View Article PDF

HbA1c is HbA that is glycated at the N-terminal valine of one or both of the beta chains. Glucose can also attach to other amino acids, but HbA1c constitutes the majority of the glycated Hb forms and is the one used for the diagnosis and monitoring of diabetes since 2011.1 From a clinician’s perspective it is important to understand the differences, if any, between results produced by different methods and laboratories. These differences will dictate whether results from different laboratories produce clinically comparable results and can be used interchangeably for screening, diagnosis and monitoring of diabetes. If so, they can be displayed and interpreted cumulatively to show trends in guiding management.

HbA1c measurement was standardised in 2001 by the International Federation for Clinical Chemistry (IFCC), resulting in marked improvement in inter-assay and inter-laboratory performance and clinical utility.2 Furthermore, the US National Glycohemoglobin Standardisation Program (NGSP) and the IFCC set out quality targets to assist manufacturers, external quality assurance (EQA) programmes and laboratories with accuracy and precision goals for their assays.3,4

Laboratories choose an assay based on many factors, including clinical and technical performance, workflow and cost-effectiveness. Having different HbA1c assays in a region has its advantages and disadvantages and is common. From an analytical perspective, differences between assays and laboratories are inevitable and can be explained by the difference in measurement methodology, the inherent imprecision and bias associated with individual assays, and changes in lots of reagents and calibrators for all assays. While standardisation and quality targets have improved the quality of results and inter-assay comparability, there can still be differences between assays that can preclude interchangeability, especially in certain patient groups.

Differences between HbA1c assays can be accentuated, or appear exclusively, in the presence of haemoglobinopathies, the existence of which the laboratory, clinician or even the patient in some cases, may not be aware of. Haemoglobinopathy is the term used for congenital abnormalities of haemoglobin (Hb) covering both Hb production abnormalities, ie, thalassaemia, and structural Hb variants. These discrepancies or interferences are due to the different ways the measurement methodology interacts with the abnormal haemoglobin molecule in the case of structural variants, eg, sickle cell Hb (HbS) in which the glutamic acid amino acid in position six of the beta chain is replaced with valine.5 On the other hand, in case of a production abnormality, eg, beta thalassemia in which there is reduction in the formation of the beta chain due to mutations, deletions or insertions, the interference is related to the quantitative reduction in beta chains and associated increase in the proportion of more primal Hb forms like foetal Hb (HbF). The latter is composed of two alpha and two gamma chains. Large quantities of HbF are known to affect several assays.6

Red blood cell (RBC) volume, measured as the mean corpuscular volume (MCV), is also reported to influence the rate of glycation of Hb.7 The smaller the volume, eg, iron deficiency anaemia, the larger the surface-area-to-volume ratio, which increases the proportion of glycated Hb. Furthermore, nutritional iron deficiency is associated with higher HbA1c levels through an effect on the glycoxidation process itself. 8–10

The effect of haemoglobinopathies on measurement and interpretation of HbA1c is a matter of clinical concern because of the increasingly diverse ethnicities in New Zealand, including from the Pacific, Middle East, and parts of Asia where the frequency of certain haemoglobinopathies is greatly increased.11

Methods

One hundred and fifty-four (154) samples sourced from the community laboratory in Auckland (Labtests) were tested between September 2017 and August 2018. Samples chosen were residual ethylenediaminetetraacetic acid (EDTA) samples for which a HbA1c had been requested, and were either known to have a haemoglobinopathy as documented in the laboratory database, or on which a haemoglobinopathy screen had been requested on the basis of clinical or laboratory suspicion. Reasons may include a relevant family history, otherwise unexplained anaemia, unexplained low mean corpuscular haemoglobin (MCH) or MCV, or an apparent discrepancy between blood glucose and HbA1c concentrations.

HbA1c (requested by the clinician) was tested on the day of collection in the community laboratory. Aliquots were made and forwarded to two other (hospital) laboratories for HbA1c testing. If not tested on the same day, the other laboratories stored the samples at 4oC until tested within the week.

HbA1c in the community laboratory was tested on a Roche Gen.3 Cobas c513 (Germany) immunoassay (IA) platform. The second laboratory tested HbA1c by capillary zone electrophoresis (CZE) using the Capillarys 2 Flex-Piercing (Sebia, France) platform and the third laboratory tested HbA1c by ion exchange chromatography (IEC) on the Bio-Rad D-100 (California) platform. These three methods are named IA, CZE and IEC hereon. All three assays are traceable to the IFCC and NGSP reference methods. Coefficients of variation (CV), a measure of precision of an assay, at the time of testing were as follows: 3.9% and 2% at 34mmol/mol and 78mmol/mol respectively for the IA; 3% at both HbA1c levels for CZE; and 2.1% and 1.4% at 34mmol/mol and 81mmol/mol respectively for IEC.

All three methods use potassium EDTA, the preservative used in all samples in this study. All three laboratories are accredited to ISO15189 standard by International Accreditation New Zealand.

During the study period there were, depending on the assay, approximately 2–10 lots of calibrators, 3–4 lots of quality control material, and 1–6 lots of reagents used.

Samples from children (less than 18 years old) and from pregnant women were excluded. Apart from basic demographics (age, sex), the state of diabetes, MCV and Hb at the time of the study, other clinical details were not sought. Therefore individuals with haematological malignancies, chronic illness or other conditions that may affect the RBC lifespan or haematological parameters were not excluded.

Concordance between methods was assessed using non-parametric Passing-Bablok regression analysis and Bland-Altman difference plots for all results. Passing-Bablok was chosen because it allows for imprecision in both comparators (x and y). The regression equation y=a+bx was used to interpret the relation between the two relevant assays. A 95% confidence interval (CI) containing 0 for the intercept a, and 1 for the slope b indicates statistical comparability within the measured concentration range. A 95% CI that does not include 0 for a, and 1 for b indicates a systematic difference and proportional bias respectively.

Bland-Altman difference plots evaluate bias between the mean differences and estimate an agreement interval within which 95% of the differences between the two relevant methods fall.

For the two largest groups of haemoglobinopathies, beta thalassaemia trait (BT) and alpha thalassaemia trait (AT), three more methods were used, namely: the Royal College of Pathologists of Australasia’s Quality Assurance Programme (RCPA-QAP) allowable limits of performance (ALP) for HbA1c based on biological variation,12 the change in HbA1c concentration (mmol/mol) that is considered clinically significant,13 and whether the difference between results can cause misclassification of an individual as diabetic or prediabetic according to current New Zealand guidelines diagnostic cut-offs.1 Assessing the median of triplicate results against the RCPA-QAP ALP was considered a mathematically useful way of assessing concordance since the RCPA-QAP is used by many local and Australasian laboratories. None of the three methods was considered a reference method and no assumption of accuracy was made.

The relation between HbA1c and Hb and MCV was assessed using correlation scatter plots and the correlation coefficient r to determine the strength of the correlation, or lack of.

Ethics

Ethics approval was not needed because this study was for laboratory quality assurance purposes only on the assay originally requested, and patients were anonymised.14

Statistical analysis

MEDCAL statistical software, MedCal software Ltd. version 19, was used to compare the three methods for HbA1c measurement with Passing-Bablok and Bland-Altman analyses, and for correlation analysis for the relation between HbA1c, and Hb and MCV.

Results

A total of 154 samples were tested, of which nine were subsequently excluded: three patients were pregnant, two were children, two patients did not have data on MCV and Hb, one sample was repeated and one result was lost. A total of 145 samples from 145 patients were included in the final analysis, which included 82 (56%) females. Ages ranged from 18 to 86 years. There were 44 individuals with known diabetes mellitus. Table 1 summarises the number of samples for the haemoglobinopathies found in the study. Twenty-five samples did not show an apparent abnormality (NAA) based on the HbH inclusion body test used in our institution. The latter is a phenotypic test for alpha thalassaemia and cannot exclude a single or two gene deletion alpha thalassaemia.

Table 1: Haemoglobinopathy findings in increasing order of frequency.

BT: beta thalassemia trait; AT: alpha thalassemia trait; NAA: no apparent abnormality- cannot exclude one or two gene deletion alpha thalassemia; N: normal haemoglobinopathy screen; HbE: haemoglobin E; HbD: haemoglobin D; HbS: haemoglobin S; HBT: homozygous beta thalassemia; *other: Heterozygous HbS, HbS/ beta thalassemia, alpha thalassemia /HbE trait, Haemoglobin H (HbH) disease, hereditary persistent HbF, and heterozygous haemoglobin C (HbC).

Figures 1a to 3b summarise the correlation and differences between the three platforms for all results.

Figure 1a: Correlation between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

The 95% CI for the intercept a (-1.50) is -3.17–1.0, and for the slope b (1.05) is 1.0–1.2. This correlation indicates statistical comparability between CZE and IEC within the measured concentration range.

Figure 1b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

Upper limit of agreement (ULOA) was 4.4 (95% CI: 3.88–5.0); mean difference was 0.6 (95% CI: 0.27–0.88); and lower limit of agreement (LLOA) was -3.3 (95% CI: -2.76–-3.84).

Figure 2a: Correlation between capillary zone electrophoresis (CZE) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.99–1.0 and for the slope b (1.0) was 1.0–1.05 indicating statistical comparability between CZE and IA within the measured concentration range.

Figure 2b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and immunoassay (IA)

The ULOA was 5.8 (95% CI: 5.08–6.6); mean difference was 0.4 (95% CI: 0.09-0.94); and LLOA was -5.1 (95% CI: -5.86–-4.23).

Figure 3a: Correlation between ion exchange chromatography (IEC) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.0–1.06 and for the slope b (1.0) was 0.97–1.0 indicating statistical comparability between IEC and IA within the measured concentration range.

Figure 3b: Bland-Altman plot for the difference between ion exchange chromatography (IEC) and immunoassay (IA).

The ULOA was 5.7 (95% CI: 4.95–6.56); the mean difference was -0.2 (95% CI: -0.79–0.33); and LLOA was -6.1 (95% CI: -5.34–-6.95).

There was only one patient with HbS/ beta thalassemia in the study. The patient had recurrent sickle cell crises and regular blood transfusions. However, the same blood sample was tested by the three platforms in this study. Table 2 summarises HbA1c levels for the four most discordant results in the study.

Table 2: Details of the four results with maximum differences between the highest and lowest results.

CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; HbS: haemoglobin S.

Results for beta thalassemia trait and alpha thalassemia trait

The RCPA-QAP ALP for HbA1c are +/-4 up to 45mmol/mol and 8% for levels higher than 45mmol/mol.12 These ALPs were used to estimate concordance between the three assays for alpha thalassemia trait (AT) (42 samples) and beta thalassemia trait (BT) (38 samples) groups using the median of the triplicates as true value. For the majority of triplicates the measured values fell within the range for the median that was based on RCPA-QAP ALP, and the maximum difference between the measured values would not have misclassified glycaemic status. Table 3 summarises results for the discordant results.

Table 3: Discordant haemoglobin A1c results for beta thalassemia trait (BT) (total 5) and alpha thalassemia trait (AT) (total 3).

ALP: allowable limits of performance; CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; RCPA: Royal College of Pathologists of Australasia; QAP: quality assurance programme. * Maximum difference between results was ≥5mmol/mol.

Six of a total of 80 samples with thalassemia (7.5%) had discrepant results that could misclassify glycaemic status.

