Evaluation of the Sebia Capillarys 3 Tera and the Bio-Rad D ...

Evaluation of the Sebia Capillarys 3 Tera and the Bio-RadD-100 Systems for the Measurement of Hemoglobin A1c

Margaux Herpol, MD, Katrien Lanckmans, PhD, Stefaan Van Neyghem, Pascale Clement,Stefanie Crevits, Kim De Crem, MS, Frans K. Gorus, MD, PhD, and Ilse Weets, MD, PhD

From the Department of Clinical Chemistry and Radio-Immunology, Universitair Ziekenhuis Brussel, Brussels, Belgium.

Key Words: Analytical performance; Capillary zone electrophoresis; Hemoglobin A1c; HPLC

Am J Clin Pathol July 2016;146:67-77

DOI: 10.1093/AJCP/AQW081

ABSTRACT

Objectives: We evaluated the Bio-Rad (Irvine, CA) D-100 and

the Sebia (Lisses, France) Capillarys 3 Tera for the measure-

ment of hemoglobin A1c (HbA1c) in venous blood samples.

Methods: Whole-blood samples and control material were

analyzed with the D-100 and Capillarys 3 Tera and com-

pared with our routine method, HLC-723G7 (Tosoh, Tokyo,

Japan). An evaluation protocol to test precision, trueness,

linearity, carryover, and selectivity was set up according to

Clinical and Laboratory Standards Institute guidelines. The

results were presented in National Glycohemoglobin

Standardization Program and International Federation of

Clinical Chemistry (IFCC) units.

Results: Both systems showed excellent precision (total co-

efficients of variation<2%, IFCC) and bias (<0.3% or

3 mmol/mol). Linearity was demonstrated for HbA1c values

from 3.8% (18 mmol/mol) to 18.5% (179 mmol/mol). Results

were correlated with the routine method using Bland-

Altman analysis, showing a mean difference of 0.33% or

3.6 mmol/mol for the D-100 and of 0.25% or 2.6 mmol/mol

for the Capillarys 3 Tera vs HLC-723G7. None of the auto-

mated instruments were prone to interferences by labile

HbA1c (�10 g/L glucose), carbamylated hemoglobin

(�0.5 mmol/L potassium cyanate), hemoglobin variants,

bilirubin (�15 mg/dL), and triglycerides (�3,360 mg/dL).

Conclusions: The Bio-Rad D-100 and the Sebia Capillarys

3 Tera instruments performed well for the determination of

HbA1c in terms of quality criteria as well as for sample

throughput.

The prevalence and incidence of diabetes are on the rise

worldwide. If current trends continue, it is expected that

there will be more than 592 million patients with diabetes

by 2035.1 Hemoglobin (Hb) A1c (HbA1c)—a result of the

nonenzymatic covalent binding of glucose on N-terminal

valine residues of hemoglobin b chains—is presently the

most important marker for monitoring of glycemic control

in daily clinical practice. It is associated with risk of chronic

complications in both type 1 and type 2 diabetes mellitus.

This risk rises exponentially with increasing mean HbA1c.2

In 2010, the American Diabetes Association (ADA)

introduced HbA1c as an additional parameter for the diagno-

sis of diabetes.3 Therefore, the role of HbA1c as a biomarker

for glycemic control and diagnosis will only become more

important.

Major efforts have been made to standardize the

HbA1c determination. HbA1c values were initially aligned

with the Diabetes Control and Complications Trial

Research Group2 and are currently standardized according

to the International Federation of Clinical Chemistry

(IFCC) reference method.4,5 This evolution toward stand-

ardization of the HbA1c determination was paralleled by

the introduction of more stringent criteria for precision and

accuracy (WIV-ISP, Scientific Institute of Public Health,

Brussels, Belgium), as well as for diabetes control.6,7 In

addition, point-of-care (POC) devices were introduced in

second-line (diabetes clinics) and first-line (general practi-

tioner) health care. Although quality of the POC systems

has improved, only few devices meet acceptable perform-

ance criteria.5,8,9 Recently, Sebia (Lisses, France) and Bio-

Rad (Irvine, CA) have launched new automated analyzers

not only to meet the more stringent analytical

© American Society for Clinical Pathology, 2016. All rights reserved.For permissions, please e-mail: [email protected]

Am J Clin Pathol 2016;146:67-77 67DOI: 10.1093/ajcp/aqw081

67

AJCP / ORIGINAL ARTICLE

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

performances and ISO15189 criteria but also to improve

usability and throughput.

In this study, we present the evaluation of beta versions

of the Sebia Capillarys 3 Tera and the Bio-Rad D-100 in

comparison with the HLC-723G7 instrument (Tosoh,

Tokyo, Japan) routinely used in our laboratory for the quan-

tification of HbA1c in whole blood. A thorough validation

of both test instruments was performed, including precision,

trueness, linearity, carryover, and specificity. Stability of

HbA1c in capillary and venous blood samples was eval-

uated, as well as the usability and throughput of both auto-

mated instruments.

Materials and Methods

Analyzers

Quantification of HbA1c was performed simultaneously

on the Bio-Rad D-100 and the Sebia Capillarys 3 Tera. Results

were compared with the HLC-723G7 (Tosoh). All analyzers

were used according to the manufacturer’s instructions.

The Bio-Rad D-100 is a fully automatic benchtop ana-

lyzer for determination of HbA1c, based on chromatographic

separation of hemoglobin fraction by cation-exchange high-

performance liquid chromatography coupled to UV detec-

tion. Whole blood (5 mL) is automatically diluted (1/300)

with hemolyzing solution and loaded onto the analytical

cartridge. The separated hemoglobin fractions are detected

at 415 nm. Calibration is obtained by Bio-Rad calibrators,

providing results in IFCC units (mmol/mol) or in National

Glycohemoglobin Standardization Program (NGSP) units

(%). For quality control samples and patient samples, the

Clinical Data Management software transforms the results

from one unit to another by using the following master

equation: IFCC (mmol/mol)¼ 10.93 NGSP (%) –23.50.4,5

HbA1c is calculated as the ratio of HbA1c to the total HbA

fractions. The Bio-Rad software detects normal and abnor-

mal peaks, such as hemoglobin variants, and reruns auto-

matically as preset by the customer. The chromatograms are

composed of seven peaks, eluting in the following order:

HbA1a, HbA1b, HbF, labile A1c, HbA1c, P3, and HbA0. This

P3 peak is a “degradation peak,” coeluting with unidentified

components just before the HbA0.

