Clinica Chimica Acta 451 (2015) 305–309

Contents lists available at ScienceDirect

Clinica Chimica Acta

journal homepage: www.elsevier.com/locate/clinchim

Performance of glycated hemoglobin (HbA1c) methods evaluated with

EQAS studies using fresh blood samples: Still space for improvements
Andrea Mosca a,⁎, Renata Paleari a, Anna Carobene b, Cas Weykamp c,d, Ferruccio Ceriotti b
a
Centro per la Riferibilità Metrologica in Medicina di Laboratorio (CIRME), Dip. di Fisiopatologia medico-chirurgica e dei trapianti, Università degli Studi di Milano, Milano, Italy
b
Laboratorio di Standardizzazione, Servizio Medicina di Laboratorio, Ospedale San Raffaele, Milano, Italy
c
Department of Clinical Chemistry, Location Queen Beatrix Hospital, Winterswijk, The Netherlands
d
European Reference Laboratory, Location Queen Beatrix Hospital, Winterswijk, The Netherlands

a r t i c l e i n f o a b s t r a c t

Article history: Background: The determination of glycated hemoglobin is a key indicator for the management of diabetic
Received 1 October 2015 patients. A reference measurement system for its determination is available and IVD manufacturers should
Received in revised form 15 October 2015 have aligned their assay to this system.
Accepted 15 October 2015 Methods: Two fresh blood samples were distributed by courier to 206 Italian laboratories asking for the determi-
Available online 22 October 2015
nation of their HbA1c concentration. Target HbA1c values were assigned by the IFCC reference measurement pro-
cedure.
Keywords:
Standardization
Results: From 193 laboratories using analytical systems from five manufacturers (Bio-Rad Laboratories, A.
Diabetes Menarini Diagnostics, Roche Diagnostics, Sebia and Tosoh), we obtained a global variability of 5.3% (in terms of
Traceability CV) and of 3.8% at an HbA1c value of 37.4 mmol/mol (sample 1) and 62.0 mmol/mol (sample 2), respectively.
Accuracy With a goal for the allowable total error (TE) of 6.0%, 70% and 77% of the participants met this criterion for sam-
ples 1 and 2, respectively. Inter-laboratory CVs, were between 3.3 and 5.0% and between 2.2 and 3.7% for samples
1 and 2, respectively. Tosoh users registered the smallest inter-laboratory CV in sample 1, and Sebia's in sample 2.
With regard to trueness, all methods had a mean bias of ≤2.8% with respect to the target values, with the excep-
tion of Tosoh (bias of +6.1 and +5.8%, for samples 1 and 2, respectively).
Conclusion: These results are in good agreement with those obtained by the CAP 2014 GH2-A survey, suggesting
then that still there is an urgent need for improving a significant part of the methods currently used to measure
HbA1c.
© 2015 Elsevier B.V. All rights reserved.

1. Introduction employing lyophilized materials of not well defined matrix and appar-
ently not commutable [8]. However, within a pilot study performed at
The accurate measurement of glycated hemoglobin (HbA1c) is of the end of 2013 [9], we have proven that it is possible to run an EQAS
pivotal importance for monitoring diabetic patients [1,2] and, according by using fresh blood samples in Italy, providing that the delivery is
to the recommendations of various scientific societies, it is also strongly being performed by courier.
indicated for the diagnosis of the illness [3]. In both cases some target Therefore, we report here the results obtained in 2014, when we
values have been defined (53 and 48 mmol/mol) [4] thus demanding have reproduced on a larger scale the pilot study previously mentioned.
to the laboratory professional the measurement of this analyte with The main goals of this exercise were essentially two. First, to prove what
great attention. was the level of agreement for the measurement of HbA1c in a larger
Recently, some recommendations have been issued, in Italy and group of Italian laboratories, by evaluating their performance using
other countries defining precise analytical goals for the measurement goals for the total error based on rigorous criteria. Second, to obtain an
of HbA1c [5,6]. Beyond these goals, the regular participation to an objective estimate of the trueness of the various methods in comparison
EQAS exercise using commutable control materials having the HbA1c with the official IFCC reference measurement procedure [10].
title assigned by the primary reference measurement procedure have
been advocated [7]. Unfortunately, up to now, no EQAS service with 2. Materials and methods
such kind of characteristics is available in Italy, those mostly used
2.1. Enrolment of the participants
⁎ Corresponding author at: Dip. di Fisiopatologia medico-chirurgica e dei trapianti, Via
Fratelli Cervi 93, 20090 Segrate, Milano, Italy. An informative campaign concerning the pilot EQAS study was car-
E-mail address: andrea.mosca@unimi.it (A. Mosca). ried out in collaboration with the main diagnostic companies involved