Results for the relation between HbA1c concentration, and mean corpuscular volume and haemoglobin

MCV and Hb concentration for 131 individuals were available for analysis. The range of MCV was 57–92fL and that for Hb 54–160g/L. The median HbA1c level for the three assays was plotted against the MCV and Hb; Figures 4 and 5 respectively.

Figure 4: The relationship between the mean corpuscular volume (MCV) and the median HbA1c concentrations.

Correlation coefficient r is 0.021 (95% CI -0.15–0.19) (P=0.8). This signifies no linear correlation between HbA1c concentration and MCV in this population.

Figure 5: The relationship between the haemoglobin (Hb) concentration and the median HbA1c concentrations.

Correlation coefficient r is -0.09 (95% CI was -0.26–0.08) P=0.3. This signifies no statistical linear correlation between HbA1c concentration and Hb in this population in spite of a suggestive trend.

Discussion

Concordance

Passing-Bablok comparisons demonstrated acceptable concordance between the three platforms. Bland-Altman plots demonstrated the highest mean difference of 0.6 mmol/mol between CZE and IEC with CZE being the higher of the two sets. However, these same platforms had the narrowest LOA compared to their individual differences with the IA platform.

Differences in two of the four discordant triplicate results were due to low IA results, one due to a high IA result, and the fourth due to a discrepantly low IEC result (Table 2). The higher concordance between CZE and IEC may be because both techniques exploit molecular charge for separation while IA techniques are based on the binding between antigen and antibody and exposure of a unique epitope (glycated amino group of N-terminal valine of the beta chain).

For purposes of this study, the median of each triplicate was presumed closest to the “true” value. The RCPA-QAP ALP was then used to determine comparability of the three results. If the results fell within the ALP range for the median value it was assumed that the results were clinically comparable. This is an unbiased criterion because the RCPA-QAP ALP is based purely on what is considered an acceptable analytical imprecision, for the known biological variation of the analyte, without the additional factors in play including the difference in methodologies, precision and calibration, lot-to-lot variation in reagents and age of the sample. For BT, one set of triplicate results had a level outside of the ALP by approximately 4.6mmol/mol (64mmol/mol versus the upper limit for the ALP of the median 59.4mmol/mol). However, this discrepancy did not cause a difference in classification in glycaemic status between platforms. For AT, one set of triplicate results had a level outside the ALP by 2mmol/mol (35mmol/mol versus the lower limit of the ALP for the median of 37mmol/mol). This discrepancy did cause a difference in classification in glycaemic status between platforms. There were no evident biological or analytical explanations for these discrepancies.

When monitoring diabetic patients, a difference of 5mmol/mol or more is usually considered significant,13 indicating an improvement or worsening of glycaemic control based on current recommended analytical standards of performance. This difference was used to determine the clinical comparability of results; if the analytical difference between any two results was 5mmol/mol or more, the inter-assay discrepancy may cause inaccurate clinical interpretation. In practice, the 5mmol/mol difference threshold is intended to be the difference between two consecutive results for an individual patient using the same assay, while in this study the same sample was tested three-fold. However it provides a useful, if crude, measure of comparability when considering a patient having serial results using different methods. In the BT group, one triplicate demonstrated a maximum difference of 5mmol/mol (32mmol/mol versus 37mmol/mol) and another 11mmol/mol (53mmol/mol versus 64mmol/mol). In the AT group there was only one set of triplicates with a maximum difference more than 5mmol/mol (35mmol/mol versus 42mmol/mol). The two BT triplicates did not demonstrate a difference in classification of glycaemia probably because the levels were far from the diagnostic cut-offs, whereas the AT triplicate did because levels were around the prediabetes cut-off of ≥41mmol/mol.

There was one result that violated all three measures of concordance: the AT triplicate of 35mmol/mol, 41mmol/mol and 42mmol/mol. There was no biological or analytical explanation.

In New Zealand, a HbA1c of 50mmol/mol or higher is the current recommended diagnostic threshold for diabetes mellitus, and 41–49mmol/mol is the diagnostic range for pre-diabetes.1 Using these cut-offs as a marker for concordance we found three triplicates in the BT group and three in the AT group had a level that misdiagnosed diabetes or pre-diabetes. According to the New Zealand guidelines1 a single HbA1c level is insufficient to diagnose diabetes unless the individual is symptomatic, in which case the HbA1c is expected to be clearly high. In asymptomatic individuals a repeat HbA1c or an additional elevated plasma glucose is needed to confirm the diagnosis. Assuming assay performance remains constant it is probable that a repeat HbA1c measurement for three of the samples, samples 1 and 6 by IA and sample 7 by CZE, would correctly reclassify the patients’ glycaemic status, aligning them with levels from the other two relevant platforms.

The assays

The IFCC model for quality targets was used to investigate the performance of HbA1c assays by 24 manufacturers in 17 countries and 2,166 laboratories.3 It is a model based on total error in which performance criteria are derived from sigma (σ) metrics. The criterion to be met was a total allowable error (TAE) of 5mmol/mol at the 2σ level for a HbA1c level of 50mmol/mol.3,15 Bio-Rad D-100 and Sebia Capillarys 2 Flex-Piercing met the IFCC criteria. Multiple Roche assays were grouped together because of their large variety, but regardless they also met the IFCC criterion for accuracy; the failure in the precision criterion was deemed due to inter-laboratory variability.3 Lenters-Westra and English16 in their comparative study of three HbA1c assays including Roche Tina-quant Gen.3 on Cobas c513 (the assay used by the community lab) demonstrated an analytical precision of 2% and 2.1% at 46mmol/mol and 72mmol/mol for the Roche assay, respectively.16

HbA1c reliability to assess dysglycaemia in patients with haemoglobinopathies and other conditions

In healthy adults, the normal physiological proportion of HbA (α2β2) is 95–98% and of HbF <1%.17 Most haemoglobinopathies, of which there are approximately 1,300, are clinically silent and do not affect RBC survival or interfere with HbA1c measurement, but the proportion of HbA can vary in the presence of some.18 Individuals with HbS/ alpha thalassemia typically have 60–70% HbA (α2β2)19 depending on the genotype, and individuals with HbE/alpha thalassemia would have 20–30% HbA.20 Also depending on the genotype, those with HbS/beta thalassemia have a maximum of 30% HbA.21

Clinical cut-offs for HbA1c assume normal physiological RBC lifespan and proportion of HbA. Therefore a significantly shorter lifespan, or lower proportion of HbA relative to serum/intraerythrocytic glucose9 are expected to influence interpretation of measured HbA1c, rendering it potentially unreliable in the diagnosis and monitoring of dysglycaemia regardless of method used. Furthermore, individuals with HbS/beta thalassemia and HBT often receive regular transfusions rendering the measured HbA1c an inaccurate reflection of glycaemic status. When the proportion of HbA is abnormal, RBC survival is short, or in case of regular blood transfusions, HbA1c may be unreliable by any method and should not be used. It is appropriate to assess glycaemia in such individuals by other means such as plasma glucose, fructosamine or (if available) glycated albumin.

Variation in RBC survival and glucose entry into RBC22 can occur even in patients with normal haematology23 and no known chronic disease or drug effects, and most of the genetic variation in HbA1c is due to non-glucose factors as shown in gene studies.24 This is an inherent, and often unrecognised, limitation of the test in all patients, not just those with known causes of unreliability, and limits the ability of HbA1c to assess estimated average glucose (eAG).25

Haematological parameters

HbA1c concentration is inversely correlated with MCV, Hb and MCH in premenopausal females without haemoglobinopathies or diabetes; the strongest correlation being with MCH.26 A similar association with MCV and MCH was found in 1,315 adults without haemoglobinopathies, diabetes or serum creatinine ≥120umol/L.27

Our study was suggestive, albeit not statistically, of an inverse correlation between HbA1c concentration and Hb, but demonstrated no correlation between HbA1c and MCV. Unlike findings in individuals with normal Hb,26 our population consisted of samples from both genders, and pre- and post-menopausal females, the majority of whom had a haemoglobinopathy. Increased RBC turnover, haemolysis and blood transfusions contribute to lowering of HbA1c. These conditions would not be represented in studies of healthy individuals but would skew our findings towards lower HbA1c at smaller MCVs.

Strengths, weaknesses and significance of this study

A strength of this study is its specific focus on haemoglobinopathies with large numbers of patients; the findings add insight and inform local clinicians on diagnosis and monitoring of their diabetic patients who have haemoglobinopathies. Furthermore, as the timeframe for the study included changes of reagent lots and calibrators, transportation, different methodologies and operators, the differences are a reflection of the ‘real life’ scenario in laboratories.

There were limited numbers of samples without a haemoglobinopathy, but differences in patients without known haemoglobinopathies were not the focus of this study and inter-method correlation would be expected to be, if anything, better. All New Zealand laboratories participate in two-weekly inter-laboratory comparisons of HbA1c using de-identified human samples of mostly normal HbA (Waikato QAP scheme, New Zealand)28 and in a wider Australasian RCPA QAP comparison using both human and non-human matrix.12,20 Laboratories using all three methods compared in this study are represented in the Waikato scheme, with 11 participants using IEC, six using IA and two using CZE.28

Although numbers for some uncommon haemoglobinopathies were small, making it difficult to make firm conclusions, this was a reflection of the numbers in the local community. Results of this study do not imply that the same correlation applies to other haemoglobinopathies or assay formats (even if from the same manufacturer) but do indicate that for the assays in this study the difference in HbA1c levels for the haemoglobinopathies included is acceptable in the large majority of cases.

Conclusion

This study was a pragmatic three-way comparison for HbA1c in non-pregnant adults, most of whom had a haemoglobinopathy. Of the six triplicate sets of results from the thalassemia groups where inter-platform results caused a change in classification of glycaemia, three may have re-classified the patients had they been repeated. There was no statistical correlation between HbA1c, and Hb and MCV.

While HbA1c measurement technology has improved greatly, there is no perfect test and HbA1c measurement still demonstrates inter-platform discrepancies, particularly using the IA method. However, the difference between methods was found not to be clinically significant in the large majority of cases. Even for individual methods there can be differences in analytical bias and imprecision between laboratories. Assuming individual methods remain well controlled in each laboratory, and results are interpreted within the broader clinical context, this implies that HbA1c results from the three assays in this study can be viewed cumulatively, interchanged, and importantly trends can be determined for monitoring purposes. However, caution should be exercised in a small proportion of patients when a diagnosis is sought and where, regardless of methodology used, the result may not accurately reflect dysglycaemia. Further studies with larger sample size, blood glucose levels for correlation and standard reference material for better assessment of accuracy are recommended.

Summary

Abstract

Aim

To determine whether glycated haemoglobin (HbA1c) results from three commonly used platforms can be interpreted cumulatively and used interchangeably in individuals with common haemoglobinopathies. A secondary goal was to assess the relationship between HbA1c concentrations, and haemoglobin and mean corpuscular volume in this population.

Method

One hundred and forty-five samples, mostly with haemoglobinopathies, were tested by each of: Roche Gen.3 Cobas c513, Capillarys 2 Flex-Piercing and Bio-Rad D-100 platforms. Statistical comparisons and limits of performance based on biological variation, international recommendations, and local diagnostic cut-offs were drawn upon to determine comparability of results.