The Sebia Capillarys 3 Tera is an automated analyzer

based on capillary zone electrophoresis and UV detection

for the quantitative analysis of HbA1c, proteins (serum and

urine), immunotyping (serum and urine), and carbohydrate

deficient transferrin. The Sebia Capillarys 3 Tera instrument

includes 12 silica capillaries functioning in parallel, allow-

ing 12 simultaneous analyses. Before each run, the capilla-

ries are washed with washing solution and prepared for the

next analysis with buffer. Whole blood (18mL) is automatic-

ally diluted (1/6) with hemolyzing solution and injected at

the anodic end of the capillary. The various hemoglobin

components are separated by high-voltage protein separ-

ation and consequently detected at the cathodic end of the

capillary at 415 nm. Calibration is obtained by the Sebia

calibrators, giving results in mmol/mol, which are then con-

verted to NGSP units using the master equation. The for-

mula used for quantification of HbA1c is based on IFCC

recommendations: HbA1c ¼HbA1c/(HbA1cþHbA0). The

Phoresis software (preliminary version 9.0.0; Sebia, Lisses,

France) detects both normal and atypical profiles (eg, hemo-

globin variants, increased HbA2). The chromatogram is

composed of at least four peaks: HbA1c, “other HbA,”

HbA0, and HbA2. In case there is an elevated HbF, a peak

appears adjacent to the HbA0 peak. The presence of a hemo-

globin variant leads to extra peaks in the chromatogram.

Samples

Quality control (QC) materials used throughout the

evaluation were Lyphochek Diabetes Control 1 and 2 sam-

ples (Bio-Rad). As for the Sebia Capillarys 3 Tera instru-

ment, Capillarys HbA1c Controls 1 and 2 (Sebia) were used

for the positioning of the HbA1c peak. Both types of QC ma-

terial are human lyophilized whole-blood controls. External

QC samples supplied by the Scientific Institute of Public

Health (WIV-ISP) were used for trueness evaluation.

Patient samples were obtained from residual material of

diabetic and nondiabetic patients whose HbA1c or hemo-

globinopathy screening was assessed routinely in our la-

boratory (UZ Brussels, Brussels, Belgium). Venous blood

samples were collected in tubes containing K-EDTA as an

anticoagulant. For the evaluation of hemoglobin variants, a

collection of patient samples containing hemoglobin vari-

ants stored at less than –70 �C was used. For the interference

testing of carbamylated HbA1c, triglycerides, and bilirubin,

erythrocytes of the patient blood samples were obtained

after centrifugation during 10 minutes at 2,000 g and remov-

ing the plasma. The samples in this study were used accord-

ing to procedures supervised by local authorities responsible

for ethical research and working in accordance with the

Helsinki Declaration of 1975.

Evaluation Protocol

Imprecision

Within-run imprecision was determined by performing

12 replicates of two patient samples, one with a normal and

one with an elevated HbA1c level. Total imprecision was

evaluated according to the Clinical and Laboratory

Standards Institute (CLSI) EP05-A2 guidelines.10 This

implies two replicate analyses per day of the Lyphochek

Herpol et al / EVALUATION CAPILLARYS 3 TERA AND D-100

68 Am J Clin Pathol 2016;146:67-77 © American Society for Clinical Pathology68 DOI: 10.1093/ajcp/aqw081

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

controls (levels 1 and 2) during 20 days. Imprecision on

Capillarys 3 was measured on each of the 12 capillaries. A

Grubbs test was performed to detect outliers, followed by

calculation of the mean, standard deviation, and percent co-

efficient of variation (CV%).

Trueness

Trueness has been evaluated according to CLSI EP9-

A2 guidelines11 by analyzing 12 external QC samples of the

national health care survey, supplied by the WIV-ISP

(2014). The experimentally obtained HbA1c results were

compared with assigned values (both in IFCC and NGSP

units). Linear regression analysis was performed and bias

was estimated at HbA1c levels of 5% (31 mmol/mol), 8%

(64 mmol/mol), and 11% (97 mmol/mol).

Linearity

Linearity was assessed in two ways: first by using the

Bio-Rad Lyphochek Hemoglobin A1c linearity kit, com-

posed of lyophilized whole-blood samples, and, second, by

using patient samples according to the CLSI EP06-A guide-

lines12 and measuring HbA1c of samples prepared by pro-

portional mixing of two samples with a high (16.2%;

154 mmol/mol) and a low (5.2%; 33 mmol/mol) HbA1c. The

patient samples had similar but slightly different total hemo-

globin concentrations (14.8 and 16.3 g/dL): it is not

excluded that this could have an impact on the linearity re-

sults. Linear regression analysis was performed using the

coefficient of correlation (R), comparing expected values

with obtained values.

Carryover

Carryover was tested by analyzing blood samples with a

high (H) and low (L) concentration of HbA1c in the following

order: L, L, L, H, H, L, H, H, L, L, L, L, H, H, L, H, H, L, H,

H, L (EP Evaluator Release 9 protocol, Data Innovations,

Burlington, VT). Carryover was calculated by the mean of the

high-low results minus the mean of the low-low results. The

error limit is defined as three times the SD of the low-low

results. Carryover could be suspected if the calculated carry-

over is higher than the error limit.

Method Comparison

HbA1c results of both test instruments were compared

with the results obtained with the routine method (Tosoh

HLC-723G7). Samples from 173 patients were included.

The HbA1c range was between 3.6% (16 mmol/mol) and

16.2% (154 mmol/mol), and none of the samples contained

hemoglobin variants. The samples were analyzed over 30

operating days, followed by outlier detection according to

the CLSI EP9-A2 guidelines.11 Data were further analyzed

using Passing-Bablok regression analysis and a Bland-

Altman plot.

Interference Testing

The protocol that was used for the assessment of inter-

fering substances has been previously described by our la-

boratory.13 According to the CLSI EP7-A2 guidelines,14 the

possible error caused by potentially interfering substances

(such as labile HbA1c, carbamylated hemoglobin, bilirubin,

triglycerides, hemoglobin content, and hemoglobin variants)

was compared with an allowable error. The applied allow-

able error was derived from the national health care survey

of the WIV-ISP—namely, 0.3% NGSP units and 3 mmol/L

IFCC units7 ( Table 1 , good absolute bias).

