Accuracy of HbA1c as Monitored by External Quality Assessment and Compared With Patient Mean Values

Abstract

Background:

The accuracy and trueness of results from a laboratory test, such as the HbA1c test, should not be taken for granted but must be checked continuously. A tool for this is the participation in external quality assessment (EQA) for all laboratories performing the HbA1c-test. An additional possibility to detect changes in trueness is to monitor variations in patient cohort mean or median values that is not explained by changes in treatment or selection of patients.

Methods:

Results reported to an EQA scheme for HbA1c during 20 years have been extracted from Equalis database. The results are compared to current analytical performance specifications (APS) and to the mean HbA1c levels for the Swedish population of persons with type 2 diabetes.

Results:

The accuracy of the HbA1c test has improved during the period. The hospital lab methods used in Sweden now fulfil APS agreed by professional organizations in Sweden. The accuracy for point-of-care tests (POCT) methods vary over time and fulfil APS for some periods. The bias found for some of the methods might explain changes seen in patient mean values for HbA1c in Sweden during the period 2007-2017.

Conclusions:

The global standardization of HbA1c has resulted in an improved comparability for HbA1c-results worldwide. But even small variation in trueness for the methods in use might have important impact on mean HbA1c values for cohorts of patients. When a systematic error is observed for a specific method it is therefore essential that manufacturers correct the method without delay.

Measurement of HbA1c is essential both for monitoring diabetes and for the diagnosis of the disease. External quality assessment (EQA) is an interlaboratory comparison where participating laboratories measure the same sample with value unknown for the laboratory and compare the result with results from users of the same method, or other methods, and with a true value if this is available. EQA is thus a tool for individual laboratories by which they can show the performance of their HbA1c method and it is also a tool to evaluate the trueness of measurements achieved by different in vitro diagnostic (IVD) products.

In 2007, a global recommendation¹ was made to make all HbA1c results metrologically traceable to one and the same reference. This was possible because the IFCC reference measurement procedure (RMP) had been approved in 2002² and two reference materials (pure HbA1c and pure HbA0) had been certified by the European Commission Joint Research Centre.³ Prior to this unification of the metrological traceability, the HbA1c results worldwide had been traceable to one of three different standardizations with results on three different scales; the JDS/JSCC in Japan, the in house Mono S procedure in Sweden, and the NGSP standardization in the United States as well as for a large part of the rest of the world.⁴

Equalis, a nonprofit organization in Sweden, has organized EQA for HbA1c since 1994. This report is a summary of the experiences from the surveys, and describes some observations in relation to the change of traceability for the results from the Mono S procedure to the IFCC RMP.

Units and Measurement Scales

Results traceable either to the IFCC RMP (unit “mmol/mol”) or to NGSP (unit “NGSP%”) represent measurements on two different scales. The use of the unit “mmol/mol” for results traceability to IFCC RMP was recommended by the IFCC committee on nomenclature, properties and units⁵ and was also suggested to distinguish between the two scales and make a dual reporting possible, which has also been agreed upon.¹ However, reporting of results on both NGSP and IFCC scales, as recommended 2007, increases the risk for confusion, such as when results are transferred to patient records. The fact that precision in terms of CV% seemingly is higher for results on the NGSP scale than on the IFCC scale might also be confusing.⁶ Therefore, the HbA1c-results reported to physicians and patients usually have been restricted to one of the scales.

The Swedish EQA Scheme

During the first years of surveys lyophilized samples were distributed. This type of material was soon discovered to lack commutability, and since 1997 only fresh, pooled, left over whole blood samples with EDTA as additive have been used. Commutability is an important property of material used for EQA and describes how well the material mimics the properties of clinical specimens. Commutability is essential to make results from different measurement methods comparable through the results in an EQA scheme.

