Good Agreement Between Hba1c Analyzed Using Capillary Electrophoresis, HPLC, Immunological and Enzymatic Methods

Purpose: Hemoglobin A1c (HbA1c) is an essential marker for assessment of glycemic control in diabetes patients. The aim of this study was to evaluate the agreement between different HbA1c methods. 
Methodology: We used blood samples to compare HbA1c results analyzed with Capillarys 3 Tera, Roche Tina-Quant HbA1c Gen 3, BioRad Variant II Turbo (3 sites), Mono S® and Abbott Architect enzymatic method. The comparisons were made as paired instrument comparisons with Capillarys 3 Tera. 
Results: The linear correlations between the HbA1c methods were as follows: 
Cobas 6000 = 0.982 x Capillarys 3 Tera + 0.975, R² = 0.994; 
Architect c8000 = 0.982 x Capillarys 3 Tera + 1.064, R² = 0.994;
Mono S® = 0.916 x Capillarys 3 Tera + 3.397, R² = 0.965; 
BioRad Variant II Turbo = 0.923 x Capillarys 3 Tera + 4.062, R² = 0.990; 
Tosoh G8 = 0.963 x Capillarys 3 Tera + 3.895, R² = 0.996.
Conclusions: The different instrument platforms showed the best agreement in the 50-70 mmol/mol interval. Above and below this range the methods separated into 2 groups, one consisting of Capillarys 3 Tera, Roche Tina-Quant and Abbott enzymatic method and the other group consisting of BioRad Variant II Turbo, Tosoh G8 and Mono S®.


Introduction
Hemoglobin A1c (HbA1c) is an important blood test for the diagnosis and monitoring of diabetes mellitus patients, providing information about the average blood glucose levels over the past 2 to 3 months. Together with glucose it is used for glycemic control assessment of both type 1 and type 2 diabetes patients [1,2]. There are a number of different methods for HbA1c testing that cover the range from point of care environments to large centralized laboratories. The diabetes patient may thus have their HbA1c measured with different methods over a time period. To be able to correctly evaluate changes in HbA1c levels it is important that these different methods provide similar test results. The accuracy goal for HbA1c results in Equalis external quality assessment (EQA) scheme is that results should deviate less than 1.5 mmol/mol + 1.65 * 0.025 * the assigned HbA1c value. This corresponds to an allowed deviation of 3 mmol/mol at the level of 37 mmol/mol, and of 5 mmol/mol at the level of 85 mmol/mol. To be able to reach this accuracy it is important that the different methods in use provide test results without bias and with a high precision. HbA1c is used to monitor the diabetes care in Sweden and is an important part of the Swedish national diabetes register (SNDR) [3,4]. SNDR strives to improve the quality of care of Swedish diabetes patients by providing tools for monitoring the performance of the care and HbA1c is one of the most important biomarkers used for this monitoring [2,5]. The registry collect very large number of patient results which means that even small differences in trueness will reach significance.
There are a number of different available methods for HbA1c determinations such as ion-exchange high-performance liquid chromatography (e.g. Mono S ® ), boronate affinity HPLC, capillary zone electrophoresis (CZE), enzymatic assays and immunoassays [6,7]. Much effort has been put into the standardization of HbA1c and it is generally recognized that all HbA1c results shall be traceable to the International Federation of Clinical Chemistry (IFCC) reference method even if the results may be reported as different units [8]. In Sweden, all HbA1c results are reported using the unit mmol/mol and the different methods should theoretically give the same test results. It is difficult to have good agreement over broad measuring ranges, so most likely the companies will give priority to the HbA1c range that is the most important for the care of diabetes patients (e.g. 50-70 mmol/mol). The aim of the present study was to compare HbA1c results provided by the new Capillarys 3 Tera, Cobas 6000 (immunological method), Architect 8000 (enzymatic method), Mono S ® , BioRad Variant II Turbo, and Tosoh G8 (three ion exchange chromatography methods) using fresh patient samples and external quality assurance pooled whole blood samples.