Labile HbA1c. The interference of labile HbA1c on

HbA1c measurements was assessed with three patient sam-

ples containing a normal (5.5%; 37 mmol/mol), medium

(8.2%; 66 mmol/mol), and high (12.2%; 109 mmol) HbA1c

level. The samples were spiked with five different amounts

of glucose solution (up to 5,000 mg/dL) and incubated at

37 �C for 3 hours with mixing every 30 minutes. After dupli-

cate analysis, the absolute bias was calculated.

Carbamylated HbA1c. Erythrocytes of three patients

with a normal (5.0%; 31 mmol/mol), medium (9.0%;

75 mmol/mol), and high (13.3%; 122 mmol) HbA1c concen-

tration were incubated with various concentrations (0.15, 0.5,

and 1 mmol/L) of potassium cyanate (KCNO) to generate

in vitro carbamylation of HbA1c. After incubation at 37 �C

Table 1Scientific Institute of Public Health (WIV-ISP) Criteria (2014) for Absolute Bias, Imprecision, and Linearity

Parameter Excellent Good Acceptable Poor Unacceptable

Criteria for HbA1c in NGSP units

Absolute bias, % <0.2 0.20-0.30 0.31-0.40 0.41-0.50 >0.5

Imprecision (CV), % <1.4 1.4-2.0 2.1-2.9 3.0-4.0 >4

Linearity (r) >0.9970 0.9950-0.9970 0.9900-0.9949 0.9800-0.9899 <0.9800

Criteria for HbA1c in IFCC units

Absolute bias, mmol/mol <2 2-3 3-4 4-5 >5

Imprecision (CV), % <2 2.0-2.99 3.0-3.99 4.0-4.99 >5

Linearity (r) >0.9970 0.9950-0.9970 0.9900-0.9949 0.9800-0.9899 <0.9800

CV, coefficient of variation; HbA1c, hemoglobin A1c; IFCC, International Federation of Clinical Chemistry; NGSP, National Glycohemoglobin Standardization Program.

AJCP / ORIGINAL ARTICLE

© American Society for Clinical Pathology Am J Clin Pathol 2016;146:67-77 69DOI: 10.1093/ajcp/aqw08169

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

for 3 hours and mixing every 30 minutes, the RBCs were

resuspended in their corresponding plasma. All samples were

analyzed in duplicate, and absolute bias values were

calculated.

Triglycerides and bilirubin. Interference of triglycer-

ides and bilirubin was performed with three patient sam-

ples with a normal (5.5%; 37 mmol/mol), medium (8.1%;

65 mmol/mol), and high (10.0%; 86 mmol/mol) HbA1c

concentration. Erythrocytes of these samples were mixed

(50:50 v/v) with various dilutions of triglyceride-rich

and bilirubin-rich plasma up to a final concentration of re-

spectively 3,361 mg/dL and 14 mg/dL. All samples were

analyzed in duplicate, and absolute bias values were

calculated.

Hemoglobin content. The effect of the hemoglobin con-

centration on the measurement of HbA1c was evaluated by

diluting a whole-blood sample containing a normal HbA1c

and a high hemoglobin concentration (16.8 g/dl) with saline,

obtaining hemoglobin values ranging from 5.0 to 16.8 g/dL.

After analysis, the absolute bias was calculated.

Hemoglobin variants. The influence of hemoglobin

variants on the measurement of HbA1c on both test instru-

ments was investigated using patient samples containing

HbS (n¼ 20), HbD (n¼ 5), HbE (n¼ 2), and HbC (n¼ 5).

The HbA1c concentrations of these samples were compared

with the results obtained with a Sebia Capillarys 2 Flex ana-

lyzer. This analyzer was evaluated in 2013 in our laboratory

for detection of hemoglobin variants.13 The HbA1c results

were further analyzed using Passing-Bablok regression ana-

lysis and a Bland-Altman plot.

Stability of HbA1c

The stability of venous blood samples at 2�C to 8 �C

was assessed by repeated analysis of three patient samples

with normal (5.5%; 37 mmol/mol), medium (8.0%;

63 mmol/mol), and high (12.5%; 113 mmol/mol) HbA1c

concentration during 3 weeks on both test instruments.

Room temperature stability of blood, contained in the

Bio-Rad Hemoglobin Capillary Collection System (HCCS)

for HbA1C analysis, was briefly evaluated by transferring

5 mL of venous blood samples (n¼ 5) in the HCCS. The ca-

pillary collection systems, containing a hemolyzing solu-

tion, were stored during 1 week and analyzed on days 0, 1,

3, 4, and 7. The samples were tested on the Bio-Rad D-100

analyzer. Chromatograms were visually evaluated, and ab-

solute bias was calculated.

Detection of Hemoglobin Variants

The capacity of demonstrating the presence of hemo-

globin variants in the HbA1c modus was investigated using

samples containing common hemoglobin variants such as

HbS, HbD, and HbE, as well as more rare variants or double

heterozygotes such as HbO-Arabe, HbSC, HbH, HbSS,

HbEE, HbCD, HbJ-Toronto, HbS/HbG Philadelphia,

HbA20, HbG Philadelphia, and HbH Barts.

Sample Throughput

A set of 150 patient samples was analyzed for HbA1c

on both systems. The set comprised K-EDTA tubes of dif-

ferent sizes and from different manufacturers. The duration

of the batch analysis was counted from startup of the system

and included analysis of two levels of internal quality con-

trol material at the beginning and end of the run.

Results

Imprecision

At normal and high HbA1c levels, within-run and total

CV% were below 1.4% (NGSP units) and below 2% (IFCC

units) for both analyzers Table 2 , showing excellent preci-

sion, based on the analytical goals from the national health

care survey of WIV-ISP (Table 1).7

Trueness

Twelve external QC samples ranging between 5.2%

(33 mmol/mol) and 9.5% (80 mmol/mol) were analyzed on

both systems. Absolute bias values at three HbA1c levels

are shown in Table 1. Linear regression analysis showed a

slope of 0.93 (IFCC units; slope¼ 0.93) and an intercept of

0.43 (IFCC units; intercept¼ 3.1) for the Bio-Rad D-100

and a slope of 1.02 (IFCC units; slope¼ 1.03) and intercept

of –0.27 (IFCC units; intercept¼ –2.7) for the Sebia

Capillarys 3 Tera. According to the analytical goals of the

WIV-ISP7 (Table 1), there is an excellent trueness on all

levels for the Sebia Capillarys 3 Tera, while the Bio-Rad

D-100 shows a slightly negative bias in the high concentra-

tion range but still acceptable according to the WIV-ISP

criteria.