The frequency of distributions in the EQA scheme was 12 per year until 2003 and has thereafter been 10 times per year. Normally one sample is distributed each distribution. All Swedish hospital laboratories have participated as well as a majority of the point-of-care laboratories. One of the purposes of the Equalis scheme was to monitor the adherence to the specific national traceability for results to the Mono S procedure and scale. Today, when all HbA1c results are traceable to IFCC RMP and reported with the unit “mmol/mol,” any laboratory in the world can participate, if it is possible to distribute and measure the fresh sample material in due time.

The sample material is collected and prepared on Monday to Tuesday, distributed by surface mail in ambient temperature on Wednesday. The samples need to be refrigerated upon reception at the laboratory, and HbA1c measured within 7 days. Approximately half of the participants measure the samples already the first day after distribution, and a further 30% in two days after the day of distribution.

During the period 1994-2010, the target values for the sample materials in the Equalis EQA surveys was calculated as the mean of a selected and representative group of laboratories using the Mono S procedure. Between 2011 to 2014 the target was set by IFCC’s secondary reference methods (SRM) at the European reference laboratory (ERL) in the Netherlands.⁷ From 2015 until today the target value has again been results from the Mono S procedure, after conversion to the IFCC scale. The target values are set by the laboratory at the Sahlgrenska hospital, Gothenburg, currently representing the Mono S dedicated comparison method in the IFCC network. The results from the Mono S procedure in the IFCC HbA1c Certification Programme are in good agreement with the RMP. The reason for the shift from the SRM to the Mono S procedure for target values in the EQA scheme was formal; to accredit the HbA1c-scheme according to the accreditation standard ISO 17043 the laboratory performing the target measurements must be accredited. However, results from the IFCC’s SRM at ERL are still collected for careful comparison with the results from the Mono S procedure within the scheme.

The number of participating instruments in the scheme has increased from 300 in 1997 to more than 800 in 2017. Data for this report have been extracted from the Equalis database which contained more than 107 000 HbA1c results in December 31 2017. For the statistics all the results deviating more than 40% from the target value have been trimmed. Approximately 0.5% of the reported data have been outside the trimming limits, most of them appearing during the first four years (1994-1997) when the assays were less standardized, and around 2011 due to reporting mistakes regarding the new IFCC units. All calculations for this report have been performed using Microsoft Excel.

The majority of results reported in the EQA scheme are from the two POCT methods in clinical use in Sweden, Alere Afinion and Siemens DCA Vantage (Table 1). It should be noted that the number of results reported in the EQA scheme is not proportional to the number of patient results produced by each method. On the contrary, the hospital methods with few results in the EQA scheme produce more patient results than the POCT methods. A rough estimate made by Equalis indicates that the methods producing most of the patient results over the last years have been, in decreasing order; Bio-Rad Variant II Turbo, Tosoh G7/G8, Roche Tina-quant and Siemens DCA Vantage.

Table 1.

The Number of Equalis EQA Results Reported per Year for the Last Four Years With Various Methods.

Method for HbA1c	2014	2015	2016	2017
Abbott Architect enzymatic	1	10	10	10
Alere Afinion	2046	2417	2791	3825
Bio-Rad D-10	29	26	20	19
Bio-Rad D-100			8	20
Bio-Rad Variant Turbo	158	166	165	157
Mono S in house HPLC	28	19	20	18
Roche Tina-quant	99	104	116	109
Sebia CAPILLARYS				7
Siemens DCA Vantage	4276	4495	4386	3977
Tosoh G7/G8	129	129	126	129

All HbA1c results reported without fully declared method, or from some very small method groups, have been excluded.

The feed-back reports from Equalis to each participant in the EQA scheme contain statistics for the different method groups represented in the survey (Figure 1) together with a graphical presentation of the results of the individual laboratory for the last 10 surveys and a comparison with the analytical performance specification (APS).

Figure 1.

Report summary from round 2017:10. The dashed red lines are quality goal limits (APS₂₀₁₀) for individual results.

The Mono S procedure served until 2010 as the basis for the standardization of HbA1c methods in Sweden. All HbA1c results reported in the Swedish health care system should be metrological traceable to the Mono S procedure, either through a direct calibration of devices to results with the Mono S procedure, or with a checked and agreed upon conversion algorithm for each method.

