Abstract
Reliable measurement of hemoglobin A1c (HbA1c) has great importance in the diagnosis and monitoring of diabetes mellitus. The aim of the present study was to compare the performance parameters of the three common methods of HbA1c assay, including the Roche, Sebia and TOSOH G8 systems. We studied 120 patients referred to a clinical laboratory for HbA1c assay. The blood samples were analyzed with the Roche, Sebia and TOSOH G8 systems based on immunoassay, capillary electrophoresis, and ion-exchange chromatography techniques, respectively. The Spearman and the Passing-Bablok regression,as well as the Bland-Altman plots, were used to compare these methods. For each assay, the patients’ classification was evaluated at the three cut-points of 6.5, 7, and 8% and the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the methods were estimated. Our results showed that there were good correlations and agreement between the methods. We found a mean difference of 0.07% for the TOSOH G8 vs. Roche, 0.06% for the TOSOH G8 vs. Sebia and − 0.01% for the Roche vs. Sebia. The methods represented very low bias, indicating the good accuracy of the results. The sensitivity and specificity of the methods were comparable as well. The three methods also performed similarly in the classification of patients at the proposed cut-off points. Based on our results, the Roche, Sebia and TOSOH G8 systems showed a very high level of agreement with comparable performance parameters and yielded similar and accurate classification of diabetic patients. Therefore, these methods can be used interchangeably.
Keywords: HbA1c, Sebia, TOSOH G8, Roche, Misclassification, Cut-offs
Introduction
Glycosylated hemoglobin (Hb) is composed of HbA1a, HbA1b, and HbA1c [1]. Hemoglobin A1c (HbA1c) is the main fraction of hemoglobin which is formed by a non-enzymatic reaction of glucose with the N-terminal valine residue of the HbA β chain [2, 3]. Due to the direct correlation between blood HbA1c and plasma glucose level over the past 2–3 months, it is used as an indicator of diabetes control in this period [2, 4]. HbA1c is considered as the gold standard of diabetes assessment. This marker is associated with the risk of long-term complications of type 1 and type 2 diabetes [5]. Over the past several years, the American Diabetes Association (ADA) and other major organizations confirmed the use of HbA1c for the screening of diabetes and proposed 6.5% as a diagnostic cut-off [3]. Reducing the concentration of HbA1c to 7% shows a significant reduction in the rate of cardiovascular complications of diabetes and it is recommended that the level of HbA1c in non-pregnant subjects should be kept below 7% (more precisely 6.5%) and 8% in people with specific disease conditions [6, 7]. There are several methods with different test principles for HbA1c analysis, such as capillary electrophoresis, ion-exchange high-performance liquid chromatography (HPLC), immunoassay and enzymatic assays [2, 4]. HbA1c assay is a high request test of the clinical laboratory; therefore, it requires a precision instrument to measure. Because HbA1c is of high clinical significance, a great standardization has been developed to coordinate the HbA1c measurement globally [4]. Different methods show different characteristics and performance. Over the past years, plenty of efforts has been made by the National Glycohemoglobin Standardization Program (NGSP) and the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) to standardize HbA1c measurement [3]. These methods display different features and performances, however, the comparability and overall quality of them are improved with respect to tracking with an acceptable international reference method [5].
This study aimed to compare the performance parameters of HbA1c measurement with three common assays systems, i.e., TOSOH G8, Cobas c 311 and Capillarys 2 Flex Piercin, in routine clinical laboratory work.
Study design
This study was performed on the blood samples provided from the leftover of 120 patient samples that were referred to the laboratory for the HbA1c test. In the fasting state, blood samples were collected from the patients in an anticoagulant tube containing K2-Ethylenediaminetetraacetic acid (EDTA). Then, HbA1c level of the samples was measured using three assay systems with different test basis including Tosoh G8 Analyzer (Tosoh Bioscience, Tokyo, Japan), Cobas c 311 (Roche Diagnostics, Germany) and Capillarys 2 Flex Piercing (Capillarys 2FP, Sebia, France). To assure a uniform distribution of samples in the measurement range and to evaluate the performance of the assay methods over different HbA1c ranges, samples with low, normal and high levels of HbA1c were selected. Samples were evaluated on the same day with all the three methods.