http://dx.doi.org/10.1016/j.cca.2015.10.014
0009-8981/© 2015 Elsevier B.V. All rights reserved.
306 A. Mosca et al. / Clinica Chimica Acta 451 (2015) 305–309

in HbA1c measurement, in order to reach a suitable numbers of partici- as for the results grouped by different methods and instruments. Results
pants, well representative of the principal methods used in Italian beyond the mean ± 3 SD were considered outliers and excluded from
laboratories. the calculation. An iterative process was used to eliminate the outliers.
In order to evaluate the performance of individual laboratories, the rel-
2.2. Preparation and shipment of EQAS materials ative difference between the results returned from the participants and
the target value obtained with the reference method was calculated and
EQAS materials consisted of two fresh blood samples obtained from compared with a total allowable error (TAE) set at 6%, as derived from
single donations from a healthy and a diabetic subject, respectively. HbA1c biological variation data [12]. All computations were performed
Blood (250 mL from each subject) was collected in transfusion bags con- using Microsoft Excel 2010 software.
taining 1.8 g/L K2EDTA, final concentration. From each bag, 500 μL blood
aliquots were prepared and then shipped by courier to the participants 3. Results
as previously described [9]. Blood was stored at + 4 °C during all the
preparation steps until the shipment was carried out during the winter Two hundred and six laboratories took part to this study. The fresh
season, without any temperature control. The participants were blood samples have been shipped at the beginning of December 2014,
instructed to store the sample at +4 °C on arrival and to analyze them and the results have been collected within a week from the shipment's
within 4 days. Sample collection was performed in the laboratory of date. Data have been provided by 193 laboratories (i.e. from 93.7% of all
one of the Authors (CW) upon informed consent. the shipments).
Fig. 1 reports the most significant results grouped by methods, in re-
2.3. HbA1c target value assignment to EQAS materials spect to the target values. These were equal to 37.4 mmol/mol and
62.0 mmol/mol, for samples 1 and 2, respectively. The relative expanded
The HbA1c target value of the samples was assigned using the IFCC uncertainties (K = 2) were 0.57 and 0.91 mmol/mol, respectively. Most
reference measurement procedure, in the version HPLC-capillary elec- of the mean values were within the limits of the uncertainties around
trophoresis (CE) [10]. The analyses were performed in the laboratory the target values, except for Tosoh G7 and G8, whose mean values
of the Centre for Metrological Traceability in Laboratory Medicine were always higher in respect to the target values.
(CIRME, University of Milano), approved member of the IFCC network Fig. 2 shows the Youden plot of all participants. Apparently, we
of reference laboratories for HbA1c and present in the database of the found no difference in the possibility that different methods may be
Joint Committee on Traceability in Laboratory Medicine” (JCTLM) [11]. able to achieve the desirable performance according to their analytical
Briefly, human erythrocytes were separated from plasma, washed principle. Indeed, either the immunochemical method or the capillary
and lysed in water. The resulting hemoglobin solution was hydrolyzed electrophoresis seems to offer, although used by a different number of
with endoproteinase Glu-C sequencing grade (Roche Diagnostics, cod. participants, the same performance as those of two HPLC systems
11,047,817,001) at 37 °C for 19 h. In the first separation step, the mix- (Bio-Rad and Menarini). On the contrary, only part of the user of a
ture of peptides obtained from hemoglobin digestion was separated third HPLC method (Tosoh) is able to provide measurements within
by reversed-phase HPLC on C18 column (Shiseido, Capcell Pak UG120, the TAE limits.
5 μm, 4,6 × 150 mm) and the peak containing glycated and non- More details on how different methods performed in our EQAS study
glycated β-N-terminal hexapeptides (released from HbA1c and HbA0, is then shown in Table 1. Essentially, 70% of the participants were able to
respectively) was manually collected. In the second separation step, achieve the TAE on sample 1, and 77% on sample 2, respectively. Looking