In 2007 the work started for worldwide transition from the old calibrations and traceability chains to the new IFCC RMP traceability. In Sweden the transition took place January 1, 2011, from which time all results were to be traceable to IFCC reference measurement procedure according to each manufacture’s specific calibration routine, and reported with the unit “mmol/mol.”

The Mono S procedure is, contrary to other procedures, not standardized by a reference material, but only by a careful standardization of the procedure itself. The relation between the results from the Mono S procedure and the IFCC RMP was established in 2004 ⁴ and is well described by the master equations for dedicated comparison methods. Results that were, and still are, produced with the in house Mono S procedure are not calibrated with a reference material, but made traceable to the IFCC RMP via a once established relation. Thus, all HbA1c-results from the Mono S procedure are recalculated by the following equation: HbA1c(IFCC) = HbA1c(Mono S) × 10.45-10.62. This relation deviates slightly from the published master equation (HbA1c IFCC = HbA1c(Mono S) × 10.11-8.94). The difference is small (corresponding to 0.3 mmol/mol at the level of 50 mmol/mol) and is due to a careful recalculation in 2004 of the data from comparisons between the Mono S procedure and the IFCC RMP for fresh frozen whole blood samples. The constant relation between results from the Mono S procedure and the IFCC RMP has then been verified.⁸

In this report all HbA1c results produced prior to 2011 with the unit “Mono S%” have, for comparison reasons, been converted by the equation above and presented with the unit “mmol/mol.”

Analytical Performance Specifications

In 1997 the first quality goal for the measurements, or APS, was agreed upon by a working group with representation from the Swedish societies for Diabetology, Endocrinology, Clinical Chemistry, and, among others, the patient organization “Diabetesforbundet.”⁹ The APS was formulated in terms of precision; the between laboratory variation should be less than 3% coefficient of variation (CV) and the within laboratory precision less than 3% CV, for results with % units on the Mono S scale. It was also decided that the between laboratory variation should be monitored by EQA. In late 2010, when the HbA1c results should be made traceable to IFCC, the former working group asked Swedish Society for Clinical Chemistry and Equalis to reformulate the APS to retain the high analytical quality for results reported on the IFCC scale and, if possible, improve it further. The new APC (APS₂₀₁₀) became accuracy based and stated that 95% of results reported from one participant, or for a method group over time (usually a year), should be within “1.5 mmol/mol ± 1.65 × 0.025 × HbA1c mmol/mol.”¹⁰ The equation was inspired by the concept of total error with an allowable bias of 1.5 mmol/mol and imprecision of 2.5% and one-sided 95% confidence interval. The limits were also believed to be close to what was technically possible to achieve and relevant from a clinical point of view. The APS₂₀₁₀ corresponds to a total allowable error of 3.6 mmol/mol (0.33 NGSP%) at the HbA1c level of 50 mmol/mol (6.7 NGSP%).

The Swedish National Diabetes Registry (NDR)

Today, the NDR covers clinical and laboratory data for most of the persons with diabetes in Sweden.¹¹ Clinical HbA1c results are thus registered, and measures like mean HbA1c-values, or fraction of persons with HbA1c > 70 mmol/mol, can easily be extracted for cohorts with different kinds of diabetes, from different regions, and so on from the database. From the start in 1996 and until today the number of persons with HbA1c results recorded in the registry has increased from 25 000 to more than 400 000. A summary of the outcome of the registry has recently been published.¹²

Results

The fraction of EQA results that fulfils APS₂₀₁₀, has increased over time (Figure 2). Since 2013 results from the hospital laboratory methods have fulfilled the APS₂₀₁₀. The results from the two POCT methods have also improved, and for 2017 the Alere Afinion HbA1c fulfilled the APS₂₀₁₀.

Figure 2.

Fraction of results fulfilling analytical performance specifications for hospital labs and POCT labs. From 2008 the results from the two POCT methods in use are shown for the specific devices.