I Samples
Blood samples were collected in K2-EDTA tubes according to the local routines at seven Swedish sites; Gothenburg, Falun, Gävle, Huddinge, Malmö, Uddevalla and Uppsala. Routine requested HbA1c samples were used for this evaluation. After the initial routine analysis at each laboratory the samples were sent to Uppsala the same or following day where the samples were reanalyzed on Capillarys 3 Tera. The study was approved by the local ethical board at Uppsala University (01-167). Sample stability for Capillarys 3 Tera was 7 days at +4°C and the calibration of the system was verified by analyzing samples that had previously had been measured with the IFCC reference method (INSTAND, Germany) [9].
The Mono S ® method is one of the designated comparison methods for HbA1c in relation to the IFCC reference measurement system [10]. and has served for many years as national standardization for HbA1c in Sweden. A fixed equation is used to convert result from Mono S ® scale to IFCC scale: HbA1c (mmol/mol) = 10.45 x HbA1c (Mono S ® , %) -10.62. The stability of the method is carefully studied in the IFCC monitoring program. The Mono S ® procedure is currently used to assign target values for Equalis external quality assurance program for HbA1c. The five evaluation sites in this study participated in six ordinary EQA surveys during the period January -June 2016.

II Architect c8000
60 patient samples were analyzed on both Architect c8000 and Capillarys 3 Tera. The equation for the correlation was y = 0.982 x Capillarys 3 Tera + 1.064, R² = 0.994. A deviation plot is shown in (Figure 1b).

III Mono S ®
60 patient samples were analyzed on both Mono S ® and Capillarys 3 Tera. The equation for the correlation was y = 0.916 x Capillarys 3 Tera + 3.397, R² = 0.965, A deviation plot is shown in (Figure 1c).

VI Comparison between different methods
The positive intercepts for the Mono S ® , Bio-Rad Variant II Turbo, and Tosoh G8 instrument in comparison with the Capillarys 3 Tera will give slightly higher values in the low range (e.g. below 40 mmol/mol) while the lower slope for the same instruments will give slightly lower values at the high end (Figure 1a-e).  (Figure 2), but the differences were less than 2 mmol/mol for all studied methods. There was a correlation (R 2 =0.38) between the deviation of the participating laboratories from the Mono S ® assigned values in the Equalis surveys and the deviation from the Capillarys 3 Tera results with fresh patient samples ( Figure 3).

VIII Results from comparison of Capillarys 3 Tera and the IFCC reference measurement procedure
The Capillarys 3 Tera device did not participate in the regular EQA surveys. Therefore, five external quality assurance samples used by Equalis for EQA were analyzed in duplicate on the Capillarys 3 Tera. The samples had been stored at -70°C after the use in regular EQA schemes and had also been measured with IFCC reference method (INSTAND, Germany) in addition to the Mono S ® assigned values [9]. The quality goal for EQA was fulfilled for duplicate measurements for all 5 samples in relation to the IFCC reference method ( Figure 4) and to the Mono S ® procedure (data not shown).