Results

Inter-platform measurements were not significantly different for the large majority of results. The four HbA1c results that showed maximum discrepancy between triplicates had the following abnormalities: heterozygous haemoglobin S/ beta thalassemia, heterozygous haemoglobin S/ alpha thalassemia, beta thalassemia trait and alpha thalassemia trait. Six triplicates of results in the thalassemia groups (7.5% of thalassemia samples) had levels that misclassified patients’ glycaemic status. There was no correlation between HbA1c concentration and mean corpuscular volume, and a weak negative correlation between HbA1c concentration and haemoglobin concentration.

Conclusion

HbA1c concentrations measured by Cobas c513, Capillarys 2 Flex-Piercing and the Bio-Rad D-100 were found to be comparable in the large majority of samples. While discordance was due to assay imprecision in some cases, in others no biological or analytical explanation could be found.

Author Information

Samarina MA Musaad, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland; George Chan, Consultant Haematologist, Haematology Department Labtests, Healthscope, Auckland; Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland.

Acknowledgements

The authors would like to thank and acknowledge Alek Latinovic and Sian Horan, medical laboratory scientists in the departments of chemical pathology at Labtests and Southern Community Laboratories respectively; and Linda Henderson, Technical Specialist in the department of specialist chemistry at Labplus, Auckland City Hospital. Alek tediously coordinated and organised sample distribution between the laboratories and all three scientists ensured timely testing and communication of results.

Correspondence

Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, PO Box 12049, Auckland 1642.

Correspondence Email

cam.kyle@labtests.co.nz

Competing Interests

Dr Kyle is Consultant Pathologist to Labtests, Auckland.

1. New Zealand Society for the Study of Diabetes. Executive Summary of position statement. http://www.nzssd.org.nz/HbA1c/2.%20NZSSD%20exec%20summary%20diagnosis%20of%20diabetes%20Sept%202011%20final.pdf (accessed 21/10/2019).

2. Little RR, Rohlfing CL, Sacks DB for the National Glycohemoglobin Standardization Program (NGSP) Steering Committee. Status of Hemoglobin A1c Measurement and Goals for Improvement: From Chaos to Order for Improving Diabetes Care. Clin Chem. 2011; 57:205–214.

3. The EurA1c Trial Group. EurA1c: the European HbA1c Trial to Investigate the Performance of HbA1c Assays in 2166 Laboratories across 17 Countries and 24 Manufacturers by Use of the IFCC Model for Quality Targets. Clin Chem. 2018; 64:8.

4. Wekcamp C. HbA1c: A Review of Analytical and Clinical Aspects. Clin Chem. 2013; 33:393–400.

5. Rhea JM, Molinaro R. Pathology Consultation on HbA1c Methods and Interferences. Am J Clin Pathol. 2014; 141:5–16.

6. HbA1c Assay Interferences. The National Glycohemoglobin Program (NGSP) http://www.ngsp.org/interf.asp (accessed 05/08/2019).

7. English E, Idris I, Smith G, Dhatariya K, et al. The effect of anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic review. Diabetologia. 2015; 58:1409–1421.

8. Selvaraj N, Bobby Z, Sathiyapriya V. Effect of lipid peroxides and antioxidants on glycation of hemoglobin: an in vitro study on human erythrocytes. Clin Chim Acta. 2006; 366:190–5.

9. Coban E, Ozdogan M, Timuragaoglu A. Effect of iron deficiency anemia on the levels of hemoglobin A1c in nondiabetic patients. Acta Haematol. 2004; 112:126–8.

10. Hashimoto K, Noguchi S, Morimoto Y, Hamada H et al. A1C but not serum glycated albumin is elevated in late pregnancy owing to iron deficiency. Diabetes Care. 2008; 31:1945–8.

11. Stats New Zealand Tatauranga Aotearoa. http://www.stats.govt.nz/information-releases/2018-census-totals-by-topic-national-highlights#download-data (accessed 21/10/2019).

12. The Royal College of Pathologists of Australasia Quality Assurance Programs. RCPA QAP http://rcpaqap.com.au/ (accessed 05/08/2019).

13. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, et al. Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus. Clin Chem. 2002; 48(3):436–472.

14. Human Tissue Act 2008 New Zealand Legislation. Parliamentary Counsel Office. http://www.legislation.govt.nz/act/public/2008/0028/latest/versions.aspx (accessed 10/02/2020).

15. Weykamp C, Siebelder C. Evaluation of Performance of Laboratories and Manufacturers Within the Framework of the IFCC model for Quality Targets of HbA1c. J. Diabetes Sci. Technol.2018; 12(4):747–752.

16. Lenters-Westra E, English E. Evaluating new HbA1c methods for adoption by the IFCC and NGSP reference networks using international quality targets. Clin Chem Lab Med. 2017; 55(9):1426–1434.

17. Burtis CA, Ashwood ER, Bruns DE, editors. Tietz Textbook of Clinical Chemistry and Molecular Diagnostics. 4th ed. Missouri: Elsevier Saunders; 2006.

18. Dunphy CH, editor. Molecular Pathology of Hematolymphoid Diseases. NY, USA: Springer; 2010. Chapter 37, Molecular Pathology of Hemoglobin and Erythrocyte Membrane disorders; page 473.

19. Huisman THJ. Trimodality in Tie Percenetages of β Chain Variants in Heterozy-Gotes: The Effect of the Number of Aczive HB α Structural Loci. Hemoglobin. 1977; 1:349–382.

20. Fucharoen S, Weatherall D. Hemoglobin E Disorders. In Steeinberg MH, Forget BG, HIggs DR, Weatherall EJ, editors. Disorders of Hemoglobin 2nd edition. Cambridge University Press, 2009. p. 417–433.

21. Sergeant GR, Ashcroft MT, Serjeant BE, Milner PF. The Clinical Features of Sickle-Cell /p Thalassaemia in Jamaica. Br J Haematol. 1973; 24:19.

22. Khera PK, Clinton HJ, Carruthers A, Lindsell CJ, et al. Evidence for Interindividual Heterogeneity in the Glucose Gradient Across the Human Red Blood Cell Membrane and Its Relationship to Hemoglobin Glycation. Diabetes 2008; 57:2445–2452.

23. Cohen RM, Franco RS, Khera PK, Smith EP, et al. Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood 2008 112:4284–4291 doi:10.1182/blood-2008-04-154112

24. Soranzo N. Genetic Determinants of Variability in Glycated Hemoglobin (HbA1c) in Humans: Review of Recent Progress and Prospects for Use in Diabetes Care Curr Diab Rep 2011; 11:562–569 DOI 10.1007/s11892-011-0232-9

25. Barth JH, Marshall SM, Watson ID. Consensus meeting on reporting glycated haemoglobin and estimated average glucose in the UK: report to the National Director for Diabetes, Department of Health. Ann Clin Biochem 2008; 45:343–344. DOI: 10.1258/acb.2008.200815

26. Koga M, Morita S, Saito H, Mukai M, et al. Association of erythrocyte indices with glycated haemoglobin in premenopausal women. Diabet Med. 2007; 24(8):843–847.

27. Simmons D, Hlaing T. Interpretation of HbA1c: association with mean cell volume and haemoglobin concentration. Diabet Med. 2014; 31(11):1387–1392.

28. Quality Assurance Programme Waikato District Health Board. http://lab.waikatodhb.health.nz/assets/QCItems/Quality-Assurance-Programme.pdf (accessed 5/8/2019).

Contact diana@nzma.org.nz
for the PDF of this article

View Article PDF

HbA1c is HbA that is glycated at the N-terminal valine of one or both of the beta chains. Glucose can also attach to other amino acids, but HbA1c constitutes the majority of the glycated Hb forms and is the one used for the diagnosis and monitoring of diabetes since 2011.1 From a clinician’s perspective it is important to understand the differences, if any, between results produced by different methods and laboratories. These differences will dictate whether results from different laboratories produce clinically comparable results and can be used interchangeably for screening, diagnosis and monitoring of diabetes. If so, they can be displayed and interpreted cumulatively to show trends in guiding management.

HbA1c measurement was standardised in 2001 by the International Federation for Clinical Chemistry (IFCC), resulting in marked improvement in inter-assay and inter-laboratory performance and clinical utility.2 Furthermore, the US National Glycohemoglobin Standardisation Program (NGSP) and the IFCC set out quality targets to assist manufacturers, external quality assurance (EQA) programmes and laboratories with accuracy and precision goals for their assays.3,4

Laboratories choose an assay based on many factors, including clinical and technical performance, workflow and cost-effectiveness. Having different HbA1c assays in a region has its advantages and disadvantages and is common. From an analytical perspective, differences between assays and laboratories are inevitable and can be explained by the difference in measurement methodology, the inherent imprecision and bias associated with individual assays, and changes in lots of reagents and calibrators for all assays. While standardisation and quality targets have improved the quality of results and inter-assay comparability, there can still be differences between assays that can preclude interchangeability, especially in certain patient groups.

Differences between HbA1c assays can be accentuated, or appear exclusively, in the presence of haemoglobinopathies, the existence of which the laboratory, clinician or even the patient in some cases, may not be aware of. Haemoglobinopathy is the term used for congenital abnormalities of haemoglobin (Hb) covering both Hb production abnormalities, ie, thalassaemia, and structural Hb variants. These discrepancies or interferences are due to the different ways the measurement methodology interacts with the abnormal haemoglobin molecule in the case of structural variants, eg, sickle cell Hb (HbS) in which the glutamic acid amino acid in position six of the beta chain is replaced with valine.5 On the other hand, in case of a production abnormality, eg, beta thalassemia in which there is reduction in the formation of the beta chain due to mutations, deletions or insertions, the interference is related to the quantitative reduction in beta chains and associated increase in the proportion of more primal Hb forms like foetal Hb (HbF). The latter is composed of two alpha and two gamma chains. Large quantities of HbF are known to affect several assays.6

Red blood cell (RBC) volume, measured as the mean corpuscular volume (MCV), is also reported to influence the rate of glycation of Hb.7 The smaller the volume, eg, iron deficiency anaemia, the larger the surface-area-to-volume ratio, which increases the proportion of glycated Hb. Furthermore, nutritional iron deficiency is associated with higher HbA1c levels through an effect on the glycoxidation process itself. 8–10

The effect of haemoglobinopathies on measurement and interpretation of HbA1c is a matter of clinical concern because of the increasingly diverse ethnicities in New Zealand, including from the Pacific, Middle East, and parts of Asia where the frequency of certain haemoglobinopathies is greatly increased.11

Methods

One hundred and fifty-four (154) samples sourced from the community laboratory in Auckland (Labtests) were tested between September 2017 and August 2018. Samples chosen were residual ethylenediaminetetraacetic acid (EDTA) samples for which a HbA1c had been requested, and were either known to have a haemoglobinopathy as documented in the laboratory database, or on which a haemoglobinopathy screen had been requested on the basis of clinical or laboratory suspicion. Reasons may include a relevant family history, otherwise unexplained anaemia, unexplained low mean corpuscular haemoglobin (MCH) or MCV, or an apparent discrepancy between blood glucose and HbA1c concentrations.

HbA1c (requested by the clinician) was tested on the day of collection in the community laboratory. Aliquots were made and forwarded to two other (hospital) laboratories for HbA1c testing. If not tested on the same day, the other laboratories stored the samples at 4oC until tested within the week.