Methods
HbA1c level was measured using the following common systems available for routine laboratory work. TOSOH G8 Analyzer for HbA1c assay is based on ion-exchange chromatography using a TSKgel G8 Variant HSi Column. In this instrument, the elution carries out based on ionic interactions differences between the cation exchange group on column resin and the hemoglobin constituents and detection carried out at dual wavelength at 415 nm and 510 nm. In this method, HbA1c is assayed with high precision (less than 2% CVs) [2].
The Roche Cobas c 311 instrument is based on a competitive turbidimetric inhibition immunoassay in hemolysate. In the first step of this method, HbA1c in the blood sample reacts with anti-HbA1c antibody and produce soluble antigen-antibody complexes. Since there is just one specific site on the HbA1c molecule, the formation of insoluble complexes dose not occur. In the second step by adding polyhapten, an insoluble antibody-polyhapten complex is formed that is measured by turbidimetry [2].
Capillarys 2 Flex Piercing® analyzer (Sebia, France) is a HbA1c measurement method that is based on the separation of the HbA1c fractions through capillary electrophoresis. The separation of hemoglobin fractions conducted using silica capillaries affecting by electrophoretic mobility and electroosmotic flow [5].
All the instruments in this study, as well as associated reagents applied for analysis, were used following the instructions of the respective manufacturer and all the systems reported the HbA1c level as a unit percentage. For TOSOH G8, we used variant elution buffers with lot HB107D/HB205D/HB308D. For the Sebia we used a buffer (lot 15,108/01), a hemolysing solution (lot 06098101), and a wash solution (lot 20,047/80). For the Roche Cobas c 311, an A1-W solution (lot number 00253327) and a HB-W solution (lot number 00253327). HbA1C quality controls were evaluated by analyzing internal quality controls along with the samples (at low and high levels). All reagents and controls supplied by the corresponding manufacturers and purchased from Fardavar Azma Iranian Company (Fard Azma CO, Iran).
Precision of methods
The imprecision of all three methods determined using two samples at low (5%) and high (8%) levels. The sample of each level was divided into five aliquots and stored at −20 °C. One aliquot of each level was running three times daily for five consecutive days. Within- run precision of the methods was assessed as described by the Clinical and Laboratory Standards Institute (CLSI) document EP05-A2 [8].
Statistical analysis
Data analysis was performed using SPSS (version 22) and MedCalc version 15 (MedCalc Software, Ostend, Belgium). Quantitative variables were shown as the mean and standard deviation of data. Frequency and percentage were used to describe the data for qualitative variables.
The agreements of the methods were evaluated by the Spearman correlation analysis and the Passing-Bablok regression analysis. The differences between any two methods were presented as a Bland-Altman plot. The confidence intervals (95%) were determined and a p < 0.05 was considered as significant.
Results
The mean HbA1c levels in the patients were 7.299 ± 2.22%, 7.239 ± 2.28% and 7.227 ± 2.23% with the TOSOH G8, Sebia and Roche methods, respectively. The HbA1c levels of patients are represented according to the range of HbA1c levels (< 5.7 normal, 6.4–5.7 pre-diabetic and > 6.4 diabetic) by all three methods in Table 1.
Table 1.
Distributions of HbA1c concentrations assessed with 3 methods in 120 patients
| Method | HbA1c Cut-offs | ||
|---|---|---|---|
| < 5.7 | 5.7-6.4 | > 6.4 | |
| TOSOH G8 | 34(28.3) | 35(29.2) | 34(28.3) |
| Sebia | 17(14.2) | 19(15.8) | 18(15) |
| Roche | 69(57.5) | 66(55) | 68(56.7) |
Data are number of patients and percentages for each method
The imprecision of all studied methods, as presented in Table 2., was lower than 2%.
