13.8
kPa (20 psi) pressure, followed by 1.00 min of prerinsing
with buffer solution containing a polyanion (chondroitin
sulfate, pH 4.5), at the same pressure. Sample was injected
for 8 s at 5 kV; this was followed by a 10-s injection with
buffer solution at 3.5 kPa (0.5 psi) to rinse the outside of
the capillary. The capillary was then transferred to another vial containing buffer solution, in which the electrophoresis was performed. Negatively charged molecules
(chondroitin sulfate, pH 4.5) in the buffer solution bind to
hemoglobin. Electrophoresis was performed with a constant current of 52 mA for 4 min with the negative
electrode at the detector site. Detection was executed with
a UV/VIS absorbance detector at 415 nm. Peak integration
for peak area measurement was performed by a Beckman
System Gold chromatography data system (vers. 8.10);
peak area percentages corrected for velocity were used.
After electrophoresis, the capillary was rinsed with 1
mol/L NaOH for 2 min at 13.8 kPa.
The relative apparent mobility of each peak was calculated according to the method described by Harris [8].
Because of a lower refractive index of the hemolysate
sample, the “hemolyzer” peak was used as an internal
Within-run variability was determined by analyzing 20
times three different patients’ samples (containing low,
medium, and high concentrations of Hb A1c) in one run.
Between-run variability was determined by analyzing the
same three samples once a day on 20 working days
(stored samples).
The comparison method, cation-exchange HPLC with
a Bio-Rex 70 column (Bio-Rad, Veenendaal, The Netherlands), has been described elsewhere [9]. In brief, 4 mL of
hemolysate (packed cells diluted 1:3 in distilled water)
was injected onto the Bio-Rex 70 HPLC column, which
was operated at 25 °C with a flow rate of 1.5 mL/min. We
eluted the sample for 5 min with buffer A (8.0 mmol/L
potassium cyanide dissolved in a 113 mmol/L sodium
phosphate buffer, pH 6.77) and then for 15 min with
buffer B (564 mmol/L sodium phosphate buffer, pH 6.42).
Detection was performed by measuring absorbance at 410
nm. Between each run the column was equilibrated for 15
min with buffer A.
Linear regression analysis of the CE and HPLC results
was followed by an outlier detection procedure described
previously [10].
The electrophoresis pattern of hemoglobin from a nondiabetic healthy volunteer is shown in Fig. 1. The hemolyzer, Hb A1c, and Hb A0 peaks are eluted at 2.63, 3.01,
and 3.29 min, respectively. Apparent mobilities (relative
Fig. 1. Hemoglobin CE pattern from a nondiabetic healthy volunteer
showing complete separation between Hb A1c and Hb A0.
The hemolyzer peak is eluted at 2.63 min, the Hb A1c peak at 3.01 min, and the
Hb A0 peak at 3.29 min.
to the internal standard, i.e., the hemolyzing reagent) are
shown in Table 1.
Reproducibility. Within-run variabilities were determined
by assaying three different blood samples. Low (4.3%),
medium (7.0%), and high (10.5%) Hb A1c samples gave
within-assay CVs of 1.7%, 2.9%, and 1.4%, respectively.
The between-run variabilities for these samples were
3.7%, 3.3%, and 1.9%, respectively. However, the aging of
the blood samples over the 20 working days (1 month,
total) of the between-assay reproducibility study resulted
in an extra peak (relative apparent mobility 1.18) between
the peaks for Hb A1c and Hb A0.
Interferences. The results for in vivo carbamylated hemoglobin are shown in Fig. 2. No interferences were observed from carbamylated or acetylated hemoglobin or
from the labile Hb A1c fraction, whose relative apparent
mobilities are shown in Table 1. Investigation of hemoglobin variants S (Fig. 3, top) and F and C (Fig. 3, bottom)
also showed no apparent interference. Hb S and Hb C
peaks corresponding to Hb S1c and Hb C1c were visible at
3.07 min and between 3.29 and 3.74 min, respectively.
Table 1. Relative apparent mobilities of hemoglobin
variants and derivatives.
Relative mobilitya
Hb F
Labile Hb A1c
Hb A1c
Aging peak
Hb A0
Hb S
Hb C
Relative to the mobility of the hemolyzer peak.