The bias (mmol/mol) in relation to the target values in the EQA scheme for each method group and year are summarized in Table 2.

Table 2a.

Mean Deviation Expressed in IFCC Units (±95% Confidence Interval Limits) from HbA1c-Target per Year and Method Group.

	Year
Method	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Abbott Architect enzymatic											−0.42 (0.71)	0.13 (0.49)	−0.58 (0.63)
Alere Afinion	−2.17 (0.87)	−2.50 (0.57)	−2.52 (0.28)	−2.44 (0.20)	−3.07 (0.21)	−1.11 (0.16)	−0.45 (0.14)	−1.01 (0.12)	−2.92 (0.13)	−1.70 (0.10)	0.13 (0.10)	−0.14 (0.06)	−0.18 (0.05)
Bio-Rad D-10							0.91 (0.82)	2.14 (0.84)	−0.16 (1.16)	0.38 (0.52)	1.79 (0.63)	1.30 (0.68)	0.71 (0.74)
Bio-Rad D-100												0.35 (0.83)	0.18 (0.64)
Bio-Rad Variant II	0.82 (0.35)	1.08 (0.26)	1.61 (0.31)	0.06 (0.60)	0.33 (0.55)	−0.33 (0.57)	3.64 (1.87)
Bio-Rad Variant Turbo				1.83 (0.54)	0.54 (0.40)	1.52 (0.28)	2.17 (0.30)	1.65 (0.22)	0.33 (0.34)	−0.12 (0.22)	0.65 (0.20)	1.28 (0.17)	0.92 (0.18)
Bio-Rad Variant Turbo*									0.53 (0.43)
Bio-Rad Variant Turbo*							0.68 (0.36)
Mono S HPLC in house	0.00 (0.27)	0.00 (0.26)	0.00 (0.27)	0.00 (0.29)	0.00 (0.30)	0.00 (0.21)	0.64 (0.42)	1.02 (0.54)	0.34 (0.61)	0.40 (0.86)	−0.33 (0.41)	0.21 (0.24)	0.29 (0.43)
NycoCard-HbA1c	0.75 (0.72)	2.03 (0.82)	2.85 (2.33)
Roche Tina-quant			0.40 (0.61)	1.58 (0.42)	1.54 (0.35)	0.31 (0.32)	−0.03 (0.40)	−0.17 (0.40)	−0.81 (0.37)	−0.88 (0.31)	0.29 (0.31)	−0.15 (0.34)	−0.50 (0.30)
Roche Tinaquant/Unimate	−0.47 (0.66)	0.31 (0.49)	1.40 (0.95)
Sebia CAPILLARYS													0.71 (0.73)
Siemens DCA 2000	−1.88 (0.08)	−1.92 (0.07)	−2.06 (0.07)	−2.31 (0.09)	0.04 (0.20)	−1.20 (0.16)	0.14 (0.32)
Siemens DCA Vantage			−0.99 (0.79)	−0.61 (0.13)	−0.55 (0.10)	−1.44 (0.07)	−0.08 (0.07)	−0.06 (0.06)	−0.03 (0.08)	−0.93 (0.07)	−0.41 (0.06)	1.39 (0.06)	2.07 (0.06)
Tosoh G7-G8	0.28 (0.47)	0.12 (0.35)	0.13 (0.28)	−0.57 (0.40)	0.33 (0.45)	−0.18 (0.29)	3.39 (0.43)	2.71 (0.43)	1.77 (0.68)	1.08 (0.32)	1.22 (0.32)	1.38 (0.37)	0.87 (0.30)
Tosoh G7-G8*								2.00 (0.45)
Tosoh G7-G8**									1.17 (0.68)	1.04 (0.35)	1.12 (0.35)	1.74 (0.44)	1.26 (0.40)
Tosoh G7-G8***							0.43 (0.28)	1.01 (0.30)	1.01 (0.34)	0.30 (0.34)	0.48 (0.35)	0.80 (0.26)	1.70 (0.59)
Source for target	Mono S	Mono S	Mono S	Mono S	Mono S	Mono S	IFCC SRM	IFCC SRM	IFCC SRM	IFCC SRM	Mono S exp	Mono S exp	Mono S exp

Cells with blue background show a negative bias of more than 1 mmol/mol, cells with red background show a positive bias of more than 1 mmol/mol, outside the confidence interval. The sources of the target values in the EQA surveys are explained on the last line of the table.