Table 2.
Within-run and total precision
| Method | Imprecision | |||
|---|---|---|---|---|
| CV Within-run (%) | CV Total (%) | |||
| Low | High | Low | High | |
| TOSOH G8 | 0.56 | 0.41 | 0.91 | 0.57 |
| Sebia | 1.32 | 1.1 | 1.42 | 1.12 |
| Roche | 1.1 | 0.72 | 0.99 | 1.18 |
The Spearman correlation analysis showed that there was a significant linear correlation between the results of these methods (Fig. 1).
Fig. 1.
Comparative evaluation of the methods using the Passing-Bablok regression analysis and the Spearman Correlation analysis. Solid blue lines represent the regression line, red dashed line represent 95% CI, and dotted line represent the line of equality. *P < 0.001
Method comparison
TOSOH G8 vs. Roche
The Passing-Bablok regression analysis of the obtained data showed a good agreement between TOSOH G8 and Roche method. The slope of 1.0 with a 95% CI of 0.973–1.02, and an intercept of −0.1 with a 95% CI of −0.211 − 0.119. Furthermore, The Spearman correlation coefficient (r) showed these methods were highly correlated r = 0.992, p < 0.001(Fig. 1).
The Bland-Altman plot revealed a mean difference of 0.07%. This means that the TOSOH G8 method measures on average 0.07% more than the Roche method. Although this bias is not significant, it seems to be related to lower concentrations of HbA1c. Limits of agreement ranged from −0.53 to +0.67 for the constant difference that means an individual result of the TOSOH G8 method may differ by as much as −0.53 to +0.67 from the Roche result. In the percentage plot, relative bias was −1.1% and limits of agreement ranged from −6.7 to +8.9% for the proportional difference. This bias is not significant and appears caused by lower measurements (Fig. 2).
Fig. 2.
Bland-Altman plots. a Solid blue line shows the mean difference, while the dashed lines show the mean difference ± 1.96 SD. b Dashed lines show the distance of the mean difference from zero (with the blue lines indicating the 95% CI) as percentages. Red dotted line (with the blue lines indicating the 95% CI)
TOSOH G8 vs. Sebia
The Passing-Bablok regression analysis of the data obtained with the TOSOH G8 and Sebia methods showed a slope of 1.0 with a 95% CI of 0.985–1.000, and an intercept of 0.1 with 95%CI of 0.098–0.196. The Spearman correlation coefficient between the methods was r = 0.992 (95% CI: 0.995–0.989, p < 0.001) (Fig. 1).
The Bland-Altman plot showed a mean absolute difference of 0.06% HbA1c (Fig. 2). This means that the TOSOH G8 method measures on average 0.06% more than the Sebia method. Although this bias is not significant, it seems to be related to lower concentrations of HbA1c. Limits of agreement (±1.96 SD of differences) ranged from −0.55 to +0.67 for the constant difference. Relative Bias was 1.0% and limits of agreement ranged from −6.7 to +8.7% for the proportional difference (Fig. 2).
Roche vs. Sebia
The Passing-Bablok regression analysis of the data obtained with the Roche vs Sebia showed a slope of +1.0 (95%CI: 0.969–1.0), and an intercept of 0.000 (95%CI: −0.00-0.19). The Spearman correlation coefficient was r = 0.991 (95% CI: 0.989–0.995, p < 0.001) (Fig. 1).
The Bland-Altman plot represented a mean difference of −0.01% HbA1c. Limits of agreement ranged from −0.56 to +0.53 for the constant difference. This bias is not significant. Relative Bias was −0.1% and limits of agreement ranged from −7.3 to +7.2% for the proportional difference (Fig. 2).
Sensitivity and specificity
The results of the patients were classified according to three cut-off values of 6.5, 7 and 8%. Each time a method was considered as the reference method and two other methods were compared with it. The results of the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) analysis are shown in Table 3.