HbA1c in the community laboratory was tested on a Roche Gen.3 Cobas c513 (Germany) immunoassay (IA) platform. The second laboratory tested HbA1c by capillary zone electrophoresis (CZE) using the Capillarys 2 Flex-Piercing (Sebia, France) platform and the third laboratory tested HbA1c by ion exchange chromatography (IEC) on the Bio-Rad D-100 (California) platform. These three methods are named IA, CZE and IEC hereon. All three assays are traceable to the IFCC and NGSP reference methods. Coefficients of variation (CV), a measure of precision of an assay, at the time of testing were as follows: 3.9% and 2% at 34mmol/mol and 78mmol/mol respectively for the IA; 3% at both HbA1c levels for CZE; and 2.1% and 1.4% at 34mmol/mol and 81mmol/mol respectively for IEC.

All three methods use potassium EDTA, the preservative used in all samples in this study. All three laboratories are accredited to ISO15189 standard by International Accreditation New Zealand.

During the study period there were, depending on the assay, approximately 2–10 lots of calibrators, 3–4 lots of quality control material, and 1–6 lots of reagents used.

Samples from children (less than 18 years old) and from pregnant women were excluded. Apart from basic demographics (age, sex), the state of diabetes, MCV and Hb at the time of the study, other clinical details were not sought. Therefore individuals with haematological malignancies, chronic illness or other conditions that may affect the RBC lifespan or haematological parameters were not excluded.

Concordance between methods was assessed using non-parametric Passing-Bablok regression analysis and Bland-Altman difference plots for all results. Passing-Bablok was chosen because it allows for imprecision in both comparators (x and y). The regression equation y=a+bx was used to interpret the relation between the two relevant assays. A 95% confidence interval (CI) containing 0 for the intercept a, and 1 for the slope b indicates statistical comparability within the measured concentration range. A 95% CI that does not include 0 for a, and 1 for b indicates a systematic difference and proportional bias respectively.

Bland-Altman difference plots evaluate bias between the mean differences and estimate an agreement interval within which 95% of the differences between the two relevant methods fall.

For the two largest groups of haemoglobinopathies, beta thalassaemia trait (BT) and alpha thalassaemia trait (AT), three more methods were used, namely: the Royal College of Pathologists of Australasia’s Quality Assurance Programme (RCPA-QAP) allowable limits of performance (ALP) for HbA1c based on biological variation,12 the change in HbA1c concentration (mmol/mol) that is considered clinically significant,13 and whether the difference between results can cause misclassification of an individual as diabetic or prediabetic according to current New Zealand guidelines diagnostic cut-offs.1 Assessing the median of triplicate results against the RCPA-QAP ALP was considered a mathematically useful way of assessing concordance since the RCPA-QAP is used by many local and Australasian laboratories. None of the three methods was considered a reference method and no assumption of accuracy was made.

The relation between HbA1c and Hb and MCV was assessed using correlation scatter plots and the correlation coefficient r to determine the strength of the correlation, or lack of.

Ethics

Ethics approval was not needed because this study was for laboratory quality assurance purposes only on the assay originally requested, and patients were anonymised.14

Statistical analysis

MEDCAL statistical software, MedCal software Ltd. version 19, was used to compare the three methods for HbA1c measurement with Passing-Bablok and Bland-Altman analyses, and for correlation analysis for the relation between HbA1c, and Hb and MCV.

Results

A total of 154 samples were tested, of which nine were subsequently excluded: three patients were pregnant, two were children, two patients did not have data on MCV and Hb, one sample was repeated and one result was lost. A total of 145 samples from 145 patients were included in the final analysis, which included 82 (56%) females. Ages ranged from 18 to 86 years. There were 44 individuals with known diabetes mellitus. Table 1 summarises the number of samples for the haemoglobinopathies found in the study. Twenty-five samples did not show an apparent abnormality (NAA) based on the HbH inclusion body test used in our institution. The latter is a phenotypic test for alpha thalassaemia and cannot exclude a single or two gene deletion alpha thalassaemia.

Table 1: Haemoglobinopathy findings in increasing order of frequency.

BT: beta thalassemia trait; AT: alpha thalassemia trait; NAA: no apparent abnormality- cannot exclude one or two gene deletion alpha thalassemia; N: normal haemoglobinopathy screen; HbE: haemoglobin E; HbD: haemoglobin D; HbS: haemoglobin S; HBT: homozygous beta thalassemia; *other: Heterozygous HbS, HbS/ beta thalassemia, alpha thalassemia /HbE trait, Haemoglobin H (HbH) disease, hereditary persistent HbF, and heterozygous haemoglobin C (HbC).

Figures 1a to 3b summarise the correlation and differences between the three platforms for all results.

Figure 1a: Correlation between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

The 95% CI for the intercept a (-1.50) is -3.17–1.0, and for the slope b (1.05) is 1.0–1.2. This correlation indicates statistical comparability between CZE and IEC within the measured concentration range.

Figure 1b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

Upper limit of agreement (ULOA) was 4.4 (95% CI: 3.88–5.0); mean difference was 0.6 (95% CI: 0.27–0.88); and lower limit of agreement (LLOA) was -3.3 (95% CI: -2.76–-3.84).

Figure 2a: Correlation between capillary zone electrophoresis (CZE) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.99–1.0 and for the slope b (1.0) was 1.0–1.05 indicating statistical comparability between CZE and IA within the measured concentration range.

Figure 2b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and immunoassay (IA)

The ULOA was 5.8 (95% CI: 5.08–6.6); mean difference was 0.4 (95% CI: 0.09-0.94); and LLOA was -5.1 (95% CI: -5.86–-4.23).

Figure 3a: Correlation between ion exchange chromatography (IEC) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.0–1.06 and for the slope b (1.0) was 0.97–1.0 indicating statistical comparability between IEC and IA within the measured concentration range.

Figure 3b: Bland-Altman plot for the difference between ion exchange chromatography (IEC) and immunoassay (IA).

The ULOA was 5.7 (95% CI: 4.95–6.56); the mean difference was -0.2 (95% CI: -0.79–0.33); and LLOA was -6.1 (95% CI: -5.34–-6.95).

There was only one patient with HbS/ beta thalassemia in the study. The patient had recurrent sickle cell crises and regular blood transfusions. However, the same blood sample was tested by the three platforms in this study. Table 2 summarises HbA1c levels for the four most discordant results in the study.

Table 2: Details of the four results with maximum differences between the highest and lowest results.

CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; HbS: haemoglobin S.

Results for beta thalassemia trait and alpha thalassemia trait

The RCPA-QAP ALP for HbA1c are +/-4 up to 45mmol/mol and 8% for levels higher than 45mmol/mol.12 These ALPs were used to estimate concordance between the three assays for alpha thalassemia trait (AT) (42 samples) and beta thalassemia trait (BT) (38 samples) groups using the median of the triplicates as true value. For the majority of triplicates the measured values fell within the range for the median that was based on RCPA-QAP ALP, and the maximum difference between the measured values would not have misclassified glycaemic status. Table 3 summarises results for the discordant results.

Table 3: Discordant haemoglobin A1c results for beta thalassemia trait (BT) (total 5) and alpha thalassemia trait (AT) (total 3).

ALP: allowable limits of performance; CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; RCPA: Royal College of Pathologists of Australasia; QAP: quality assurance programme. * Maximum difference between results was ≥5mmol/mol.

Six of a total of 80 samples with thalassemia (7.5%) had discrepant results that could misclassify glycaemic status.

Results for the relation between HbA1c concentration, and mean corpuscular volume and haemoglobin

MCV and Hb concentration for 131 individuals were available for analysis. The range of MCV was 57–92fL and that for Hb 54–160g/L. The median HbA1c level for the three assays was plotted against the MCV and Hb; Figures 4 and 5 respectively.

Figure 4: The relationship between the mean corpuscular volume (MCV) and the median HbA1c concentrations.

Correlation coefficient r is 0.021 (95% CI -0.15–0.19) (P=0.8). This signifies no linear correlation between HbA1c concentration and MCV in this population.

Figure 5: The relationship between the haemoglobin (Hb) concentration and the median HbA1c concentrations.

Correlation coefficient r is -0.09 (95% CI was -0.26–0.08) P=0.3. This signifies no statistical linear correlation between HbA1c concentration and Hb in this population in spite of a suggestive trend.

Discussion

Concordance

Passing-Bablok comparisons demonstrated acceptable concordance between the three platforms. Bland-Altman plots demonstrated the highest mean difference of 0.6 mmol/mol between CZE and IEC with CZE being the higher of the two sets. However, these same platforms had the narrowest LOA compared to their individual differences with the IA platform.

Differences in two of the four discordant triplicate results were due to low IA results, one due to a high IA result, and the fourth due to a discrepantly low IEC result (Table 2). The higher concordance between CZE and IEC may be because both techniques exploit molecular charge for separation while IA techniques are based on the binding between antigen and antibody and exposure of a unique epitope (glycated amino group of N-terminal valine of the beta chain).

For purposes of this study, the median of each triplicate was presumed closest to the “true” value. The RCPA-QAP ALP was then used to determine comparability of the three results. If the results fell within the ALP range for the median value it was assumed that the results were clinically comparable. This is an unbiased criterion because the RCPA-QAP ALP is based purely on what is considered an acceptable analytical imprecision, for the known biological variation of the analyte, without the additional factors in play including the difference in methodologies, precision and calibration, lot-to-lot variation in reagents and age of the sample. For BT, one set of triplicate results had a level outside of the ALP by approximately 4.6mmol/mol (64mmol/mol versus the upper limit for the ALP of the median 59.4mmol/mol). However, this discrepancy did not cause a difference in classification in glycaemic status between platforms. For AT, one set of triplicate results had a level outside the ALP by 2mmol/mol (35mmol/mol versus the lower limit of the ALP for the median of 37mmol/mol). This discrepancy did cause a difference in classification in glycaemic status between platforms. There were no evident biological or analytical explanations for these discrepancies.

When monitoring diabetic patients, a difference of 5mmol/mol or more is usually considered significant,13 indicating an improvement or worsening of glycaemic control based on current recommended analytical standards of performance. This difference was used to determine the clinical comparability of results; if the analytical difference between any two results was 5mmol/mol or more, the inter-assay discrepancy may cause inaccurate clinical interpretation. In practice, the 5mmol/mol difference threshold is intended to be the difference between two consecutive results for an individual patient using the same assay, while in this study the same sample was tested three-fold. However it provides a useful, if crude, measure of comparability when considering a patient having serial results using different methods. In the BT group, one triplicate demonstrated a maximum difference of 5mmol/mol (32mmol/mol versus 37mmol/mol) and another 11mmol/mol (53mmol/mol versus 64mmol/mol). In the AT group there was only one set of triplicates with a maximum difference more than 5mmol/mol (35mmol/mol versus 42mmol/mol). The two BT triplicates did not demonstrate a difference in classification of glycaemia probably because the levels were far from the diagnostic cut-offs, whereas the AT triplicate did because levels were around the prediabetes cut-off of ≥41mmol/mol.

There was one result that violated all three measures of concordance: the AT triplicate of 35mmol/mol, 41mmol/mol and 42mmol/mol. There was no biological or analytical explanation.