Linearity

Using the Bio-Rad linearity kit, linearity was demon-

strated for HbA1c values ranging from 3.8% (18 mmol/mol)

to 18.5% (179 mmol/mol). For both analyzers, linearity was

excellent (R> 0.997) according to the analytical goals of

the WIV-ISP7 (Table 1). Slope, intercept, and correlation

coefficient are shown in Table 2. Since there is a significant

bias for the Bio-Rad linearity kit and no bias for routine

samples, it is likely that there is a matrix effect for this

linearity kit analyzed by capillary electrophoresis. Results

were similar when using mixtures of two patient sam-

ples with high and low HbA1c, although the range was

Herpol et al / EVALUATION CAPILLARYS 3 TERA AND D-100

70 Am J Clin Pathol 2016;146:67-77 © American Society for Clinical Pathology70 DOI: 10.1093/ajcp/aqw081

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

slightly narrower—namely, 5.0% (31 mmol/mol) to 15.3%

(143 mmol/mol). Linearity of the Sebia method using the

sample dilution protocol was good (NGSP units) to excel-

lent (IFCC units).

Carryover

A carryover of 0.12% and –0.02% was calculated for

respectively the Bio-Rad D-100 and Sebia Capillarys 3 Tera

when analyzed in NGSP units. Both results are lower than

the respective error limits (1.16 for the Bio-Rad D-100 and

0.00 for the Sebia Capillarys 3 Tera). Therefore, there was

no significant carryover detected between high and low

HbA1c samples. Similar results (ie, no carryover) were

found when analyzing the data in IFCC units.

Method Comparison

Passing-Bablok regression analysis of the data obtained

with the Bio-Rad D-100 vs Tosoh HLC-723G7 showed a slope

(95% confidence interval [CI]) of 0.92 (0.91-0.93; NGSP

units) and 0.92 (IFCC units) with an intercept (95% CI) of

0.31 (0.24-0.38; NGSP units) and 1.29 (0.86-1.80; IFCC units)

Figure 1A . Passing-Bablok regression on the data obtained

with the Sebia Capillarys 3 Tera yielded a slope (95% CI) of

0.97 (0.95 to 0.98; NGSP units) and 0.97 (0.96 to 0.98; IFCC

units) with an intercept (95% CI) of 0.02 (–0.07 to 0.10; NGSP

units) and –0.69 (–1.29 to –0.08; IFCC units) Figure 1C . The

Bland-Altman plot for the Bio-Rad D-100 Figure 1B showed

a mean difference of 0.33% or 3.6 mmol/mol. For the Sebia

Capillarys 3 Tera Figure 1D , the Bland-Altman plot showed

a mean difference of 0.25% or 2.6 mmol/mol.

Because of the known positive bias of the Tosoh HLC-

723G7 in the higher ranges,15 the results obtained by the D-

100 and the Capillarys 3 Tera were also compared. Passing-

Bablok regression analysis of results by the D-100 vs the

Capillarys 3 Tera showed a slope (95% CI) of 1.05 (1.04 to

1.06; NGSP units) and 1.05 (1.04 to 1.06; IFCC units) with

an intercept (95% CI) of –0.31 (–0.38 to –0.21; NGSP units)

and –1.90 (–2.63 to –1.36; IFCC units) Figure 2A . A

mean difference of –0.07% or –1 mmol/mol was found

Figure 2B in the Bland-Altman analysis.

Interference Studies

Labile HbA1c

On the Bio-Rad D-100, the peaks identified as labile

A1c and P3 increased as a function of glucose concentration.

Concerning HbA1c, the calculated absolute bias for the three

samples (low, medium, and high HbA1c) at different glucose

concentrations varied between –0.9% and 0.2% or –9.5 and

Table 2Overall Performance Characteristics for the Measurement of HbA1c on the Bio-Rad D-100 and Sebia Capillarys 3 Tera

Bio-Rad D-100 Sebia Capillarys 3 Tera

Characteristic NGSP IFCC NGSP IFCC

Imprecision (EP5)

Within-run, %CV at 5.5% (37 mmol/mol) 0.95 1.41 1.27 1.84

CV at 12.0% (108 mmol/mol) 0.68 0.82 0.64 0.76

Total, %CV at 5.0% (31 mmol/mol) 0.89 1.84 1.07 1.8

CV at 9.5% (80 mmol/mol) 0.87 0.88 0.80 1.88

Trueness (EP9)a

Absolute bias at 5% (31 mmol/mol) 0.09 1.11 –0.16 –1.77

Absolute bias at 8% (64 mmol/mol) –0.12 –1.08 –0.09 –0.77

Absolute bias at 11% (97 mmol/mol) –0.33 –3.27 –0.03 0.23

Linearity: Bio-Rad linearity kit

Slope 0.97 0.95 1.21 1.19

Intercept 0.16 1.68 –1.44 –10.22

Coefficient of correlation (r) 0.9995 0.9999 0.9980 0.9997

Linearity: mixtures of patient samples

Slope 1.00 0.99 1.01 1.02

Intercept –0.12 –1.20 –0.35 –3.47

Coefficient of correlation (r) 0.9990 0.9994 0.9967 0.9971

Interference studies (EP7): absolute bias, %Labile HbA1c (�10 g/L glucose) –0.3 to 0.2 –2.5 to 1.5 –0.3 to 0.1 –3.0 to 1.5

Carbamylated hemoglobin (�1 mmol/L KCNO) �–0.4 �–4.0 �–0.3 �–2.0

Bilirubin (�15 mg/dL) –0.1 to 0.1 –1.0 to 1.0 �–0.2 �2.5

Triglycerides (�3,360 mg/dL) <0.1 –0.5 to 1.0 –0.1 to 0.1 –0.5 to 1.0

Hemoglobin content (1.7-16.8 g/dL) �–0.2 �–2.0 �–0.2 –2.0 to 1.0

CV, coefficient of variation; HbA1c, hemoglobin A1c; IFCC, International Federation of Clinical Chemistry; KCNO, potassium cyanate; NGSP, National Glycohemoglobin

Standardization Program.aValues are presented as percentages for NGSP and as mmol/mol for IFCC.