the glycated and non-glycated hexapeptides were completely separated to the distributions of the results obtained by each method, we found
and quantified by CE performed on ProteomeLab PA800 (Beckman that the shape of these distributions (typically Gaussian) is not linked
Coulter) equipped with a fused-silica capillary, 75 μm(I.D), 77 cm in to the HbA1c concentrations of each sample.
length (AB Sciex, cod. 338473). Separation was performed at 18 KV for
35 min, at 25 °C, in 100 mmol/L sodium phosphate buffer, pH 2.5. The 4. Discussion
peak areas of the glycated and non-glycated hexapeptide were carefully
measured and HbA1c concentration was calculated from the ratio of This is the first voluntary EQAS study performed in Italy after our
hexapeptide areas by mean of a calibration function. A set of six calibra- first pilot exercise of 1997 [13]. At that time two lyophilized control ma-
tors of the IFCC Network consisting of mixtures of pure HbA1c and HbA0 terials were shipped to 117 laboratories. The samples were of proven
(lot pcal 2012, ERL/MCA Laboratory, Winterswijk, The Netherlands) commutability, as demonstrated some years later [14]. The results
were used for calibration. For each EQAS sample, two digestions were showed a global variability of 16% on the sample at physiological level
performed and each digestion product was analyzed in triplicate by of HbA1c, and of approximately 15% at the pathological level. Our pres-
CE, for a total of six measurements for sample. ent data show therefore that the global variability did significantly
The HbA1c target values were calculated from the mean of the six decrease over the years, although a rigorous judgment cannot be
measurements. The combined standard uncertainty of each target done, due to the fact that many laboratories of today were not the
value was calculated by considering three components: the measure- same as those of that time, that almost all methods changed during
ment imprecision expressed as coefficient of variation (CV), the uncer- these years, and that the materials object of the studies were not the
tainty of the HbA1c values assigned to the calibrators by the IFCC same. Similar improvements in the quality of HbA1c results have been
network coordinator and a fixed uncertainty covering other factors also reported in other countries [15]. However, a possible limitation of
and estimated on the base of our experience. The following formula our study is that the participation to this exercise was again voluntary,
was used: (CV2 + uSRM2 + uS2)1/2 where CV is the standard uncertain- not mandatory, and this fact could have potentially induced a bias in
ty derived by the imprecision, uSRM is the uncertainty of the values the recruitment of participants and in result interpretation. However,
assigned to the calibrators (0.52%) and uS is the fixed uncertainty since this was the first time that fresh blood was used as EQAS material
(0.29%). From the combined uncertainty, the expanded uncertainty in a larger study, our wish was to limit at most other potential sources of
was estimated using a factor of 2 (confidence interval of 95%). variability.
In Fig. 3 we use the model recently described by the IFCC Task Force
2.4. Results collection and elaborations on HbA1c to compare performances of the present Italian study and
those from the College of American Pathologists (CAP) Survey 2014
For each EQAS sample, the mean, the standard deviation (SD) and GH2-A, which was performed at the same time as our study i.e. by the
the inter-laboratory CV were evaluated for the overall results, as well end of 2014 [16]. Also in the CAP survey fresh blood samples were
 A. Mosca et al. / Clinica Chimica Acta 451 (2015) 305–309 307

Fig. 1. Distribution of the mean values per group of methods, respect to the target values estimated by the IFCC reference method procedure on the EQAS distributed samples. The uncer-
tainty around the target value (continuous bold line) is reported as a gray area. All the values are reported as means ± SD. The results obtained by the methods used by 2 or b2 laboratories
are not reported in this plot. A) Sample with physiological concentration of HbA1c (sample 1); target value of 37.4 ± 0.57 mmol/mol (target ± expanded uncertainty). B) Sample with
pathological concentration of HbA1c (sample 2); target value of 62.0 ± 0.91 mmol/mol (target ± expanded uncertainty).