In New Zealand, a HbA1c of 50mmol/mol or higher is the current recommended diagnostic threshold for diabetes mellitus, and 41–49mmol/mol is the diagnostic range for pre-diabetes.1 Using these cut-offs as a marker for concordance we found three triplicates in the BT group and three in the AT group had a level that misdiagnosed diabetes or pre-diabetes. According to the New Zealand guidelines1 a single HbA1c level is insufficient to diagnose diabetes unless the individual is symptomatic, in which case the HbA1c is expected to be clearly high. In asymptomatic individuals a repeat HbA1c or an additional elevated plasma glucose is needed to confirm the diagnosis. Assuming assay performance remains constant it is probable that a repeat HbA1c measurement for three of the samples, samples 1 and 6 by IA and sample 7 by CZE, would correctly reclassify the patients’ glycaemic status, aligning them with levels from the other two relevant platforms.

The assays

The IFCC model for quality targets was used to investigate the performance of HbA1c assays by 24 manufacturers in 17 countries and 2,166 laboratories.3 It is a model based on total error in which performance criteria are derived from sigma (σ) metrics. The criterion to be met was a total allowable error (TAE) of 5mmol/mol at the 2σ level for a HbA1c level of 50mmol/mol.3,15 Bio-Rad D-100 and Sebia Capillarys 2 Flex-Piercing met the IFCC criteria. Multiple Roche assays were grouped together because of their large variety, but regardless they also met the IFCC criterion for accuracy; the failure in the precision criterion was deemed due to inter-laboratory variability.3 Lenters-Westra and English16 in their comparative study of three HbA1c assays including Roche Tina-quant Gen.3 on Cobas c513 (the assay used by the community lab) demonstrated an analytical precision of 2% and 2.1% at 46mmol/mol and 72mmol/mol for the Roche assay, respectively.16

HbA1c reliability to assess dysglycaemia in patients with haemoglobinopathies and other conditions

In healthy adults, the normal physiological proportion of HbA (α2β2) is 95–98% and of HbF <1%.17 Most haemoglobinopathies, of which there are approximately 1,300, are clinically silent and do not affect RBC survival or interfere with HbA1c measurement, but the proportion of HbA can vary in the presence of some.18 Individuals with HbS/ alpha thalassemia typically have 60–70% HbA (α2β2)19 depending on the genotype, and individuals with HbE/alpha thalassemia would have 20–30% HbA.20 Also depending on the genotype, those with HbS/beta thalassemia have a maximum of 30% HbA.21

Clinical cut-offs for HbA1c assume normal physiological RBC lifespan and proportion of HbA. Therefore a significantly shorter lifespan, or lower proportion of HbA relative to serum/intraerythrocytic glucose9 are expected to influence interpretation of measured HbA1c, rendering it potentially unreliable in the diagnosis and monitoring of dysglycaemia regardless of method used. Furthermore, individuals with HbS/beta thalassemia and HBT often receive regular transfusions rendering the measured HbA1c an inaccurate reflection of glycaemic status. When the proportion of HbA is abnormal, RBC survival is short, or in case of regular blood transfusions, HbA1c may be unreliable by any method and should not be used. It is appropriate to assess glycaemia in such individuals by other means such as plasma glucose, fructosamine or (if available) glycated albumin.

Variation in RBC survival and glucose entry into RBC22 can occur even in patients with normal haematology23 and no known chronic disease or drug effects, and most of the genetic variation in HbA1c is due to non-glucose factors as shown in gene studies.24 This is an inherent, and often unrecognised, limitation of the test in all patients, not just those with known causes of unreliability, and limits the ability of HbA1c to assess estimated average glucose (eAG).25

Haematological parameters

HbA1c concentration is inversely correlated with MCV, Hb and MCH in premenopausal females without haemoglobinopathies or diabetes; the strongest correlation being with MCH.26 A similar association with MCV and MCH was found in 1,315 adults without haemoglobinopathies, diabetes or serum creatinine ≥120umol/L.27

Our study was suggestive, albeit not statistically, of an inverse correlation between HbA1c concentration and Hb, but demonstrated no correlation between HbA1c and MCV. Unlike findings in individuals with normal Hb,26 our population consisted of samples from both genders, and pre- and post-menopausal females, the majority of whom had a haemoglobinopathy. Increased RBC turnover, haemolysis and blood transfusions contribute to lowering of HbA1c. These conditions would not be represented in studies of healthy individuals but would skew our findings towards lower HbA1c at smaller MCVs.

Strengths, weaknesses and significance of this study

A strength of this study is its specific focus on haemoglobinopathies with large numbers of patients; the findings add insight and inform local clinicians on diagnosis and monitoring of their diabetic patients who have haemoglobinopathies. Furthermore, as the timeframe for the study included changes of reagent lots and calibrators, transportation, different methodologies and operators, the differences are a reflection of the ‘real life’ scenario in laboratories.

There were limited numbers of samples without a haemoglobinopathy, but differences in patients without known haemoglobinopathies were not the focus of this study and inter-method correlation would be expected to be, if anything, better. All New Zealand laboratories participate in two-weekly inter-laboratory comparisons of HbA1c using de-identified human samples of mostly normal HbA (Waikato QAP scheme, New Zealand)28 and in a wider Australasian RCPA QAP comparison using both human and non-human matrix.12,20 Laboratories using all three methods compared in this study are represented in the Waikato scheme, with 11 participants using IEC, six using IA and two using CZE.28

Although numbers for some uncommon haemoglobinopathies were small, making it difficult to make firm conclusions, this was a reflection of the numbers in the local community. Results of this study do not imply that the same correlation applies to other haemoglobinopathies or assay formats (even if from the same manufacturer) but do indicate that for the assays in this study the difference in HbA1c levels for the haemoglobinopathies included is acceptable in the large majority of cases.

Conclusion

This study was a pragmatic three-way comparison for HbA1c in non-pregnant adults, most of whom had a haemoglobinopathy. Of the six triplicate sets of results from the thalassemia groups where inter-platform results caused a change in classification of glycaemia, three may have re-classified the patients had they been repeated. There was no statistical correlation between HbA1c, and Hb and MCV.

While HbA1c measurement technology has improved greatly, there is no perfect test and HbA1c measurement still demonstrates inter-platform discrepancies, particularly using the IA method. However, the difference between methods was found not to be clinically significant in the large majority of cases. Even for individual methods there can be differences in analytical bias and imprecision between laboratories. Assuming individual methods remain well controlled in each laboratory, and results are interpreted within the broader clinical context, this implies that HbA1c results from the three assays in this study can be viewed cumulatively, interchanged, and importantly trends can be determined for monitoring purposes. However, caution should be exercised in a small proportion of patients when a diagnosis is sought and where, regardless of methodology used, the result may not accurately reflect dysglycaemia. Further studies with larger sample size, blood glucose levels for correlation and standard reference material for better assessment of accuracy are recommended.

Summary

Abstract

Aim

To determine whether glycated haemoglobin (HbA1c) results from three commonly used platforms can be interpreted cumulatively and used interchangeably in individuals with common haemoglobinopathies. A secondary goal was to assess the relationship between HbA1c concentrations, and haemoglobin and mean corpuscular volume in this population.

Method

One hundred and forty-five samples, mostly with haemoglobinopathies, were tested by each of: Roche Gen.3 Cobas c513, Capillarys 2 Flex-Piercing and Bio-Rad D-100 platforms. Statistical comparisons and limits of performance based on biological variation, international recommendations, and local diagnostic cut-offs were drawn upon to determine comparability of results.

Results

Inter-platform measurements were not significantly different for the large majority of results. The four HbA1c results that showed maximum discrepancy between triplicates had the following abnormalities: heterozygous haemoglobin S/ beta thalassemia, heterozygous haemoglobin S/ alpha thalassemia, beta thalassemia trait and alpha thalassemia trait. Six triplicates of results in the thalassemia groups (7.5% of thalassemia samples) had levels that misclassified patients’ glycaemic status. There was no correlation between HbA1c concentration and mean corpuscular volume, and a weak negative correlation between HbA1c concentration and haemoglobin concentration.

Conclusion

HbA1c concentrations measured by Cobas c513, Capillarys 2 Flex-Piercing and the Bio-Rad D-100 were found to be comparable in the large majority of samples. While discordance was due to assay imprecision in some cases, in others no biological or analytical explanation could be found.

Author Information

Samarina MA Musaad, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland; George Chan, Consultant Haematologist, Haematology Department Labtests, Healthscope, Auckland; Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland.

Acknowledgements

The authors would like to thank and acknowledge Alek Latinovic and Sian Horan, medical laboratory scientists in the departments of chemical pathology at Labtests and Southern Community Laboratories respectively; and Linda Henderson, Technical Specialist in the department of specialist chemistry at Labplus, Auckland City Hospital. Alek tediously coordinated and organised sample distribution between the laboratories and all three scientists ensured timely testing and communication of results.

Correspondence

Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, PO Box 12049, Auckland 1642.

Correspondence Email

cam.kyle@labtests.co.nz

Competing Interests

Dr Kyle is Consultant Pathologist to Labtests, Auckland.

1. New Zealand Society for the Study of Diabetes. Executive Summary of position statement. http://www.nzssd.org.nz/HbA1c/2.%20NZSSD%20exec%20summary%20diagnosis%20of%20diabetes%20Sept%202011%20final.pdf (accessed 21/10/2019).

2. Little RR, Rohlfing CL, Sacks DB for the National Glycohemoglobin Standardization Program (NGSP) Steering Committee. Status of Hemoglobin A1c Measurement and Goals for Improvement: From Chaos to Order for Improving Diabetes Care. Clin Chem. 2011; 57:205–214.

3. The EurA1c Trial Group. EurA1c: the European HbA1c Trial to Investigate the Performance of HbA1c Assays in 2166 Laboratories across 17 Countries and 24 Manufacturers by Use of the IFCC Model for Quality Targets. Clin Chem. 2018; 64:8.

4. Wekcamp C. HbA1c: A Review of Analytical and Clinical Aspects. Clin Chem. 2013; 33:393–400.

5. Rhea JM, Molinaro R. Pathology Consultation on HbA1c Methods and Interferences. Am J Clin Pathol. 2014; 141:5–16.

6. HbA1c Assay Interferences. The National Glycohemoglobin Program (NGSP) http://www.ngsp.org/interf.asp (accessed 05/08/2019).

7. English E, Idris I, Smith G, Dhatariya K, et al. The effect of anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic review. Diabetologia. 2015; 58:1409–1421.

8. Selvaraj N, Bobby Z, Sathiyapriya V. Effect of lipid peroxides and antioxidants on glycation of hemoglobin: an in vitro study on human erythrocytes. Clin Chim Acta. 2006; 366:190–5.

9. Coban E, Ozdogan M, Timuragaoglu A. Effect of iron deficiency anemia on the levels of hemoglobin A1c in nondiabetic patients. Acta Haematol. 2004; 112:126–8.

10. Hashimoto K, Noguchi S, Morimoto Y, Hamada H et al. A1C but not serum glycated albumin is elevated in late pregnancy owing to iron deficiency. Diabetes Care. 2008; 31:1945–8.

11. Stats New Zealand Tatauranga Aotearoa. http://www.stats.govt.nz/information-releases/2018-census-totals-by-topic-national-highlights#download-data (accessed 21/10/2019).

12. The Royal College of Pathologists of Australasia Quality Assurance Programs. RCPA QAP http://rcpaqap.com.au/ (accessed 05/08/2019).

13. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, et al. Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus. Clin Chem. 2002; 48(3):436–472.