AJCP / ORIGINAL ARTICLE

© American Society for Clinical Pathology Am J Clin Pathol 2016;146:67-77 71DOI: 10.1093/ajcp/aqw08171

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

1.5 mmol/mol. Up to a glucose concentration of 10 g/L, the

bias was below the maximum allowable bias set by the

WIV-ISP (Table 2).

On the Sebia Capillarys 3 Tera instrument, there was no

visual effect on the electropherogram or the quantification

of HbA1c. Up to the maximal tested glucose concentration

(50 g/L), the absolute bias ranged between –0.3% and 0.1%

or –3.0 and 1.5 mmol/mol, which meets the criterion of

0.3% or 3 mmol/mol.

Carbamylated Hemoglobin

On the Bio-Rad D-100, the in vitro experiments with

potassium cyanate showed a similar effect on the chromato-

grams as for the experiments with high glucose concentra-

tions—namely, increasing labile HbA1c and P3 peaks with

increasing KCNO concentrations. The absolute bias for

HbA1c was a maximum of –0.4% (–4.0 mmol/mol) for all

three patient samples at the different KCNO concentrations.

On the Sebia Capillarys 3 Tera, the KCNO had a visual effect

on the “other HbA” peak, but did not affect the quantification

of HbA1c. A maximal absolute bias of –0.3% (–2.0 mmol/

mol) was found. For both analyzers, the trueness criteria of

the WIV-ISP (0.3% or 3 mmol/mol) were met, except for the

samples containing the highest KCNO concentration

(1 mmol/L) and analyzed with the D-100 analyzer.

Triglycerides and Bilirubin

Up to a concentration of 15 mg/dL bilirubin and a con-

centration of 3,360 mg/dL triglycerides, the measured

HbA1c values for both analyzers were within 0.2%

(2.5 mmol/mol) of the baseline HbA1c values. There was no

observed interference on the quantification of HbA1c or in

terms of additional/changing peaks in the chromatogram/

electropherogram.

0 20 40 60 80 100 120 140 160

20

40

60

80

100

120

140

160

0

Hb

A1c

(mm

ol/

mo

l)—B

io-R

ad D

-100

HbA1c (mmol/mol)—Tosoh G7

Mean3.6

–1.96 SD–0.9

+1.96 SD8.1

500 100 150 200

10

12

14

6

8

4

0

2

–2

Hb

A1c

(mm

ol/

mo

l)—

To

soh

G7

and

Bio

-Rad

D-1

00

Mean HbA1c (mmol/mol) of Tosoh G7 and Bio-Rad D-100

Mean2.6

–1.96 SD–0.8

+1.96 SD6.1

500 100 150 200

6

8

10

4

2

–2

0

–4

Hb

A1c

(mm

ol/

mo

l)—

To

soh

G7

and

Seb

ia C

apill

arys

3 T

era

Mean HbA1c (mmol/mol) of Tosoh G7 and Sebia Capillarys 3 Tera

0 50 80 150 200

20

40

60

80

100

120

140

160

0Hb

A1c

(mm

ol/

mo

l)—S

ebia

Cap

illar

y 3

Tera

HbA1c (mmol/mol)—Tosoh G7

A B

C D

Figure 1 Passing-Bablok regression analysis and Bland-Altman plot of hemoglobin A1c (HbA1c) results (International

Federation of Clinical Chemistry units) using whole-blood patient samples, obtained with the Bio-Rad D-100 (A, B) and the

Sebia Capillarys 3 Tera (C, D) in comparison with the Tosoh HLC-723G7. Passing-Bablok: blue solid line (with 95% confidence

interval [CI]) represents the linear regression curve. Bland-Altman Plot: blue solid line (with 95% CI) represents mean absolute

difference, while the brown dashed lines represent 95% limits of agreement.

Herpol et al / EVALUATION CAPILLARYS 3 TERA AND D-100

72 Am J Clin Pathol 2016;146:67-77 © American Society for Clinical Pathology72 DOI: 10.1093/ajcp/aqw081

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

Hemoglobin Content

Serial dilutions of a patient sample with a hemoglobin

concentration of 16.9 g/dL and a normal HbA1c level re-

sulted in a hemoglobin range between 1.7 and 16.9 g/dL.

The Bio-Rad D-100 system reported total areas from 10,714

to 78,321 units. The calculated maximum bias was –0.2%

(–2.0 mmol/mol). The same evaluation was done on the

Sebia Capillarys 3 Tera, where optic densities were noted

from 0.033 to 0.258, resulting in a maximum bias of –0.2%

(–2.0 mmol/mol) at the lowest hemoglobin level. For both

analyzers, there was no significant bias detected, even for a

hemoglobin concentration of 1.7 g/dL. Both automates gave

warnings for the lowest hemoglobin concentrations, al-

though HbA1c results were not affected.

Hemoglobin Variants

HbS and HbC were visually well separated from the

HbA0 peak, both for the Bio-Rad and Sebia Capillarys 3

Tera Figure 3A , Figure 3B , Figure 4A , and Figure 4B .

On the Bio-Rad D-100, four samples from three patients

with HbAS (n¼ 3) and one with HbAC (n¼ 1) showed an

absolute bias ranging from –0.5% to 0.7% (NGSP units) and

from –7 to 5 mmol/mol (IFCC units) in comparison with the

Sebia Capillarys 2 Flex Piercing. One of the patients had a

very low HbA1c level of 3.8% (18 mmol/mol), and the other

patients had normal or high HbA1c values.

On the Bio-Rad D-100, HbD and HbE eluted shortly

after the HbA0 peak Figure 4C and Figure 4D . In one pa-

tient with a high HbA1c value (13.4% or 123 mmol/mol), an

absolute bias of –2.2% (–24 mmol/mol) was observed,

whereas another patient with a low HbA1c value (3.7% or

17 mmol/mol) showed an absolute bias of 0.9% (–11 mmol/

mol) on the Bio-Rad D-100 compared with the Capillarys 2

Flex Piercing. For the other three patients with normal

HbA1c values and heterozygous for HbD, a bias ranging be-

tween –0.6% and –0.4% (5-6 mmol/L) was found. On the

Sebia Capillarys 3 Tera analyzer, HbD and HbE were nicely

separated from the HbA0 peak Figure 3C and Figure 3D .

In comparison with the Sebia Capillarys 2 Flex Piercing, bias

was below 0.4% or 4 mmol/mol for all variants tested.