distributed to the participants. The CAP survey was mandatory and meet the 2 sigma criterion of the Sigma-metrix model in both countries
the Italian study was voluntary. In the Italian study target values and one manufacturer touches even the minimum performance level
are set with the IFCC Reference Measurement Procedure. In the CAP sur- of the biological variation model. Fig. 3 also indicates the source of
vey targets are set with the network of NGSP reference laboratories. poor performance: relatively high imprecision for Bio-Rad and Roche
IFCC- and NGSP Reference Systems are linked with each other with (along with a small bias), and relatively high bias for Tosoh (along
the Master Equation and continuously monitored [17]. Therefore results with a good precision). Tosoh in particular has recently communicated
are well comparable. Fig. 3 shows the condensed results of: a) 3277 lab- to its users that a set of corrective actions have been undertaken in
oratories in CAP survey sample GH2-01 with an HbA1c concentration of order to improve the performance of their HbA1c methods, by circulating
48 mmol/mol, and b) of 193 laboratories in the Italian study with an new lots of calibrators. Apparently, from recent data obtained through
HbA1c of 49.7 mmol/mol. It can be seen that the overall performance the Equalis exercise [18], using also fresh blood samples, the mean bias
of all laboratories in Italy borderline met the 2 sigma criterion of the between Tosoh results and the target values provided by the European
Sigma-metrix model, but was far away from the minimum performance Reference Laboratory decreased from + 2.7 mmol/mol (at the target
level of the biological variation model, although slightly better than the value of 37.3 mmol/mol) in the 2012 exercise to + 1.1 mmol/mol
overall performance in US. It can also be seen that the mean perfor- (at the target value of 50.0 mmol/mol) in the most recent exercise.
mance between the manufacturer groups is variable, but quite similar In conclusion, we feel that only when performing EQAS studies sim-
within manufacturer groups in both countries. Three manufacturers ilar to the one we have just described, it is possible to obtain a realistic
308 A. Mosca et al. / Clinica Chimica Acta 451 (2015) 305–309

Fig. 2. Youden plot for HbA1c results of samples 1 and 2 by all individual laboratories
participating to the EQAS SIBioC project. The results obtained by different analytical sys-
Fig. 3. Quality target models applied to 4 manufacturer/instruments means in the present
tems are plotted with different colors. Continuous horizontal and vertical blue lines repre-
study and in the CAP 2014 GH2-A survey. Mean within-manufacturer interlaboratory CV
sent the target values. The gray area delimited by light dotted lines indicates the expanded
on the x axis; mean manufacturer absolute bias on the y axis. The gray stars represent
uncertainty (U) around the target values, while the square with the bold continuous line
the overall mean of all laboratories in Italy (star inscribed IT) and United States (US).
delimits the area of the TE of 6.0%. In order to plot all the data, some results have been
The dots represent the means of analytical devices of manufacturers in both countries:
increased or reduced by 0.1–0.2 mmol/mol.
Tosoh (red), Roche (green), Sebia (blue), and Bio-Rad (yellow).

representation of what is the state-of-the-art of the measurements of

HbA1c in a country, such as Italy. Of course we hope that the results ob- been undertaken to launch it in 2016 in at least three European coun-
tained by our study may constitute a strong stimulus to move forward. tries (Belgium, Germany and The Netherlands) involving probably a
As it has been recently outlined [19], improvements will only take place total of approximately 800 laboratories. In this context the use of fresh
if all stakeholders (governments and international organizations, scien- blood samples is important, because lyophilized materials are not com-
tific societies, manufacturers and specialists in laboratory medicine) act mutable for all methods. In addition, there are more and more point of
in a coordinated way to take corrective action every time EQAS clearly care instruments on the market and most of them can only work with
indicates a suboptimal situation, i.e. when the quality of the measure- fresh whole blood. Finally the use of the allowable total error of ± 6%
ment is not adequate to its clinical use. In our National context we can- has been recently criticized as being too liberal when using the HbA1c
not forget that many Regions have not launched a mandatory EQAS in the diagnosis of diabetes mellitus [20] and therefore this aspect will
scheme and that therefore many laboratories are not obligated to probably need further re-evaluation.
prove their analytical performance in an objective way.
In an international context we may think of a EurA1c trial: sharing
samples with other European EQA organizers using fresh whole blood Conflict of interest
samples and retrospectively evaluate the results in the respective coun-
tries. From our point, such an approach is feasible, and contacts have None declared.