Table 3.
Sensitivity, Specificity, PPV, and NPV of the three methods at different Cut points compared to the assumed reference method.PPV, positive predictive value; NPV, negative predictive value
| TOSOH G8 | |||||
| Cut off | Method | SN | SP | PPV | NPV |
| 6.5 | Sebia |
92.75 (83.9–97.6) |
96.08 (86.5–99.5) |
97 (89.5–99.6) |
90.7 (79.7–96.6) |
| Roche |
95.65 (87.8–99.1) |
96.08 (86.5–99.5) |
97.1 (89.8–99.6) |
94.2 (84.1–98.8) |
|
| 7 | Sebia |
94.74 (85.4–98.9) |
98.41 (91.5–100) |
98.2 (90.3–100) |
95.4 (87.1–99) |
| Roche |
94.7 (85.4–98.9) |
100 (94.3–100) |
100 (93.4–100) |
95.5 (87.3–99.1) |
|
| 8 | Sebia |
92.11 (87.6–98.3) |
100 (95.6–100) |
100 (90–100) |
96.5 (90–99.3) |
| Roche |
86.84 (71.9–95.6) |
100 (95.6–100) |
100 (89.4–100) |
94.3 (87.1–98.1) |
|
| Roche | |||||
| Cut off | methods | SN | SP | PPV | NPV |
| 6.5 | TOSOH G8 |
97.06 (89.8–99.6) |
94.23 (84.1–98.8) |
95.7 (87.8–99.1) |
96.1 (86.5–99.5) |
| Sebia |
95.59 (87.6–99.1) |
98.08 (89.7–100) |
98.5 (91.8–100) |
94.4 (84.6–98.8) |
|
| 7 | TOSOH G8 |
100 (93.4–100) |
95.45 (87.3–99.1) |
94.7 (85.4–98.9) |
100 (94.3–100) |
| Sebia |
98.15 (90.1–100) |
96.97 (89.5–99.6) |
96.4 (87.5–99.6) |
98.5 (91.7–100) |
|
| 8 | TOSOH G8 |
100 (89.4–100) |
94.25 (87.1–98.1) |
86.8 (71.9–95.6) |
100 (95.6–100) |
| Sebia |
96.97 (84.2–99.9) |
96.55 (90.3–99.3) |
91.4 (76.9–98.2) |
98.8 (93.6–100) |
|
| Sebia | |||||
| Cut off | methods | SN | SP | PPV | NPV |
| 6.5 | TOSOH G8 |
96.97 (89.5–99.6) |
90.74 (79.7–96.9) |
92.8 (83.9–97.6) |
96.1 (86.5–99.5) |
| Roche |
98.48 (91.8–100) |
94.44 (84.6–98.8) |
95.6 (87.6–99.1) |
98.1 (89.7–100) |
|
| 7 | TOSOH G8 |
98.18 (90.3–100) |
95.38 (97.1–99) |
94.7 (85.4–98.9) |
98.4 (91.5–100) |
| Roche |
96.36 (87.5–99.6) |
98.46 (91.7–100) |
98.1 (90.1–100) |
97 (89.5–99.6) |
|
| 8 | TOSOH G8 |
100 (90–100) |
96.47 (90–99.3) |
92.1 (78.6–98.3) |
100 (95.6–100) |
| Roche |
91.43 (76.9–98.2) |
98.82 (93.6–100) |
97 (84.2–99.9) |
96.6 (90.3–99.3) |
|
If the TOSOH G8 method was considered as the reference method, the Roche method provided better performance than the Sebia method at the cut-off of 6.5%, however, at the cut-off of 7.0%, they showed equal sensitivity, but the specificity and PPV of the Roche method were better than the Sebia method. At the cut-off of 8.0%, the Sebia method represented better performance than the Roche method. The performance of the methods changed with the increase in the cut-off value.