Discussion
In the present study we compared HbA1c results from Roche Tina-Quant HbA1c Gen 3, BioRad Variant II Turbo (3 instruments), Tosoh G8, Mono S ® and Architect c8000 with results with Capillarys 3 Tera. The comparisons were made as paired instrument comparisons with Capillarys 3 Tera. HbA1c is measured in whole blood. Even if HbA1c has a much better stability than blood cell counts at room temperature or at 4C, the sample stability has to be taken into account when performing method comparisons. This method comparison evaluated eight different instruments at different sites. A consecutive testing strategy would take minimum seven days if we estimate one day per site including shipping and testing. HbA1c is usually not measured by routine laboratories during Saturdays and Sundays in Sweden which means additional days between first and last measurement. Freezing or transferring samples to multiple secondary tubes may also cause handling errors. We believe that it is important to do this type of comparison with fresh patient samples rather than with controls or healthy blood donors. To be able to use fresh patient samples we chose to do paired comparisons. We chose to do the comparison with the reference method on only one instrument for logistic reasons. The methods used in this study were all CE labelled which includes requirements of interference studies. The present study did not include such studies which is a limitation of the study. Another limitation of the paired comparisons is that it is not the same samples that are run on all instruments. The difference in patient selection and number of samples could affect the results of the study and small differences in slope or intercept should be due to these differences. Several different assay principles were compared in the present study. The measurement principle of the Architect method is based on the enzymatic quantification of fructosyl dipeptides by a fructosyl peptide oxidase after digestion of the β chains of HbA1c with a proteinase [11]. The reaction produces hydrogen peroxide which reacts with a substrate. The reaction is measured at 660 nm.
The CAPILLARYS3 Tera HbA1c assay is based on the principle of capillary electrophoresis in free solution.12 Hemoglobin fractions are separated in silica capillaries by their electrophoretic mobility and electroosmotic flow at a high voltage in an alkaline buffer. Hemoglobin fractions are directly detected at 415 nm. The Roche assay is based on immunoturbidimetric determination of HbA1c. Monoclonal antibodies attached to latex particles bind to the β-N-terminal HbA1c fragments. A synthetic polymer with multiple β-N-terminal standard HbA1c copies is used to agglutinate remaining free antibody. The turbidity change measured at 552 nm is inversely related to the concentration of bound HbA1c. The BioRad method is based on chromatographic separation of HbA1c on a cation exchange cartridge. The various forms of hemoglobin exhibit charge differences at the acidic pH of the mobile phase, and thus can be separated on a support that is negatively charged. The separation is achieved by increasing the concentration of buffer B containing sodium perchlorate. Hemoglobin fractions are detected at 415 nm. The Tosoh G8 method is based on chromatographic separation of hemoglobin forms, including HbA1c, on a cation exchange column. The separation is achieved by increasing the concentration of buffer B. The hemoglobin fractions are eluted from the column and the eluted fractions are detected at 415 nm. The Mono S method is also based on cation exchange chromatography combined with detection of eluted fractions at 415 nm. The separation is achieved by increasing the concentration of buffer B. Both Capillarys 3 Tera and Bio-Rad, most recently in the form of the D-100 instrument, has been shown to perform very well for the determination of HbA1c [12].
The best agreements with Capillarys 3 Tera were found for Cobas 6000 and Architect c8000 with intercepts of approximately 1 mmol/mol and slopes of 0.98. We were positively surprised over the very good compatibility between these assays as they represent three completely different assay technologies. Cobas 6000, Architect c8000, Bio-Rad Variant II Turbo, and Tosoh G8 had all strong correlations with Capillarys 3 Tera of (R² 0.988-0.996). This is in agreement with the high accuracy and good reproducibility shown for these instruments in the first CAP survey of 2016 for Cobas 6000, Architect c8000, Tosoh G8 and Bio-Rad, Variant II Turbo(http://www.ngsp.org/CAP/CAP16a.pdf). The Capillary 3 Tera is so new on the market that it was not part of the CAP survey, but its predecessor the Capillarys 2 also had a low CV. Immunological methods have sometimes been reported to be analytically inferior for the measurement of HbA1c, but not in this study. The three Bio-Rad Variant II Turbo instruments had in comparison with Capillarys 3 Tera positive intercepts of 3.5-5 mmol/mol which was partly compensated by a slope of 0.923-0.953 [13,14]. The Mono S ® and Tosoh G8 systems showed similar intercepts and slopes as the Bio-Rad instruments. All studied instruments thus showed a good agreement in the 50-70 mmol/mol range. It seems likely that the manufacturers have focused their efforts to obtain agreement with the international calibrators in this clinically important range rather than at the high or low ends of the HbA1c range. It is important to be aware of these differences in research projects studying non-diabetic individuals with HbA1c within the lower end of the reference interval or diabetes patients with very high HbA1c values. The present study shows that there is a difference between the two instrument groups at the high and low end of the HbA1c range, but it cannot be used to define which of the two instrument groups that are the more correct in these regions. The Mono S ® method was for many years the basis for the national standardization of HbA1c in Sweden. Today, the majority of the Swedish laboratories have replaced Mono S ® with other routine methods for HbA1c measurement, but the Mono S ® procedure is still used, e.g. for assigning target values for EQA program for HbA1c. By use of the known relation to the IFCC scale, the results are made traceable to the IFCC reference measurement procedure. In the present study the agreement between the results from the routine methods and target values for EQA samples was good, as was the agreement between the routine methods and the Capillarys 3 Tera results for fresh patient samples. We also compared the deviation of the participating laboratories from the Mono S ® assigned values in the Equalis surveys with the deviation from the Capillarys 3 Tera results with fresh patient samples. The correlation between these two deviations is a strong indicator that the external quality assurance materials, consisting of pooled left over whole blood patient samples, detect the same kind of method deviations as seen with fresh patient samples. The accuracy of the Capillarys 3 Tetra was verified through measurements of samples with target values produced by IFCC reference measurement procedure. The results thus confirm the usefulness of a common traceability for various routine methods to the IFCC reference measurement procedure. A positive bias was observed for results for the Tosoh G8 and the BioRad Variant Turbo results both in relation the Mono S ® target in the EQA scheme with pooled whole blood samples and in relation to the Capillarys Tera 3 results for native patient samples ( Figure 5). In conclusion, HbA1c is today used as a diagnostic test for diabetes with a cut-off value of 48 mmol/mol according to WHO recommendation [15]. The use of international guidelines emphasizes the importance of co-calibration of HbA1c methods used clinically. The present study showed a good agreement between all studied instruments in the clinically important 50-70 mmol/L range and supports the use of the international cut-off value for the methods studied. The data also supports the use of external quality assurance material to detect method differences for HbA1c.