No. of
Doelman et al.: Capillary electrophoresis determination of Hb A1c
Fig. 2. Hemoglobin CE pattern of blood containing in vivo carbamylated
hemoglobin; the carbamylated hemoglobin (eluted at 2.88 min) is well
separated from Hb A1c (eluted at 3.01 min).
Blood was obtained from an insulin-dependent diabetic patient having a serum
urea concentration of 54 mmol/L.
None of these potentially interfering substances, including the sample-aging peak mentioned above, influenced
the CE assay of Hb A1c.
Results of the comparison between CE and Bio-Rex 70
HPLC gave a good linear correlation (Fig. 4): CE Hb A1c 5
Fig. 4. Correlation between the Bio-Rex 70 HPLC results and those by
CE (n 5 100).
The two samples that have a much larger deviation from the regression line than
the others both contained carbamylated hemoglobins, which seem to interfere in
the Bio-Rex 70 method.
21.41 1 1.02 HPLC Hb A1c (r 5 0.98). No outliers were
Fig. 3. Electrophoresis patterns of heterozygous Hb S (upper panel)
and Hb C and Hb F (lower panel).
(Top) Hb S is eluted at 3.47 min, Hb A1c at 2.99 min in this run. The peak at 3.07
min might be Hb S1c. (Bottom) Hb F elutes at 2.74 min and Hb C at 3.74 min.
The Hb A0 peak is shown at 3.29 min.
The DCCT clearly showed that the risk of late complications of diabetes mellitus is related to the percentage of
Hb A1c in a patient’s blood. The Bio-Rex 70 HPLC method,
described by Goldstein et al. [9] and having been shown
to be very accurate, precise, and useful for calibrating
different methods [11], was used to calibrate all of the Hb
A1c methods used in that study. The therapy goals of
diabetic therapy (i.e., Hb A1c values as low as reasonably
achievable) defined in the DCCT study are therefore
related to the Bio-Rex 70 HPLC results [2]. The AACC
subcommittee on standardization of Hb A1c analyses has
chosen to use this method as an anchor and recommends
that routine assays be calibrated in terms of the Bio-Rex 70
HPLC method [12].
The CE separation technique is rapid, uses low
amounts of reagents, and is easily automated. Although
separation of hemoglobin variants and derivatives in
uncoated fused-silica capillaries is limited by adsorbance
of proteins to the capillary wall and by variable rates of
electroosmotic flow, some methods for separating hemoglobin variants this way have been described [13, 14].
Those investigators tried to overcome the problems by
performing electrophoresis at a relatively high pH, which
induces a strong and constant electroosmotic flow; this
reduced the resolution, however, so long capillaries had
to be used, which increased the analysis time [13, 14].
Others have shown that capillary isoelectric focusing with
coated capillaries could also be used to separate hemoglobin variants [15–18], but only two of these publications
demonstrated the possibility of Hb A1c determination
Clinical Chemistry 43, No. 4, 1997
(they did not report complete separation of Hb A1c from
Hb A0) [17, 18].
The CE method used here is performed at an acidic pH
(4.5) and is based on an ion-pairing effect between hemoglobin and a negatively charged molecule in the running
buffer solution. The equilibrium of this ion pairing depends on the charges carried by the hemoglobin molecule
and on the accessibility of these charges. At acidic pH, the
amino group of hemoglobin is more positively charged
and more accessible than is the amino group of glycohemoglobin. Thus, the glycohemoglobin is eluted first because it is less strongly attached to the negatively charged
At the working pH of 4.5, the electroosmotic flow is
low, unstable, and highly variable from capillary to capillary. We overcame this difficulty by applying a dynamic
coating to the capillary. This coating is made in two steps.
The capillary is rinsed with buffer containing a polycation, which binds to the negatively charged silica surface
of the capillary; this approach can diminish or even
reverse the electroosmotic flow, depending on the nature
of the polycation and its concentration. In the second step,
the capillary is rinsed with buffer containing a polyanion,
which adds a layer of negative charges over the polycation layer. As a result, the internal surface of the capillary
will present a controlled and reproducible high number of
charges from the polyanion, thereby controlling the electroosmotic flow. The hemoglobin complexed with the
negatively charged molecule is then electrophoresed over
the coated capillary. The double-layer coating is removed
after each run by a simple rinse with NaOH [7]
The principle of coating the capillary makes possible
rapid separation of these hemoglobin variants (within a
few minutes)—which is the new development in this
It is generally accepted that to be useful an assay for Hb
A1c should not be influenced by Hb F or by carbamylated
or acetylated hemoglobin derivatives; hemoglobin variants should be detected but should not interfere. Analysts
should also be aware that hemoglobinopathies may reduce erythrocyte life span, which will result in artifactually low Hb A1c values both in nondiabetic and diabetic
subjects [19]. Knowledge of the presence of hemoglobin
variants in a patient’s sample or of other factors that
reduce erythrocyte life span is therefore necessary. An Hb
A1c method that identifies samples containing hemoglobin variants would be useful.