Table 1
Results of the study grouped by method and instruments.

Methods and instruments Sample 1 Sample 2

N° labs (outliers) Mean SD CV N° labs N°labs Mean SD CV N° labs within TAE

mmol/mol mmol/mol % within TAE (outliers) mmol/mol mmol/mol % 6.0%
6.0%

Abbott Architect c4000 1 (0) 37.0 – – 1 (100) 1 (0) 63.0 – – 1 (100)

B-Analyst 1 (0) 37.0 – – 1 (100) 1 (0) 63.0 – – 1 (100)
Bio-Rad D-10 11 (2) 36.7 1.79 4.9 9 (82) 11 (2) 62.5 2.34 3.7 10 (91)
Bio-RadVariant II 25 (1) 36.7 2.03 5.5 19 (76) 25 (1) 63.1 2.35 3.7 22 (88)
Bio-Rad Variant IITurbo 14(0) 37.7 1.44 3.8 12 (86) 14 (0) 62.6 1.86 3.0 13 (93)
Menarini - HA8160 VP 12 (0) 37.9 1.38 3.6 9 (75) 12 (0) 64.3 2.23 3.5 8 (67)
Menarini - HA8160 TP 20 (0) 36.8 1.80 4.9 17(85) 20 (0) 63.3 2.49 3.9 16 (80)
Menarini - HA8180 V 5 (0) 37.8 0.45 1.2 5 (100) 5 (0) 62.6 0.55 0.9 5 (100)
Roche - Cobas C501 11(0) 36.8 0.84 2.3 10 (91) 11 (0) 60.8 1.36 2.2 11 (100)
Roche - Integra 400 3 (0) 37.1 1.55 4.2 3 (100) 3 (0) 62.5 2.53 4.0 3 (100)
Roche - Modular 2 (0) 40.5 – – 0 (0) 2 (0) 66.0 – – 0 (0)
Sebia - Capillarys HbA1c 22 (0) 36.7 1.29 3.5 18 (82) 22 (0) 62.5 1.01 1.6 22 (100)
Sebia - Minicap HbA1c 3 (0) 37.3 1.53 4.1 3 (100) 3 (0) 60.0 1.73 2.9 3 (100)
Tosoh - G7 8 (0) 40.0 1.60 4.0 3 (38) 8 (0) 66.4 2.07 3.1 3 (38)
Tosoh - G8 49 (0) 39.7 1.24 3.1 25 (51) 49 (0) 65.6 1.63 2.5 25 (51)
Tosoh - GX 2 (0) 37.5 – – 0 (0) 2 (0) 63.0 – – 2 (100)
Other 1 (0) 42.0 – – 0 (0) 0 (1)
 A. Mosca et al. / Clinica Chimica Acta 451 (2015) 305–309 309