14. Human Tissue Act 2008 New Zealand Legislation. Parliamentary Counsel Office. http://www.legislation.govt.nz/act/public/2008/0028/latest/versions.aspx (accessed 10/02/2020).

15. Weykamp C, Siebelder C. Evaluation of Performance of Laboratories and Manufacturers Within the Framework of the IFCC model for Quality Targets of HbA1c. J. Diabetes Sci. Technol.2018; 12(4):747–752.

16. Lenters-Westra E, English E. Evaluating new HbA1c methods for adoption by the IFCC and NGSP reference networks using international quality targets. Clin Chem Lab Med. 2017; 55(9):1426–1434.

17. Burtis CA, Ashwood ER, Bruns DE, editors. Tietz Textbook of Clinical Chemistry and Molecular Diagnostics. 4th ed. Missouri: Elsevier Saunders; 2006.

18. Dunphy CH, editor. Molecular Pathology of Hematolymphoid Diseases. NY, USA: Springer; 2010. Chapter 37, Molecular Pathology of Hemoglobin and Erythrocyte Membrane disorders; page 473.

19. Huisman THJ. Trimodality in Tie Percenetages of β Chain Variants in Heterozy-Gotes: The Effect of the Number of Aczive HB α Structural Loci. Hemoglobin. 1977; 1:349–382.

20. Fucharoen S, Weatherall D. Hemoglobin E Disorders. In Steeinberg MH, Forget BG, HIggs DR, Weatherall EJ, editors. Disorders of Hemoglobin 2nd edition. Cambridge University Press, 2009. p. 417–433.

21. Sergeant GR, Ashcroft MT, Serjeant BE, Milner PF. The Clinical Features of Sickle-Cell /p Thalassaemia in Jamaica. Br J Haematol. 1973; 24:19.

22. Khera PK, Clinton HJ, Carruthers A, Lindsell CJ, et al. Evidence for Interindividual Heterogeneity in the Glucose Gradient Across the Human Red Blood Cell Membrane and Its Relationship to Hemoglobin Glycation. Diabetes 2008; 57:2445–2452.

23. Cohen RM, Franco RS, Khera PK, Smith EP, et al. Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood 2008 112:4284–4291 doi:10.1182/blood-2008-04-154112

24. Soranzo N. Genetic Determinants of Variability in Glycated Hemoglobin (HbA1c) in Humans: Review of Recent Progress and Prospects for Use in Diabetes Care Curr Diab Rep 2011; 11:562–569 DOI 10.1007/s11892-011-0232-9

25. Barth JH, Marshall SM, Watson ID. Consensus meeting on reporting glycated haemoglobin and estimated average glucose in the UK: report to the National Director for Diabetes, Department of Health. Ann Clin Biochem 2008; 45:343–344. DOI: 10.1258/acb.2008.200815

26. Koga M, Morita S, Saito H, Mukai M, et al. Association of erythrocyte indices with glycated haemoglobin in premenopausal women. Diabet Med. 2007; 24(8):843–847.

27. Simmons D, Hlaing T. Interpretation of HbA1c: association with mean cell volume and haemoglobin concentration. Diabet Med. 2014; 31(11):1387–1392.

28. Quality Assurance Programme Waikato District Health Board. http://lab.waikatodhb.health.nz/assets/QCItems/Quality-Assurance-Programme.pdf (accessed 5/8/2019).

Contact diana@nzma.org.nz
for the PDF of this article

View Article PDF

HbA1c is HbA that is glycated at the N-terminal valine of one or both of the beta chains. Glucose can also attach to other amino acids, but HbA1c constitutes the majority of the glycated Hb forms and is the one used for the diagnosis and monitoring of diabetes since 2011.1 From a clinician’s perspective it is important to understand the differences, if any, between results produced by different methods and laboratories. These differences will dictate whether results from different laboratories produce clinically comparable results and can be used interchangeably for screening, diagnosis and monitoring of diabetes. If so, they can be displayed and interpreted cumulatively to show trends in guiding management.

HbA1c measurement was standardised in 2001 by the International Federation for Clinical Chemistry (IFCC), resulting in marked improvement in inter-assay and inter-laboratory performance and clinical utility.2 Furthermore, the US National Glycohemoglobin Standardisation Program (NGSP) and the IFCC set out quality targets to assist manufacturers, external quality assurance (EQA) programmes and laboratories with accuracy and precision goals for their assays.3,4

Laboratories choose an assay based on many factors, including clinical and technical performance, workflow and cost-effectiveness. Having different HbA1c assays in a region has its advantages and disadvantages and is common. From an analytical perspective, differences between assays and laboratories are inevitable and can be explained by the difference in measurement methodology, the inherent imprecision and bias associated with individual assays, and changes in lots of reagents and calibrators for all assays. While standardisation and quality targets have improved the quality of results and inter-assay comparability, there can still be differences between assays that can preclude interchangeability, especially in certain patient groups.

Differences between HbA1c assays can be accentuated, or appear exclusively, in the presence of haemoglobinopathies, the existence of which the laboratory, clinician or even the patient in some cases, may not be aware of. Haemoglobinopathy is the term used for congenital abnormalities of haemoglobin (Hb) covering both Hb production abnormalities, ie, thalassaemia, and structural Hb variants. These discrepancies or interferences are due to the different ways the measurement methodology interacts with the abnormal haemoglobin molecule in the case of structural variants, eg, sickle cell Hb (HbS) in which the glutamic acid amino acid in position six of the beta chain is replaced with valine.5 On the other hand, in case of a production abnormality, eg, beta thalassemia in which there is reduction in the formation of the beta chain due to mutations, deletions or insertions, the interference is related to the quantitative reduction in beta chains and associated increase in the proportion of more primal Hb forms like foetal Hb (HbF). The latter is composed of two alpha and two gamma chains. Large quantities of HbF are known to affect several assays.6

Red blood cell (RBC) volume, measured as the mean corpuscular volume (MCV), is also reported to influence the rate of glycation of Hb.7 The smaller the volume, eg, iron deficiency anaemia, the larger the surface-area-to-volume ratio, which increases the proportion of glycated Hb. Furthermore, nutritional iron deficiency is associated with higher HbA1c levels through an effect on the glycoxidation process itself. 8–10

The effect of haemoglobinopathies on measurement and interpretation of HbA1c is a matter of clinical concern because of the increasingly diverse ethnicities in New Zealand, including from the Pacific, Middle East, and parts of Asia where the frequency of certain haemoglobinopathies is greatly increased.11

Methods

One hundred and fifty-four (154) samples sourced from the community laboratory in Auckland (Labtests) were tested between September 2017 and August 2018. Samples chosen were residual ethylenediaminetetraacetic acid (EDTA) samples for which a HbA1c had been requested, and were either known to have a haemoglobinopathy as documented in the laboratory database, or on which a haemoglobinopathy screen had been requested on the basis of clinical or laboratory suspicion. Reasons may include a relevant family history, otherwise unexplained anaemia, unexplained low mean corpuscular haemoglobin (MCH) or MCV, or an apparent discrepancy between blood glucose and HbA1c concentrations.

HbA1c (requested by the clinician) was tested on the day of collection in the community laboratory. Aliquots were made and forwarded to two other (hospital) laboratories for HbA1c testing. If not tested on the same day, the other laboratories stored the samples at 4oC until tested within the week.

HbA1c in the community laboratory was tested on a Roche Gen.3 Cobas c513 (Germany) immunoassay (IA) platform. The second laboratory tested HbA1c by capillary zone electrophoresis (CZE) using the Capillarys 2 Flex-Piercing (Sebia, France) platform and the third laboratory tested HbA1c by ion exchange chromatography (IEC) on the Bio-Rad D-100 (California) platform. These three methods are named IA, CZE and IEC hereon. All three assays are traceable to the IFCC and NGSP reference methods. Coefficients of variation (CV), a measure of precision of an assay, at the time of testing were as follows: 3.9% and 2% at 34mmol/mol and 78mmol/mol respectively for the IA; 3% at both HbA1c levels for CZE; and 2.1% and 1.4% at 34mmol/mol and 81mmol/mol respectively for IEC.

All three methods use potassium EDTA, the preservative used in all samples in this study. All three laboratories are accredited to ISO15189 standard by International Accreditation New Zealand.

During the study period there were, depending on the assay, approximately 2–10 lots of calibrators, 3–4 lots of quality control material, and 1–6 lots of reagents used.

Samples from children (less than 18 years old) and from pregnant women were excluded. Apart from basic demographics (age, sex), the state of diabetes, MCV and Hb at the time of the study, other clinical details were not sought. Therefore individuals with haematological malignancies, chronic illness or other conditions that may affect the RBC lifespan or haematological parameters were not excluded.

Concordance between methods was assessed using non-parametric Passing-Bablok regression analysis and Bland-Altman difference plots for all results. Passing-Bablok was chosen because it allows for imprecision in both comparators (x and y). The regression equation y=a+bx was used to interpret the relation between the two relevant assays. A 95% confidence interval (CI) containing 0 for the intercept a, and 1 for the slope b indicates statistical comparability within the measured concentration range. A 95% CI that does not include 0 for a, and 1 for b indicates a systematic difference and proportional bias respectively.

Bland-Altman difference plots evaluate bias between the mean differences and estimate an agreement interval within which 95% of the differences between the two relevant methods fall.

For the two largest groups of haemoglobinopathies, beta thalassaemia trait (BT) and alpha thalassaemia trait (AT), three more methods were used, namely: the Royal College of Pathologists of Australasia’s Quality Assurance Programme (RCPA-QAP) allowable limits of performance (ALP) for HbA1c based on biological variation,12 the change in HbA1c concentration (mmol/mol) that is considered clinically significant,13 and whether the difference between results can cause misclassification of an individual as diabetic or prediabetic according to current New Zealand guidelines diagnostic cut-offs.1 Assessing the median of triplicate results against the RCPA-QAP ALP was considered a mathematically useful way of assessing concordance since the RCPA-QAP is used by many local and Australasian laboratories. None of the three methods was considered a reference method and no assumption of accuracy was made.

The relation between HbA1c and Hb and MCV was assessed using correlation scatter plots and the correlation coefficient r to determine the strength of the correlation, or lack of.

Ethics

Ethics approval was not needed because this study was for laboratory quality assurance purposes only on the assay originally requested, and patients were anonymised.14

Statistical analysis

MEDCAL statistical software, MedCal software Ltd. version 19, was used to compare the three methods for HbA1c measurement with Passing-Bablok and Bland-Altman analyses, and for correlation analysis for the relation between HbA1c, and Hb and MCV.

Results

A total of 154 samples were tested, of which nine were subsequently excluded: three patients were pregnant, two were children, two patients did not have data on MCV and Hb, one sample was repeated and one result was lost. A total of 145 samples from 145 patients were included in the final analysis, which included 82 (56%) females. Ages ranged from 18 to 86 years. There were 44 individuals with known diabetes mellitus. Table 1 summarises the number of samples for the haemoglobinopathies found in the study. Twenty-five samples did not show an apparent abnormality (NAA) based on the HbH inclusion body test used in our institution. The latter is a phenotypic test for alpha thalassaemia and cannot exclude a single or two gene deletion alpha thalassaemia.

Table 1: Haemoglobinopathy findings in increasing order of frequency.