In this study we evaluated these potentially interfering
substances. The fact that Hb F and carbamylated and
acetylated hemoglobins do not interfere in Hb A1c measurement (Fig. 2) seems to be the main advantage of the
CE assay. The most common hemoglobinopathies Hb S
and Hb C also don’t appear to interfere. Probably, the
N-terminal glycated products of Hb S and Hb C are also
separated from Hb A0 and Hb A1c. The question, of
course, is which percentage—that of Hb A1c vs Hb A0, or
of Hb S1c/Hb C1c vs Hb S0/Hb C0— corresponds best with
the metabolic control of the patient. This should be
investigated in a diabetic population with heterozygous
The CE results correlated very well with those by the
Bio-Rex 70 HPLC method (Fig. 4). Two samples visually
out of line but not statistical outliers contained carbamylated hemoglobins; we concluded, therefore, that Hb A1c
measurement in the Bio-Rex 70 HPLC system is influenced by carbamylated hemoglobin.
Also, the negative intercept at 1.4% in the correlation
plot indicates that other hemoglobin derivatives (e.g.,
carbamylated or acetylated forms) are coeluting with Hb
A1c in the cation-exchange HPLC system. A comparable
difference was observed by Turpeinen et al. [20], comparing the Diamat method (cation-exchange chromatography) and HPLC with PolyCAT A; they found that the
PolyCAT A values were 2–3% lower than the Diamat
values. (The Diamat assay is easily calibrated to the
Goldstein et al. Bio-Rex 70 HPLC method.) From these
results, one might conclude that the CE method described
here is also somewhat biased (by ;1%) in comparison
with the PolyCAT A HPLC. We speculate that the integration method of the chromatograms (valley-to-valley
for PolyCAT A) and electrophoresis patterns (forward
horizontal for CE) might account for this 1% difference.
For lack of a more satisfactory explanation, perhaps a
study should be performed on comparability of the integration techniques in the various Hb A1c assays. In any
event, this situation underlines the lack of a “golden”
reference assay and the need for Hb A1c standardization.
Somewhat disappointing was the high interassay variability. The sample aging produces a peak between Hb
A1c and Hb A0, which during 1 month progressively
influences the outcome of the Hb A1c measurement (data
not shown). Therefore, the interassay variability was
increased, especially at low Hb A1c values. This should
not present problems in routine analyses, however, because the procedure calls for Hb A1c to be measured
within 1 week after blood collection.
The question arises as to whether this CE assay will be
used in a routine setting for Hb A1c measurement in
clinical laboratories. Until now, the test has been used as
a reference test for other routinely used Hb A1c assays. By
the time multichannel CE systems are available and the
throughput of samples is substantially increased, the
assay might be more suited for routine use. However, CE
equipment is rather expensive, whereas other Hb A1c
assays, e.g., immunoassays, can be run on routine clinical
chemistry analyzers. Nonetheless, because of its fast and
complete separation of Hb A1c from Hb A0 and from
hemoglobin derivatives and variants, the CE method
might be useful for clinical laboratories. The cost effectiveness of CE vs the Bio-Rex 70 method is presented in
Table 2. The main disadvantage of the Bio-Rex 70 HPLC
method is the low throughput of samples (only 5 per day
vs 40 per day by CE). This makes running the CE method,
Doelman et al.: Capillary electrophoresis determination of Hb A1c
Table 2. Cost-effectiveness of CE and Bio-Rex 70 HPLC
assays of Hb A1c.
Bio-Rex 70 HPLC
Equipment, $ per year
Maintenance, $ per year
Sample throughput
Tests per day
Tests per year
Technician, $ per year
Reagents, estimated, $ per year
Costs per sample, $
48 000
48 000
23 500
The HPLC method is not suitable for routine use because of its low throughput
(5 samples a day). Multichannel CE systems, when available, should have greatly
increased sample throughput, making them attractive for routine use.
despite the high investment costs, about one-fifth as
expensive as the method of Goldstein et al. [9].