Acknowledgments [10] J.O. Jeppsson, U. Kobold, J. Barr, A. Finke, W. Hoelzel, T. Hoshino, et al., Approved IFCC
reference method for the measurement of HbA1c in human blood, Clin. Chem. Lab.
Med. 40 (2002) 78–89.
Research project fully supported by SIBioC – Medicina di Laboratorio [11] I. Infusino, F. Braga, R. Paleari, A. Mosca, M. Panteghini, The joint committee for
(Italian Society of Clinical Biochemistry and Clinical Molecular Biology). traceability in laboratory medicine (JCTLM): a global cooperation to promote the
standardisation of test results in laboratory medicine, Biochim. Clin. 35 (2011)
377–381.
References [12] F. Braga, A. Dolci, M. Montagnana, F. Pagani, R. Paleari, G.C. Guidi, et al., Revaluation
of biological variation of glycated hemoglobin (HbA1c) using an accurately designed
[1] D.B. Sacks, M. Arnold, G.L. Bakris, D.E. Bruns, A.R. Horvath, M.S. Kirkman, et al., protocol and an assay traceable to the IFCC reference system, Clin. Chim. Acta 412
Guidelines and recommendations for laboratory analysis and management of (2011) 1412–1416.
diabetes mellitus, Clin. Chem. 57 (2011) e1–e47. [13] A. Mosca, R. Paleari, A. Trapolino, F. Capani, G. Pagano, M. Plebani, A re-evaluation of
[2] American Diabetes Association, Glycemic targets, Diabetes Care 38 (2015) S33–S40. glycohemoglobin standardization: the Italian experience with 119 laboratories and
[3] A. Lapolla, A. Mosca, D. Fedele, The general use of glycated haemoglobin for the 12 methods, Eur. J. Clin. Chem. Clin. Biochem. 35 (1997) 243–248.
diagnosis of diabetes and other categories of glucose intolerance: still a long way [14] A. Mosca, R. Paleari, A. Madè, C. Ferrero, M. Locatelli, F. Ceriotti, Commutability
to go, Nutr. Metab. Cardiovasc. 21 (2011) 467–475. of control materials in glycohemoglobin determinations, Clin. Chem. 44 (1998)
[4] Associazione Medici Diabetologi (AMD), Società Italiana di Diabetologia (SID), Stan- 632–638.
dard italiani per la cura del diabete mellito, 2014 58 (www.standarditaliani.it, access [15] R.R. Little, C.L. Rohlfing, The long and winding road to optimal HbA1c measurement,
verified in date 2 July 2015). Clin. Chim. Acta 418 (2013) 63–71.
[5] C.W. Weykamp, A. Mosca, P. Gillery, M. Panteghini, The analytical goals for hemo- [16] C. Weykamp, G. John, P. Gillery, E. English, L. Ji, E. Lenters-Westra, et al., Investigation
globin A1c measurement in IFCC units and national glycohemoglobin standardiza- of 2 models to set and evaluate quality targets for Hb A1c: biological variation and
tion program units are different, Clin. Chem. 57 (2011) 1204–1206. sigma-metrics, Clin. Chem. 61 (2015) 752–759.
[6] F. Braga, M. Panteghini, Standardization and analytical goals for glycated hemoglo- [17] C. Weykamp, W.G. John, A. Mosca, T. Hoshino, R. Little, J.O. Jeppsson, et al., The IFCC
bin measurement, Clin. Chem. Lab. Med. 51 (2013) 1719–1726. reference measurement system for HbA1c: a 6-year progress report, Clin. Chem. 54
[7] A. Mosca, M.T. Branca, M. Carta, M.L. Genna, C.B. Giorda, R. Ghidelli, et al., Recom- (2008) 240–248.
mendations for the implementation of the international standardization of glycated [18] www.equalis.se/en/start/ (accessed October 15, 2015).
hemoglobin measurement in Italy, Biochim. Clin. 33 (2009) 258–261. [19] F. Ceriotti, The role of external quality assessment schemes in monitoring and
[8] A. Mosca, Some practical advices on how to implement the international standard- improving the standardization process, Clin. Chim. Acta 432 (2014) 77–81.
ization of glycated hemoglobin measurement in Italy, Biochim. Clin. 35 (2011) [20] A. Åsberg, I.H. Odsæter, S. Magnus Carlsen, G. Mikkelsen, Using the likelihood ratio
36–41. to evaluate allowable total error — an example with glycated hemoglobin
[9] A. Mosca, C. Weykamp, Feasibility of an EQAS for HbA1c in Italy using fresh blood (HbA1c), Clin. Chem. Lab. Med. 53 (2015) 1459–1464.
samples, Clin. Chem. Lab. Med. 57 (2014) e151–e153.