BT: beta thalassemia trait; AT: alpha thalassemia trait; NAA: no apparent abnormality- cannot exclude one or two gene deletion alpha thalassemia; N: normal haemoglobinopathy screen; HbE: haemoglobin E; HbD: haemoglobin D; HbS: haemoglobin S; HBT: homozygous beta thalassemia; *other: Heterozygous HbS, HbS/ beta thalassemia, alpha thalassemia /HbE trait, Haemoglobin H (HbH) disease, hereditary persistent HbF, and heterozygous haemoglobin C (HbC).

Figures 1a to 3b summarise the correlation and differences between the three platforms for all results.

Figure 1a: Correlation between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

The 95% CI for the intercept a (-1.50) is -3.17–1.0, and for the slope b (1.05) is 1.0–1.2. This correlation indicates statistical comparability between CZE and IEC within the measured concentration range.

Figure 1b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and ion exchange chromatography (IEC).

Upper limit of agreement (ULOA) was 4.4 (95% CI: 3.88–5.0); mean difference was 0.6 (95% CI: 0.27–0.88); and lower limit of agreement (LLOA) was -3.3 (95% CI: -2.76–-3.84).

Figure 2a: Correlation between capillary zone electrophoresis (CZE) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.99–1.0 and for the slope b (1.0) was 1.0–1.05 indicating statistical comparability between CZE and IA within the measured concentration range.

Figure 2b: Bland-Altman plot for the difference between capillary zone electrophoresis (CZE) and immunoassay (IA)

The ULOA was 5.8 (95% CI: 5.08–6.6); mean difference was 0.4 (95% CI: 0.09-0.94); and LLOA was -5.1 (95% CI: -5.86–-4.23).

Figure 3a: Correlation between ion exchange chromatography (IEC) and immunoassay (IA).

The 95% CI for the intercept a (0.0) was -1.0–1.06 and for the slope b (1.0) was 0.97–1.0 indicating statistical comparability between IEC and IA within the measured concentration range.

Figure 3b: Bland-Altman plot for the difference between ion exchange chromatography (IEC) and immunoassay (IA).

The ULOA was 5.7 (95% CI: 4.95–6.56); the mean difference was -0.2 (95% CI: -0.79–0.33); and LLOA was -6.1 (95% CI: -5.34–-6.95).

There was only one patient with HbS/ beta thalassemia in the study. The patient had recurrent sickle cell crises and regular blood transfusions. However, the same blood sample was tested by the three platforms in this study. Table 2 summarises HbA1c levels for the four most discordant results in the study.

Table 2: Details of the four results with maximum differences between the highest and lowest results.

CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; HbS: haemoglobin S.

Results for beta thalassemia trait and alpha thalassemia trait

The RCPA-QAP ALP for HbA1c are +/-4 up to 45mmol/mol and 8% for levels higher than 45mmol/mol.12 These ALPs were used to estimate concordance between the three assays for alpha thalassemia trait (AT) (42 samples) and beta thalassemia trait (BT) (38 samples) groups using the median of the triplicates as true value. For the majority of triplicates the measured values fell within the range for the median that was based on RCPA-QAP ALP, and the maximum difference between the measured values would not have misclassified glycaemic status. Table 3 summarises results for the discordant results.

Table 3: Discordant haemoglobin A1c results for beta thalassemia trait (BT) (total 5) and alpha thalassemia trait (AT) (total 3).

ALP: allowable limits of performance; CZE: capillary electrophoresis; IA: immunoassay; IEC: ion exchange chromatography; RCPA: Royal College of Pathologists of Australasia; QAP: quality assurance programme. * Maximum difference between results was ≥5mmol/mol.

Six of a total of 80 samples with thalassemia (7.5%) had discrepant results that could misclassify glycaemic status.

Results for the relation between HbA1c concentration, and mean corpuscular volume and haemoglobin

MCV and Hb concentration for 131 individuals were available for analysis. The range of MCV was 57–92fL and that for Hb 54–160g/L. The median HbA1c level for the three assays was plotted against the MCV and Hb; Figures 4 and 5 respectively.

Figure 4: The relationship between the mean corpuscular volume (MCV) and the median HbA1c concentrations.

Correlation coefficient r is 0.021 (95% CI -0.15–0.19) (P=0.8). This signifies no linear correlation between HbA1c concentration and MCV in this population.

Figure 5: The relationship between the haemoglobin (Hb) concentration and the median HbA1c concentrations.

Correlation coefficient r is -0.09 (95% CI was -0.26–0.08) P=0.3. This signifies no statistical linear correlation between HbA1c concentration and Hb in this population in spite of a suggestive trend.

Discussion

Concordance

Passing-Bablok comparisons demonstrated acceptable concordance between the three platforms. Bland-Altman plots demonstrated the highest mean difference of 0.6 mmol/mol between CZE and IEC with CZE being the higher of the two sets. However, these same platforms had the narrowest LOA compared to their individual differences with the IA platform.

Differences in two of the four discordant triplicate results were due to low IA results, one due to a high IA result, and the fourth due to a discrepantly low IEC result (Table 2). The higher concordance between CZE and IEC may be because both techniques exploit molecular charge for separation while IA techniques are based on the binding between antigen and antibody and exposure of a unique epitope (glycated amino group of N-terminal valine of the beta chain).

For purposes of this study, the median of each triplicate was presumed closest to the “true” value. The RCPA-QAP ALP was then used to determine comparability of the three results. If the results fell within the ALP range for the median value it was assumed that the results were clinically comparable. This is an unbiased criterion because the RCPA-QAP ALP is based purely on what is considered an acceptable analytical imprecision, for the known biological variation of the analyte, without the additional factors in play including the difference in methodologies, precision and calibration, lot-to-lot variation in reagents and age of the sample. For BT, one set of triplicate results had a level outside of the ALP by approximately 4.6mmol/mol (64mmol/mol versus the upper limit for the ALP of the median 59.4mmol/mol). However, this discrepancy did not cause a difference in classification in glycaemic status between platforms. For AT, one set of triplicate results had a level outside the ALP by 2mmol/mol (35mmol/mol versus the lower limit of the ALP for the median of 37mmol/mol). This discrepancy did cause a difference in classification in glycaemic status between platforms. There were no evident biological or analytical explanations for these discrepancies.

When monitoring diabetic patients, a difference of 5mmol/mol or more is usually considered significant,13 indicating an improvement or worsening of glycaemic control based on current recommended analytical standards of performance. This difference was used to determine the clinical comparability of results; if the analytical difference between any two results was 5mmol/mol or more, the inter-assay discrepancy may cause inaccurate clinical interpretation. In practice, the 5mmol/mol difference threshold is intended to be the difference between two consecutive results for an individual patient using the same assay, while in this study the same sample was tested three-fold. However it provides a useful, if crude, measure of comparability when considering a patient having serial results using different methods. In the BT group, one triplicate demonstrated a maximum difference of 5mmol/mol (32mmol/mol versus 37mmol/mol) and another 11mmol/mol (53mmol/mol versus 64mmol/mol). In the AT group there was only one set of triplicates with a maximum difference more than 5mmol/mol (35mmol/mol versus 42mmol/mol). The two BT triplicates did not demonstrate a difference in classification of glycaemia probably because the levels were far from the diagnostic cut-offs, whereas the AT triplicate did because levels were around the prediabetes cut-off of ≥41mmol/mol.

There was one result that violated all three measures of concordance: the AT triplicate of 35mmol/mol, 41mmol/mol and 42mmol/mol. There was no biological or analytical explanation.

In New Zealand, a HbA1c of 50mmol/mol or higher is the current recommended diagnostic threshold for diabetes mellitus, and 41–49mmol/mol is the diagnostic range for pre-diabetes.1 Using these cut-offs as a marker for concordance we found three triplicates in the BT group and three in the AT group had a level that misdiagnosed diabetes or pre-diabetes. According to the New Zealand guidelines1 a single HbA1c level is insufficient to diagnose diabetes unless the individual is symptomatic, in which case the HbA1c is expected to be clearly high. In asymptomatic individuals a repeat HbA1c or an additional elevated plasma glucose is needed to confirm the diagnosis. Assuming assay performance remains constant it is probable that a repeat HbA1c measurement for three of the samples, samples 1 and 6 by IA and sample 7 by CZE, would correctly reclassify the patients’ glycaemic status, aligning them with levels from the other two relevant platforms.

The assays

The IFCC model for quality targets was used to investigate the performance of HbA1c assays by 24 manufacturers in 17 countries and 2,166 laboratories.3 It is a model based on total error in which performance criteria are derived from sigma (σ) metrics. The criterion to be met was a total allowable error (TAE) of 5mmol/mol at the 2σ level for a HbA1c level of 50mmol/mol.3,15 Bio-Rad D-100 and Sebia Capillarys 2 Flex-Piercing met the IFCC criteria. Multiple Roche assays were grouped together because of their large variety, but regardless they also met the IFCC criterion for accuracy; the failure in the precision criterion was deemed due to inter-laboratory variability.3 Lenters-Westra and English16 in their comparative study of three HbA1c assays including Roche Tina-quant Gen.3 on Cobas c513 (the assay used by the community lab) demonstrated an analytical precision of 2% and 2.1% at 46mmol/mol and 72mmol/mol for the Roche assay, respectively.16

HbA1c reliability to assess dysglycaemia in patients with haemoglobinopathies and other conditions

In healthy adults, the normal physiological proportion of HbA (α2β2) is 95–98% and of HbF <1%.17 Most haemoglobinopathies, of which there are approximately 1,300, are clinically silent and do not affect RBC survival or interfere with HbA1c measurement, but the proportion of HbA can vary in the presence of some.18 Individuals with HbS/ alpha thalassemia typically have 60–70% HbA (α2β2)19 depending on the genotype, and individuals with HbE/alpha thalassemia would have 20–30% HbA.20 Also depending on the genotype, those with HbS/beta thalassemia have a maximum of 30% HbA.21

Clinical cut-offs for HbA1c assume normal physiological RBC lifespan and proportion of HbA. Therefore a significantly shorter lifespan, or lower proportion of HbA relative to serum/intraerythrocytic glucose9 are expected to influence interpretation of measured HbA1c, rendering it potentially unreliable in the diagnosis and monitoring of dysglycaemia regardless of method used. Furthermore, individuals with HbS/beta thalassemia and HBT often receive regular transfusions rendering the measured HbA1c an inaccurate reflection of glycaemic status. When the proportion of HbA is abnormal, RBC survival is short, or in case of regular blood transfusions, HbA1c may be unreliable by any method and should not be used. It is appropriate to assess glycaemia in such individuals by other means such as plasma glucose, fructosamine or (if available) glycated albumin.

Variation in RBC survival and glucose entry into RBC22 can occur even in patients with normal haematology23 and no known chronic disease or drug effects, and most of the genetic variation in HbA1c is due to non-glucose factors as shown in gene studies.24 This is an inherent, and often unrecognised, limitation of the test in all patients, not just those with known causes of unreliability, and limits the ability of HbA1c to assess estimated average glucose (eAG).25

Haematological parameters

HbA1c concentration is inversely correlated with MCV, Hb and MCH in premenopausal females without haemoglobinopathies or diabetes; the strongest correlation being with MCH.26 A similar association with MCV and MCH was found in 1,315 adults without haemoglobinopathies, diabetes or serum creatinine ≥120umol/L.27

Our study was suggestive, albeit not statistically, of an inverse correlation between HbA1c concentration and Hb, but demonstrated no correlation between HbA1c and MCV. Unlike findings in individuals with normal Hb,26 our population consisted of samples from both genders, and pre- and post-menopausal females, the majority of whom had a haemoglobinopathy. Increased RBC turnover, haemolysis and blood transfusions contribute to lowering of HbA1c. These conditions would not be represented in studies of healthy individuals but would skew our findings towards lower HbA1c at smaller MCVs.