In conclusion: The new method described here for measuring Hb A1c separates hemoglobin variants and derivatives by CE. This method is fast and reproducible, can be
automated, and uses low amounts of reagents. The performance of the assay in a routine setting needs to be
evaluated in future studies, as well as calibration with the
assay of Goldstein et al. used in the DCCT [9].
1. Nathan DM. Long-term complications of diabetes mellitus. N Engl
J Med 1993;328:1676 – 85.
2. The Diabetes Control and Complications Trial Research Group.
The effect of intensive treatment of diabetes on the development
and progression of long-term complications in insulin-dependent
diabetes mellitus. N Engl J Med 1993;329:977– 86.
3. Weykamp CW, Penders TJ, Siebelder CWM, Muskiet FAJ, Van der
Slik W. Interference of carbamylated and acetylated hemoglobins
in assays of glycohemoglobin by HPLC, electrophoresis, affinity
chromatography, and enzyme immunoassay. Clin Chem 1993;39:
138 – 42.
4. Weykamp CW, Penders TJ, Muskiet FAJ, Van der Slik W. Influence
of hemoglobin variants and derivatives on glycohemoglobin determinations, as investigated by 102 laboratories using 16 methods.
Clin Chem 1993;39:1717–23.
5. Weykamp CW, Martina WV, Van der Dijs FPL, Penders TJ, Van der
Slik W, Muskiet FAJ. Hemoglobin S, C reference values for
glycohemoglobin in heterozygous, double-heterozygous, and homozygous subjects, as established by 13 methods. Clin Chim Acta
Li SFY. Capillary electrophoresis (Vol. 52 in J Chromatogr Library).
Amsterdam: Elsevier, 1993.
US patent 08/412,032; European patent 733900.
Harris DC. Quantitative chemical analysis, 4th ed. New York: WH
Freeman Co., 1993:713–29.
Goldstein DE, Little RR, England JD, Wiedmeyer H-M, McKenzie
EM. Methods for quantitating glycosylated hemoglobins: high
performance liquid chromatography and thiobarbituric acid colorimetry. In: Clarke WL, Larner J, Pohl SL, eds. Methods of diabetic
research, Vol 2: Clinical methods. New York: John Wiley 1986:
Steigstra H, Jansen RTP, Baadenhuijsen H. Combi scheme: new
combined internal/external quality-assessment scheme in The
Netherlands. Clin Chem 1991;37:1196 –204.
Bodor GS, Little RR, Garrett N, Brown W, Goldstein DE, Nahm MH.
Standardization of glycohemoglobin determinations in the clinical
laboratory: three years of experience. Clin Chem 1992;38:
2114 – 8.
AACC subcommittee on GHB standardization. National Glycohemoglobin Standardization Program; protocol for glycohemoglobin
standardization. Washington, DC: AACC, 1996.
Ishioka N, Iyori N, Noji J, Kurioka S. Detection of abnormal
hemoglobin by capillary electrophoresis and structural identification. Biomed Chromatogr 1992;6:224 – 6.
Ong CN, Liau LS, Ong HY. Separation of globins using free zone
capillary electrophoresis. J Chromatogr Biomed Appl 1992;576:
346 –50.
Zhu M, Wehr T, Levi V, Rodriguez R, Shiffer K, Cao Z. Capillary
electrophoresis of abnormal hemoglobins associated with alphathalassemias. J Chromatogr 1993;652:119 –29.
Castagnola M, Messana I, Cassiano L, Rabino R, Rossetti DV,
Giardina B. The use of capillary electrophoresis for the determination of hemoglobin variants. Electrophoresis 1995;16:1492– 8.
Hempe JM, Craver RD. Quantification of hemoglobin variants by
capillary isoelectric focusing. Clin Chem 1994;40:2288 –95.
Molteni S, Frischknecht H, Thormann W. Application of dynamic
capillary isoelectric focusing to the analysis of human hemoglobin
variants. Electrophoresis 1994;15:22–30.
Sacks DB. Carbohydrates. In: Burtis CA, Ashwood ER, eds. Tietz
textbook of clinical chemistry, 2nd ed. Philadelphia: WB Saunders, 1994:928 –1001.
Turpeinen U, Karjaleinen U, Stenman U. Three assays for glycohemoglobin compared. Clin Chem 1995;41:191–5.