Stability of HbA1c in Capillary and Venous Blood

Samples

Three venous patient samples were stored at 2�C to

8 �C during 3 weeks. Analysis on the Bio-Rad D-100 instru-

ment showed a bias of less than 0.4% (<4 mmol/mol) up to

13 days. The chromatograms showed a continuously

increasing P3 peak with longer storage time. For the Sebia

Capillarys 3 Tera, the bias was less than 0.4% (4 mmol/mol)

up to 19 days. The electropherogram showed an increasing

“other HbA” peak with time but with no significant effect

on the quantification of HbA1c.

The room temperature stability of HbA1c in blood,

stored in the HCCS, was only performed with the Bio-Rad

D-100 analyzer, because the application was under develop-

ment on the Sebia system. Data showed stability of at least 7

days with a maximal bias below 0.15% or 2.5 mmol/mol.

Detection of Hemoglobin Variants

Both test instruments were able to detect the most import-

ant hemoglobin variants: HbS, HbD, HbE, and HbC (see

above). The presence of the hemoglobin variants was visually

evaluated on the chromatogram/electropherogram (Figures 3

and 4). For these variants, the Bio-Rad D-100 software identi-

fied the specific hemoglobin variant, based on retention time.

However, identification should be confirmed by other

Mean–1.0

–1.96 SD–4.7

+1.96 SD2.7

500 100 150 200

2

4

6

0

–2

–6

–4

–8

Hb

A1c

(mm

ol/

mo

l)—

Bio

-Rad

D-1

00 a

nd S

ebia

Cap

illar

ys 3

Ter

a

Mean HbA1c (mmol/mol) of Bio-Rad D-100 and Sebia Capillarys 3 Tera

0 20 40 60 80 100 120 140 160

20

40

60

80

100

120

140

160

0

Hb

A1c

(mm

ol/

mo

l)—

Seb

ia C

apill

arys

3 T

era

HbA1c (mmol/mol)—Bio-Rad D-100

A B

Figure 2 Passing-Bablok regression analysis (A) and Bland-Altman plot (B) of hemoglobin A1c (HbA1c) results (International

Federation of Clinical Chemistry units) using whole-blood samples obtained with the Bio-Rad D-100 in comparison with the

Sebia Capillarys 3 Tera.

AJCP / ORIGINAL ARTICLE

© American Society for Clinical Pathology Am J Clin Pathol 2016;146:67-77 73DOI: 10.1093/ajcp/aqw08173

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

methods. Less frequent variants or variant combinations in our

population, such as HbO-Arabe, HbSC, HbSS, HbCD, HbG

Philadelphia, HbS/HbG Philadelphia, HbJ-Toronto, and

HbEE, were detected in the chromatograms/electropherograms

and identified by the software on both test instruments. The

Sebia Capillarys 3 Tera was also able to detect HbA20 and

other not yet identified variants, due to a clear separation of the

peaks. In the HbA1c application, none of the tested instruments

was able to detect HbH or Hb Barts.

Workflow

Bio-Rad D-100

Startup time after shutdown of the system was 130 se-

conds. Results of two internal quality control (IQC) samples

were available after 9 minutes and 22 seconds. The first re-

sult of patient samples was available 2 minutes and 20 se-

conds after obtaining and verification of IQC results. The

last result of 150 patient samples was available after 1 hour

and 56 minutes. Of the 150 samples, 11 samples were rerun

because of a “too low area” error code. The instrument was

able to rerun automatically problematic samples. The whole

run of patient samples and IQC was completed after

128 minutes. Sample throughput after startup was 77 sam-

ples per hour.

Sebia Capillarys 3 Tera

Startup time after shutdown was 20 minutes and 14 se-

conds. ICQ results from all 12 capillaries were available

after 26 minutes and 30 seconds. The first patient results

were available 17 minutes later. The whole run of 150 pa-

tient samples (including IQC at the beginning and the end of

the run) was completed after 191 minutes. None of the sam-

ples had to be rerun. Sample throughput after startup was 62

samples per hour. This throughput might not be final since

the system was operating with preliminary software.

Discussion

In this study, we evaluated two beta versions of newly

launched HbA1c analyzers from Sebia (Capillarys 3 Tera)

and Bio-Rad. The two systems have been developed to meet

the current stringent performance criteria as well as NEN-

EN-ISO 15189:2012 (2012) criteria for traceability of cali-

bration data, reagents, and internal quality control batches.16

The Belgian Scientific Institute of Health (WIV-ISP)

lowered the CV acceptance criteria for individual methods

300280260240220200180160140120100806040200

HbA1c

HbA2

HbS

HbA05

OtherHbA

HbF orvariant

300280260240220200180160140120100806040200

HbA1c

HbC

HbA0

4

5

OtherHbA

300280260240220200180160140120100806040200

HbA1c

HbA2

HbD

HbA0

5

OtherHbA HbF or

variant

300280260240220200180160140120100806040200

HbA1c

HbA2

HbE

HbA0

5

OtherHbA

HbF orvariant

A B

C D

Figure 3 Hemoglobin A1c (HbA1c) electropherograms on the Sebia Capillarys 3 Tera for some variant hemoglobins showing

the resolution of the separation: HbS (A), HbC (B), HbD (C), and HbE (D).

Herpol et al / EVALUATION CAPILLARYS 3 TERA AND D-100

74 Am J Clin Pathol 2016;146:67-77 © American Society for Clinical Pathology74 DOI: 10.1093/ajcp/aqw081

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

in 20147 to minimize the contribution of assay imprecision

to the HbA1c values. Within-laboratory CV should be below

2% (NGSP units), even better below 1.4% (NGSP units).

College of America Pathologists surveys have shown

continuous improvement of between-laboratory CVs but

also that several laboratories do not fulfill the criteria.15,17

Both automated analyzers showed excellent trueness

(bias<0.3% or 3 mmol/mol) in the absence of hemoglobin

11.6

4

0.00 10.00 20.00 30.00 40.00

15

10

5

0

20

Hb

A1c

(% N

GS

P)

Time (s)

22.0

4

18.9

6 P

3

24.1

2

Hb

A0

28.4

3

14.0

Hb

A1c

11.5

2

8.32

Hb

F7.

14

4.91

Hb

A1a

/Hb

A1b

5.76

HbS

0.00 10.00 20.00 30.00 40.00

15

10

5

0

20

Hb

A1c

(% N

GS

P)

Time (s)

5.79 7.