Strengths, weaknesses and significance of this study

A strength of this study is its specific focus on haemoglobinopathies with large numbers of patients; the findings add insight and inform local clinicians on diagnosis and monitoring of their diabetic patients who have haemoglobinopathies. Furthermore, as the timeframe for the study included changes of reagent lots and calibrators, transportation, different methodologies and operators, the differences are a reflection of the ‘real life’ scenario in laboratories.

There were limited numbers of samples without a haemoglobinopathy, but differences in patients without known haemoglobinopathies were not the focus of this study and inter-method correlation would be expected to be, if anything, better. All New Zealand laboratories participate in two-weekly inter-laboratory comparisons of HbA1c using de-identified human samples of mostly normal HbA (Waikato QAP scheme, New Zealand)28 and in a wider Australasian RCPA QAP comparison using both human and non-human matrix.12,20 Laboratories using all three methods compared in this study are represented in the Waikato scheme, with 11 participants using IEC, six using IA and two using CZE.28

Although numbers for some uncommon haemoglobinopathies were small, making it difficult to make firm conclusions, this was a reflection of the numbers in the local community. Results of this study do not imply that the same correlation applies to other haemoglobinopathies or assay formats (even if from the same manufacturer) but do indicate that for the assays in this study the difference in HbA1c levels for the haemoglobinopathies included is acceptable in the large majority of cases.

Conclusion

This study was a pragmatic three-way comparison for HbA1c in non-pregnant adults, most of whom had a haemoglobinopathy. Of the six triplicate sets of results from the thalassemia groups where inter-platform results caused a change in classification of glycaemia, three may have re-classified the patients had they been repeated. There was no statistical correlation between HbA1c, and Hb and MCV.

While HbA1c measurement technology has improved greatly, there is no perfect test and HbA1c measurement still demonstrates inter-platform discrepancies, particularly using the IA method. However, the difference between methods was found not to be clinically significant in the large majority of cases. Even for individual methods there can be differences in analytical bias and imprecision between laboratories. Assuming individual methods remain well controlled in each laboratory, and results are interpreted within the broader clinical context, this implies that HbA1c results from the three assays in this study can be viewed cumulatively, interchanged, and importantly trends can be determined for monitoring purposes. However, caution should be exercised in a small proportion of patients when a diagnosis is sought and where, regardless of methodology used, the result may not accurately reflect dysglycaemia. Further studies with larger sample size, blood glucose levels for correlation and standard reference material for better assessment of accuracy are recommended.

Summary

Abstract

Aim

To determine whether glycated haemoglobin (HbA1c) results from three commonly used platforms can be interpreted cumulatively and used interchangeably in individuals with common haemoglobinopathies. A secondary goal was to assess the relationship between HbA1c concentrations, and haemoglobin and mean corpuscular volume in this population.

Method

One hundred and forty-five samples, mostly with haemoglobinopathies, were tested by each of: Roche Gen.3 Cobas c513, Capillarys 2 Flex-Piercing and Bio-Rad D-100 platforms. Statistical comparisons and limits of performance based on biological variation, international recommendations, and local diagnostic cut-offs were drawn upon to determine comparability of results.

Results

Inter-platform measurements were not significantly different for the large majority of results. The four HbA1c results that showed maximum discrepancy between triplicates had the following abnormalities: heterozygous haemoglobin S/ beta thalassemia, heterozygous haemoglobin S/ alpha thalassemia, beta thalassemia trait and alpha thalassemia trait. Six triplicates of results in the thalassemia groups (7.5% of thalassemia samples) had levels that misclassified patients’ glycaemic status. There was no correlation between HbA1c concentration and mean corpuscular volume, and a weak negative correlation between HbA1c concentration and haemoglobin concentration.

Conclusion

HbA1c concentrations measured by Cobas c513, Capillarys 2 Flex-Piercing and the Bio-Rad D-100 were found to be comparable in the large majority of samples. While discordance was due to assay imprecision in some cases, in others no biological or analytical explanation could be found.

Author Information

Samarina MA Musaad, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland; George Chan, Consultant Haematologist, Haematology Department Labtests, Healthscope, Auckland; Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, Auckland.

Acknowledgements

The authors would like to thank and acknowledge Alek Latinovic and Sian Horan, medical laboratory scientists in the departments of chemical pathology at Labtests and Southern Community Laboratories respectively; and Linda Henderson, Technical Specialist in the department of specialist chemistry at Labplus, Auckland City Hospital. Alek tediously coordinated and organised sample distribution between the laboratories and all three scientists ensured timely testing and communication of results.

Correspondence

Campbell Kyle, Consultant Chemical Pathologist, Chemical Pathology Department, Labtests, Healthscope, PO Box 12049, Auckland 1642.

Correspondence Email

cam.kyle@labtests.co.nz

Competing Interests

Dr Kyle is Consultant Pathologist to Labtests, Auckland.

1. New Zealand Society for the Study of Diabetes. Executive Summary of position statement. http://www.nzssd.org.nz/HbA1c/2.%20NZSSD%20exec%20summary%20diagnosis%20of%20diabetes%20Sept%202011%20final.pdf (accessed 21/10/2019).

2. Little RR, Rohlfing CL, Sacks DB for the National Glycohemoglobin Standardization Program (NGSP) Steering Committee. Status of Hemoglobin A1c Measurement and Goals for Improvement: From Chaos to Order for Improving Diabetes Care. Clin Chem. 2011; 57:205–214.

3. The EurA1c Trial Group. EurA1c: the European HbA1c Trial to Investigate the Performance of HbA1c Assays in 2166 Laboratories across 17 Countries and 24 Manufacturers by Use of the IFCC Model for Quality Targets. Clin Chem. 2018; 64:8.

4. Wekcamp C. HbA1c: A Review of Analytical and Clinical Aspects. Clin Chem. 2013; 33:393–400.

5. Rhea JM, Molinaro R. Pathology Consultation on HbA1c Methods and Interferences. Am J Clin Pathol. 2014; 141:5–16.

6. HbA1c Assay Interferences. The National Glycohemoglobin Program (NGSP) http://www.ngsp.org/interf.asp (accessed 05/08/2019).

7. English E, Idris I, Smith G, Dhatariya K, et al. The effect of anaemia and abnormalities of erythrocyte indices on HbA1c analysis: a systematic review. Diabetologia. 2015; 58:1409–1421.

8. Selvaraj N, Bobby Z, Sathiyapriya V. Effect of lipid peroxides and antioxidants on glycation of hemoglobin: an in vitro study on human erythrocytes. Clin Chim Acta. 2006; 366:190–5.

9. Coban E, Ozdogan M, Timuragaoglu A. Effect of iron deficiency anemia on the levels of hemoglobin A1c in nondiabetic patients. Acta Haematol. 2004; 112:126–8.

10. Hashimoto K, Noguchi S, Morimoto Y, Hamada H et al. A1C but not serum glycated albumin is elevated in late pregnancy owing to iron deficiency. Diabetes Care. 2008; 31:1945–8.

11. Stats New Zealand Tatauranga Aotearoa. http://www.stats.govt.nz/information-releases/2018-census-totals-by-topic-national-highlights#download-data (accessed 21/10/2019).

12. The Royal College of Pathologists of Australasia Quality Assurance Programs. RCPA QAP http://rcpaqap.com.au/ (accessed 05/08/2019).

13. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK, et al. Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus. Clin Chem. 2002; 48(3):436–472.

14. Human Tissue Act 2008 New Zealand Legislation. Parliamentary Counsel Office. http://www.legislation.govt.nz/act/public/2008/0028/latest/versions.aspx (accessed 10/02/2020).

15. Weykamp C, Siebelder C. Evaluation of Performance of Laboratories and Manufacturers Within the Framework of the IFCC model for Quality Targets of HbA1c. J. Diabetes Sci. Technol.2018; 12(4):747–752.

16. Lenters-Westra E, English E. Evaluating new HbA1c methods for adoption by the IFCC and NGSP reference networks using international quality targets. Clin Chem Lab Med. 2017; 55(9):1426–1434.

17. Burtis CA, Ashwood ER, Bruns DE, editors. Tietz Textbook of Clinical Chemistry and Molecular Diagnostics. 4th ed. Missouri: Elsevier Saunders; 2006.

18. Dunphy CH, editor. Molecular Pathology of Hematolymphoid Diseases. NY, USA: Springer; 2010. Chapter 37, Molecular Pathology of Hemoglobin and Erythrocyte Membrane disorders; page 473.

19. Huisman THJ. Trimodality in Tie Percenetages of β Chain Variants in Heterozy-Gotes: The Effect of the Number of Aczive HB α Structural Loci. Hemoglobin. 1977; 1:349–382.

20. Fucharoen S, Weatherall D. Hemoglobin E Disorders. In Steeinberg MH, Forget BG, HIggs DR, Weatherall EJ, editors. Disorders of Hemoglobin 2nd edition. Cambridge University Press, 2009. p. 417–433.

21. Sergeant GR, Ashcroft MT, Serjeant BE, Milner PF. The Clinical Features of Sickle-Cell /p Thalassaemia in Jamaica. Br J Haematol. 1973; 24:19.

22. Khera PK, Clinton HJ, Carruthers A, Lindsell CJ, et al. Evidence for Interindividual Heterogeneity in the Glucose Gradient Across the Human Red Blood Cell Membrane and Its Relationship to Hemoglobin Glycation. Diabetes 2008; 57:2445–2452.

23. Cohen RM, Franco RS, Khera PK, Smith EP, et al. Red cell life span heterogeneity in hematologically normal people is sufficient to alter HbA1c. Blood 2008 112:4284–4291 doi:10.1182/blood-2008-04-154112

24. Soranzo N. Genetic Determinants of Variability in Glycated Hemoglobin (HbA1c) in Humans: Review of Recent Progress and Prospects for Use in Diabetes Care Curr Diab Rep 2011; 11:562–569 DOI 10.1007/s11892-011-0232-9

25. Barth JH, Marshall SM, Watson ID. Consensus meeting on reporting glycated haemoglobin and estimated average glucose in the UK: report to the National Director for Diabetes, Department of Health. Ann Clin Biochem 2008; 45:343–344. DOI: 10.1258/acb.2008.200815

26. Koga M, Morita S, Saito H, Mukai M, et al. Association of erythrocyte indices with glycated haemoglobin in premenopausal women. Diabet Med. 2007; 24(8):843–847.

27. Simmons D, Hlaing T. Interpretation of HbA1c: association with mean cell volume and haemoglobin concentration. Diabet Med. 2014; 31(11):1387–1392.

28. Quality Assurance Programme Waikato District Health Board. http://lab.waikatodhb.health.nz/assets/QCItems/Quality-Assurance-Programme.pdf (accessed 5/8/2019).

Contact diana@nzma.org.nz
for the PDF of this article

Subscriber Content

The full contents of this pages only available to subscribers.

LOGINSUBSCRIBE