If the Roche method was considered as the reference method, the sensitivity and NPV for the TOSOH G8 method were higher than the Sebia method, but the specificity and PPV of the Sebia method were better than the TOSOH G8 method, for all the cut-off points.
If the Sebia method was considered as the reference method, the Roche method displayed better performance than the TOSOH G8 method at the cut-off of 6.5%. In the cut-off values of 7% and 8%, the sensitivity and NPV of the TOSOH G8 method and the specificity and PPV of the Roche method were higher than the other method. Altogether, the differences were very small between the three methods.
Misclassification
Similarly, the results of the patients were classified according to three cut-off values of 6.5, 7 and 8%. At each cut-off point, the patients who were wrongly classified were determined (Table 3).
At the cut-off <6.5%, in the case of the Roche or Sebia methods as the reference, the TOSOH G8 method correctly categorized 96.1% of patients (49 out of 51 patients). At the cut-off ≥6.5%, if the Roche method was considered as the reference method, the Sebia method correctly categorized 98.5% of patients (65 out of 66 patients). In the cut-off value of 6%, the highest agreement between the Sebia and Roche methods was found in patient classification (96.7%).
At the cut-off <7.0%, in comparison to the Roche method as the reference, the TOSOH G8 method correctly categorizes all the patients and the two other comparisons, Sebia vs. TOSOH G8 and Roche vs. Sebia represented 98.5% (62 out of 63 patients) and 98.4% (64 out of 65 patients) agreement, respectively. At the cut-off ≥7.0%, when the Roche method was the reference method, the Sebia method correctly categorized 96.4% of patients (53 out of 55 patients), and the two other evaluations, the Sebia vs. TOSOH G8 and the Roche vs. TOSOH G8, were both equally categorized 94.7% of patients correctly (54 out of 57 patients). The highest agreement level between these methods at the cut-off point of 7.0 was 97.5%.
At the cut-off <8.0%, in case of the Roche or Sebia methods as the reference method, the TOSOH G8 method correctly categorized all the patients (82 out of 82 the patients) and in comparison, of the Roche vs. Sebia methods, 98.8% patients were correctly classified (84 out of 85 patients). At the cut-off ≥8.0%, when the Sebia method was the reference, the TOSOH G8 method correctly categorized 92.1% of the patients (35 out of 38 patients), and two other evaluations Roche vs. Sebia and Roche vs. TOSOH G8 methods categorized the patients correctly up to 91.4% (32 out of 35 patients) and 86.81% (33 out of 38 patients), respectively. The highest agreement level between the Sebia vs. TOSOH G8 methods at the cut-off value ≥8.0 was 97.5%.
Discussion
It is essential to measure HbA1c with an accurate and precise method to diagnose and monitor diabetes mellitus [9]. The precision of methods and agreement with a reference method are critical parameters in clinical decision-making and assuring the quality of clinical trials [10].
In this study, three common methods for measuring HbA1c were evaluated and it found that all the methods showed acceptable accuracy and precision. The comparative analysis showed that the methods had good agreements with a narrow scattering around the regression lines.
The results of patients with the test methods compared to the reference method showed high sensitivity and specificity in three cut-off points of 6.5%, 7%, and 8%. The PPV and NPV of methods were more than 86.8% and 90.7%, respectively, which indicates that the results above and below the level of clinical decision-making are very reliable.
The Passing-Bablock regression analysis on the HbA1c measurement results of the test methods did not show statistically significant differences between the methods. Although all the three methods showed a linear regression, regression of the TOSOH G8 -Roche and the Roche-Sebia methods encompassed zero that indicated no significant difference. But regression analysis of the TOSOH G8 - Sebia methods showed minor differences as it did not encompass zero. However, the difference was very small with no relative systematic difference which is not clinically relevant. Similar to our results, Hanita et al. demonstrated that the TINA-QUANT Roche COBAS 6000 instrument showed acceptable precision and good recovery compared with the TOSOH G8 method [11]. Other studies have also shown that the HbA1c results obtained with the TOSOH G8 and Sebia methods have a good correlation [1, 5, 12].