148.

34

11.6

2

14.0

7

21.9

2

19.0

4

24.0

625

.91

Hb

A0

P3

Hb

A1a

/Hb

A1b

Hb

A1c

Hb

F

HbE

0.00 10.00 20.00 30.00 40.00

15

10

5

0

20

Hb

A1c

(% N

GS

P)

Time (s)

4.84

5.81 7.

158.

18

11.9

2

14.1

1

22.1

6

24.2

4

27.8

7

Hb

A0

P3

Hb

A1a

/Hb

A1b

Hb

A1c

Hb

F

HbD

11.6

4

0.00 10.00 20.00 30.00 40.00

15

10

5

0

20

Hb

A1c

(% N

GS

P)

Time (s)

4.88

7.16 8.31

14.0

9

18.8

9

24.0

4

30.1

5

Hb

A0

Hb

A1c

P3

Hb

A1b

Hb

F

HbC

A B

C D

Figure 4 Hemoglobin A1c (HbA1c) chromatograms on the Bio-Rad D-100 for some variant hemoglobins showing the resolution

of the separation: HbS (A), HbC (B), HbD (C), and HbE (D). The gray shaded peak is HbA1c.

AJCP / ORIGINAL ARTICLE

© American Society for Clinical Pathology Am J Clin Pathol 2016;146:67-77 75DOI: 10.1093/ajcp/aqw08175

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

variants. Except in the high HbA1c range, the Bio-Rad

D-100 system experienced a small but acceptable absolute

bias of –0.33% (or –3.27 mmol/mol). Both the Capillarys 3

Tera and the D-100 systems were not subjected to interfer-

ence by the labile HbA1c fraction (up to 10 g/L glucose),

carbamylated hemoglobin (up to 0.5 mmol/L), triglycerides

(up to 3,361 mg/dL), or bilirubin (up to 14 mg/dL).

Although the electropherograms/chromatograms showed

some profile alterations in the presence of extremely high

concentrations of potential interferences, the bias remained

within quality criteria that were put forward. Only at very

high levels of interfering substances, such as 50 g/L glucose

and 1 mmol/L KCN, a bias could be detected with the Bio-

Rad system. Since such high concentrations of interfering

peaks are not physiologic, this bias problem can be

neglected.

HbA1c values are used not only for follow-up of gly-

cemic control in diabetic patients but also for diagnosis. The

ADA has set a cutoff for diagnosis at an HbA1c level of

6.5% (48 mmol/mol).18,19 The diagnosis should be con-

firmed by a repeat test. It has been shown that diagnosis

based on HbA1c levels identifies other patients than diagno-

sis based on oral glucose tolerance test results.17,20

Individual susceptibility to glycation or variation in HbA1c

between different ethnic groups is poorly understood but

relevant when using HbA1c for the diagnosis of diabetes.20-22

In addition, a potential bias of the method used could lead to

over- or underdiagnosis of diabetes.

In clinical practice of follow-up of diabetic patients, an

absolute HbA1c difference of 0.5% between two consecutive

patient samples is considered a significant change in dia-

betes control.17,18 In this respect, excellent reproducibility is

mandatory.

As clinical care evolves toward near-patient diagnosis,

POC methods are developed as well as high-throughput

automates with STAT function. A recent study by Lenters-

Westra and Slingerland8 of seven HbA1c POC devices

showed that three instruments did not meet the analytical

performance criteria. In addition, improper handling

and analysis by nonlaboratory staff could contribute to er-

roneous HbA1c results. Therefore, these instruments should

be introduced in clinical practice with caution and supervi-

sion of the laboratory. As an alternative, high-throughput

automated instruments are now able to provide results

within a short time frame. Especially in hospital settings,

they represent the best option for accurate and precise

HbA1c results available within the time frame of a patient

visit. For a relatively small batch of samples, the sample

throughput of the Bio-Rad D-100 is superior to the Sebia

Capillarys 3 Tera, because of its relative short startup time.

As the sample batch increases, the Sebia analyzer

gains in analysis speed, since 12 samples are analyzed in

parallel and wash/injection/analysis steps can be performed

simultaneously.

As the increasingly stringent analytical performance

criteria are met by most of the automated instruments, non-

glucose-related biological variation becomes more of an

issue when using HbA1c as a diagnostic criterion. All condi-

tions that affect erythrocyte life span may cause falsely low

or elevated HbA1c levels.18,23,24 Shortened erythrocyte life

span—in conditions such as renal anemia, chronic liver dis-

ease, chronic or hemolytic anemia, acute blood loss, or re-

cent transfusion—results in falsely low HbA1c.18,23,24

Especially in iron deficiency anemia, HbA1c should be inter-

preted with caution. Prolonged erythrocyte life span contrib-

utes to falsely high HbA1c in iron deficiency, whereas

altered glycation is found—and therefore a decrease in

HbA1c levels—in anemic patients after iron supplementa-

tion.18,25,26 In addition, hemoglobin variants could interfere

with HbA1c analysis in variable ways depending on the

method and the effect of the variant on erythrocyte life

span.18,23,24 In this study, the ability of the analyzers to de-

tect the presence of a hemoglobin variant was evaluated.

HbS, HbD, HbE, Hb C, HbO-Arabe, HbSC, HbSS, HbCD,

HbG Philadelphia, HbS/HbG Philadelphia, HbJ-Toronto,

and HbEE were detected, while HbH and Hb Barts were

not. The identification of the variants should be confirmed

by other methods. The detection of hemoglobin variants is

an advantage of chromatographic or electrophoretic analysis

techniques.

To improve patient care, home sampling by capillary

sampling and/or dried blood spots on filter paper is an op-

tion that could be interesting for patients with poor meta-

bolic control or for patients reluctant to have regular venous

blood sampling such as children, adolescents, and the eld-

erly. A previous study by our group on the Sebia Capillarys

2 Flex Piercing showed excellent agreement between ven-

ous and capillary blood samples as well as stability of the

capillary sample for at least 5 days at room temperature.13

At the time of testing, the D-100 was ready to analyze capil-

lary blood samples. A preliminary study showed a room

temperature stability of HbA1c in the HCCS of 7 days. The

application was still in development for the Sebia Capillarys

3 Tera, and therefore this stability study was only evaluated

with the Bio-Rad instrument.