In this study, we used the Bland-Altman plot to compare the results of methods and to evaluate their agreement. Although there was no systematic difference in the results of the Bland-Altman plots, the plots showed a positive bias for the TOSOH G8 vs. Roche and the TOSOH G8 vs. Sebia and a negative bias for the Roche vs. Sebia. The clinically acceptable bias for HbA1c measurement was suggested as desirable ≤ ± 1.5% and as appropriate ≤ ± 0.8% [10, 13] which were within the 95% limits of agreement obtained in this study. Despite the minor difference, statistical analyses proved that these methods were in agreement and provide comparable HbA1c results and so the methods are interchangeable. The minor differences may be due to imprecision in the operation of the techniques.
Similar to Wu et al., we showed the between methods HbA1c values were in good agreement using the Bland Altman plots [3]. In the Klingenberg et al. study, the HPLC-G7 method showed slightly higher results, probably due to the need to factorize the results of the TOSOH G7 and TOSOH G8 multiplied by 0.955 (as recommended by the EQA program) [2]. Aksungar et al. reported that the HbA1c results of the Capillary 2 Flex-Piercing instrument and the Roche-Cobas Integra 400 were fully comparable which provide reliable means of control and diagnosis of diabetes mellitus [14]. Rollborn et al. suggested that the Capillaries 3 Tera instrument provided a high assay capacity for HbA1c and showed good precision and agreement with the Roche 6000 instrument [4]. Finally, the HbA1c results of these methods showed that they can similarly and accurately classify the patients considering the proposed cut-off points.
This study bears some limitations as we did not take the influence of hemoglobin variants (hemoglobinopathies, carbamylated and labile hemoglobin) into account. The role of other interferences such as bilirubin, urea, ascorbic acid, rheumatoid factor and altered half-life of red blood cells (in case of hemolytic anemia, renal failure, etc.) were not investigated as well. More studies are required to investigate the influence of these factors involved in HbA1c measurement.
Conclusion
In summary, the Roche, Sebia and TOSOH G8 systems commonly used to measure HbA1c provided comparable and reliable results. The analytical performance of the three methods accord with the desirable performance standards for HbA1c analysis required for clinical laboratories for routine practice and the methods could correctly classify the diabetic patients. The agreement and interchangeability of these methods is an advantage and is in favor of the economic measures of laboratories.
Acknowledgements
The present study was supported by a grant from the Hyperlipidemia (HLRC-9208) ResearchCenter, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran. We are grateful to the staffs of Dr. Jalali laboratory for their kind cooperation.
Abbreviations
- HbA1c
Hemoglobin A1c
- SD
Standard Deviation
- NPV
Negative Predictive Value
- HPLC
Ion exchange High-performance Liquid Chromatography
- NGSP
National Glycohemoglobin Standardization Program
- IFCC
International Federation of Clinical Chemistry and Laboratory Medicine
- K2-EDTA
K2-Ethylenediaminetetraacetic acid
- ADA
American Diabetes Association
- PPV
Positive Predictive Value
- CI
Confidence Interval
Compliance with ethical standards
Conflict of interest
The authors declare that there is no conflict of interest.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Tesija Kuna A, Dukic K, Nikolac Gabaj N, et al. Comparison of enzymatic assay for HbA1c measurement (Abbott Architect) with capillary electrophoresis (Sebia Minicap Flex piercing analyser) Lab Med. 2018;49(3):231–238. doi: 10.1093/labmed/lmx090. [DOI] [PubMed] [Google Scholar]
- 2.Klingenberg O, Furuset T, Hestbråten CR, Hallberg MH, Steiro A, Orset IR, Berg JP. HbA1c analysis by capillary electrophoresis–comparison with chromatography and an immunological method. Scand J Clin Lab Invest. 2017;77(6):458–464. doi: 10.1080/00365513.2017.1338747. [DOI] [PubMed] [Google Scholar]
- 3.Wu X, Chao Y, Wan Z, Wang Y, Ma Y, Ke P, Wu X, Xu J, Zhuang J, Huang X. A comparative evaluation of the analytical performances of Capillarys 2 flex piercing, Tosoh HLC-723 G8, premier Hb9210, and Roche Cobas c501 Tina-quant gen 2 analyzers for HbA1c determination. Biochemia medica: Biochem Medica. 2016;26(3):353–364. doi: 10.11613/BM.2016.039. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Rollborn N, Åkerfeldt T, Nordin G, Xu XY, Mandic-Havelka A, Hansson LO, Larsson A. Analysis of HbA1c on an automated multicapillary zone electrophoresis system. Scand J Clin Lab Invest. 2017;77(1):15–18. doi: 10.1080/00365513.2016.1238507. [DOI] [PubMed] [Google Scholar]
- 5.Jaisson S, Leroy N, Meurice J, Guillard E, Gillery P. First evaluation of Capillarys 2 flex piercing®(Sebia) as a new analyzer for HbA1c assay by capillary electrophoresis. Clin Chem Lab Med. 2012;50(10):1769–1775. doi: 10.1515/cclm-2012-0017. [DOI] [PubMed] [Google Scholar]
- 6.Association AD Standards of medical care in diabetes–2014. Diabetes Care. 2014;37(Supplement 1):S14–S80. doi: 10.2337/dc14-S014. [DOI] [PubMed] [Google Scholar]
- 7.Razi F, Esfahani EN, Farzami MR, et al. Effect of the different assays of HbA1c on diabetic patients monitoring. J Diabetes Metab Disord. 2015;14(1):65. [DOI] [PMC free article] [PubMed]
- 8.Clinical and Laboratory Standards Institute. Evaluation of precision performance of quantitative measurement methods; approved guideline. 2nd ed, document EP05-A2. Wayne: CLSI;2004.
- 9.Heylen O, Van Neyghem S, Exterbille S, et al. Evaluation of the Sebia CAPILLARYS 2 flex piercing for the measurement of HbA1c on venous and capillary blood samples. Am J Clin Pathol. 2014;141(6):867–877. doi: 10.1309/AJCPRU5QC2JBANSV. [DOI] [PubMed] [Google Scholar]
- 10.Genc S, Gurdol F, Kanmaz-Ozer M, et al. The analytical performances of four different glycated hemoglobin methods. Med Chem. 2014;4(501):5. [Google Scholar]
- 11.Hanita O, Hanisah AH, Subashini CT, et al. The performance evaluation of Tina-Quant Roche Cobas 6000 in the measurement of HbA1c. Pathology. 2009;41:69. [Google Scholar]
- 12.Dessi M, Pieri M, Pignalosa S, Martino FG, Zenobi R. Performances of capillary electrophoresis and HPLC methods in HbA1c determination: diagnostic accuracy in HbS and HbD-Iran variants' presence. J Clin Lab Anal. 2015;29(1):57–60. doi: 10.1002/jcla.21728. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Rohlfing CL, Connolly SM, England JD, Hanson SE, Moellering CM, Bachelder JR, Little RR. The effect of elevated fetal hemoglobin on hemoglobin A1c results: five common hemoglobin A1c methods compared with the IFCC reference method. Am J Clin Pathol. 2008;129(5):811–814. doi: 10.1309/YFVTUD0GHJF7D16H. [DOI] [PubMed] [Google Scholar]
- 14.Aksungar FB, Serteser M, Coşkun A, Ünsal İ. A comparison between turbidimetric inhibition immunoassay and capillary electrophoresis in glycated hemoglobin (HbA1c) measurement. Clin Chem Lab Med. 2013;51(8):e191–e193. doi: 10.1515/cclm-2013-0033. [DOI] [PubMed] [Google Scholar]