In conclusion, in terms of quality, performance, and

throughput, the Bio-Rad D-100 and the Sebia Capillarys 3

Tera instruments perform similarly and meet the stringent

criteria of the WIV-ISP for the quantification of HbA1c in

whole blood.

Disclosure: Bio-Rad Laboratories Belgium and Sebia

Benelux provided the D-100 and Capillarys 3 Tera ana-

lyzers and reagents used in this evaluation.

Herpol et al / EVALUATION CAPILLARYS 3 TERA AND D-100

76 Am J Clin Pathol 2016;146:67-77 © American Society for Clinical Pathology76 DOI: 10.1093/ajcp/aqw081

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022

Corresponding author: Ilse Weets, Dept of Clinical Chemistry and

Radio-Immunology, Universitair Ziekenhuis, Laarbeeklaan 101,

Brussels, Belgium; [email protected].

References

1. International Diabetes Federation. IDF atlas. 2014. http://www.idf.org/diabetesatlas. Accessed September 15, 2015.

2. The Diabetes Control and Complications Trial ResearchGroup. The effect of intensive treatment of diabetes on thedevelopment and progression of long-term complications ininsulin-dependent diabetes mellitus. N Engl J Med.1993;329:977-986.

3. American Diabetes Association. Standards of medical carein diabetes—2010. Diabetes Care. 2010;33(suppl 1):S11-S61.

4. Hoelzel W, Weykamp C, Jeppsson J-O, et al. IFCC referencesystem for measurement of hemoglobin A1c in human bloodand the national standardization schemes in the UnitedStates, Japan, and Sweden: a method-comparison study. ClinChem. 2004;50:166-174.

5. Little RR, Rohlfing CL, Sacks DB; NationalGlycohemoglobin Standardization Program (NGSP)Steering Committee. Status of hemoglobin A1c measure-ment and goals for improvement: from chaos to order for im-proving diabetes care. Clin Chem. 2011;57:205-214.

6. Chapelle J-P, Teixeira J, Maisin D, et al. Multicentre evalu-ation of the Tosoh HbA1c G8 analyser. Clin Chem Lab Med.2010;48:365-371.

7. Wetenschappelijk Instituut Volksgezondheid. WIV informa-tion for analytical goals for measurement of HbA1c. https://www.wiv-isp.be/qml/activities/external_quality/rapports/_down/hba1c/2014/JR-2014.pdf. Accessed June 28, 2015.

8. Lenters-Westra E, Slingerland RJ. Three of 7 hemoglobinA1c point-of-care instruments do not meet generally ac-cepted analytical performance criteria. Clin Chem.2014;60:1062-1072.

9. Sanchez-Mora C, Rodriguez-Oliva MS, Fern�andez-Riejos P,et al. Evaluation of two HbA1c point-of-care analyzers. ClinChem Lab Med. 2011;49:653-657.

10. Clinical and Laboratory Standards Institute (CLSI).Evaluation of Precision Performance of QuantitativeMeasurement Methods: Approved Guideline—Second Edition.Wayne, PA: CLSI; 2004. CLSI document EP05-A2.

11. Clinical and Laboratory Standards Institute (CLSI). MethodComparison and Bias Estimation Using Patient Samples:Approved Guideline—Second Edition (Interim Revision).Wayne, PA: CLSI; 2010. CLSI document EP09-A2-IR.

12. Clinical and Laboratory Standards Institute (CLSI).Evaluation of the Linearity of Quantitative MeasurementProcedures: A Statistical Approach; Approved Guideline.Wayne, PA: CLSI; 2003. CLSI document EP6-A.

13. Heylen O, Van Neyghem S, Exterbille S, et al. Evaluation ofthe Sebia CAPILLARYS 2 flex piercing for the measurementof HbA1c on venous and capillary blood samples. Am J ClinPathol. 2014;141:867-877.

14. Clinical and Laboratory Standards Institute (CLSI).Interference Testing in Clinical Chemistry: ApprovedGuideline—Second Edition. CLSI; 2005. Wayne, PA: CLSIdocument EP7-A2.

15. College of America Pathologists. CAP survey. http://www.ngsp.org/CAP/CAP14a.pdf. Accessed September 18, 2015.

16. NEN-EN-ISO 15189:2012. https://www.iso.org/obp/ui/#iso:std:iso:15189:ed-3:v2:en. Accessed September 16, 2015.

17. Little RR, Rohlfing CL. The long and winding road to opti-mal HbA1c measurement. Clin Chim Acta. 2013;418:63-71.

18. Weykamp C. HbA1c: a review of analytical and clinical as-pects. Ann Lab Med. 2013;33:393-400.

19. American Diabetes Association. Diagnosis and classificationof diabetes mellitus. Diabetes Care. 2014;37(suppl 1):S81-S90.

20. Bonora E, Tuomilehto J. The pros and cons of diagnosingdiabetes with A1c. Diabetes Care. 2011;34(suppl2):S184-S190.

21. Hempe JM, Gomez R, McCarter RJ, et al. High and lowhemoglobin glycation phenotypes in type 1 diabetes: a chal-lenge for interpretation of glycemic control. J DiabetesComplications. 2002;16:313-320.

22. Khera PK, Joiner CH, Carruthers A, et al. Evidence for inter-individual heterogeneity in the glucose gradient across thehuman red blood cell membrane and its relationship tohemoglobin glycation. Diabetes. 2008;57:2445-2452.

23. English E, Idris I, Smith G, et al. The effect of anaemia andabnormalities of erythrocyte indices on HbA1c analysis: a sys-tematic review. Diabetologia. 2015;58:1409-1421.

24. Higgins T. HbA(1c)—an analyte of increasing importance.Clin Biochem. 2012;45:1038-1045.

25. Sundaram RC, Selvaraj N, Vijayan G, et al. Increased plasmamalondialdehyde and fructosamine in iron deficiency an-emia: effect of treatment. Biomed Pharmacother.2007;61:682-685.

26. Tarim €O, Kucukerdogan A, Gunay €U, et al. Effects of irondeficiency anemia on hemoglobin A1c in type 1 diabetes mel-litus. Pediatr Int. 1999;41:357-362.

AJCP / ORIGINAL ARTICLE

© American Society for Clinical Pathology Am J Clin Pathol 2016;146:67-77 77DOI: 10.1093/ajcp/aqw08177

Dow

nloaded from https://academ

ic.oup.com/ajcp/article/146/1/67/1730623 by guest on 31 August 2022