^*Measurements after calibration with calibrators from the MCA Laboratory in Netherlands. **Measurements with system calibrated with target values from a Mono S procedure. ***Measurements after calibration with calibrator material from Bio-Rad.

At the time for the transition of the traceability from the Mono S procedure to the IFCC RMP in 2011 an unexpected positive bias (>3.0 mmol/mol, 0.27 NGSP%) appeared for results from both the Bio-Rad and the Tosoh HPLC methods. A large fraction of results from these methods, both representing a large volume of patient results, therefore did not fulfil APS₂₀₁₀. Despite discussions and investigations in collaboration with the manufacturers the reason for the bias was never fully clarified. Some of the concerned users of these methods decided to change the metrological traceability for their results by using alternative calibration procedures, for example, by aligning the method with the Mono S procedure, or by using a calibrator material from other sources than from the manufacturer of the device.

If the sample material used in an EQA scheme is commutable it can be assumed that the results concerning systematic errors also reflect systematic errors for clinical HbA1c results. Therefore, persons monitored for HbA1c in 2010-2011 with results from the Tosoh G7-G8, are supposed to experience a positive bias of 3.4 mmol/mol (0.31 NGSP%) in addition to changes due to real changes of HbA1c for these persons. The bias has been reduced over the years to 0.9 mmol/mol (0.08 NGSP%) in 2017. A person monitored by Roche Tina-quant HbA1c has experienced a variable bias between 1.6 mmol/mol (0.15 NGSP%) in 2008 and −0.9 mmol/mol (0.08 NGSP%) in 2014.

Mean HbA1c Values From the Swedish National Diabetes Registry

Yearly mean HbA1c values in NDR between 1996 and 2017 for subjects with type 2 diabetes in primary health care and in special clinics are shown in Figure 3. Data for the period until 2015 have been published earlier.¹²

Figure 3.

Mean HbA1c values (mmol/mol left y-axis, NGSP% right y-axis) per year for patients in primary health care (red triangles) and T2D in special clinics (blue circles).

During the period 1996-2007 the mean HbA1c value decreased almost constantly, but for the period 2007-2012 it increased by 2 mmol/mol. After 2012 the mean values have decreased again. The same trends are seen for persons with T1D (data not shown). Patients treated within special clinics have generally a more severe T2D than persons treated in primary health care, which explain the higher mean HbA1c values for this group.

Discussion

A commutable reference material is crucial for EQA. During the last 20 years Equalis has used fresh pooled left over whole blood samples. To our knowledge no other type of sample material is more commutable than fresh whole blood.

There are many models to express the outcome of EQA-results in relation to APS. One model is the sigma scale,^13,14 where sigma is defined as [(“total allowable error” – “absolute value of bias”)/“SD for reproducibility”]. The benefit of the sigma scale is the possibility to compare performance for different measurands. However, when only one measurand is concerned, in this case HbA1c, the fraction of results fulfilling analytical performance specifications in relation to an accuracy goal is easy to understand and describe graphically and is selected for this report (Figure 2).

The accuracy of HbA1c results has improved over time thanks to the efforts to standardize the measurements. Today the accuracy of the methods in use in Sweden seems in general to be good, and the hospital laboratory methods do fulfil the agreed APS₂₀₁₀. The accuracy for the POCT methods varies from one year to another. This variation in accuracy is larger than the variation for the hospital methods, and might indicate a lot-lot variation in the production of the reagent cassettes. With less lot-lot variation it is possible also for the POCT methods currently in use in Sweden to fulfil APS₂₀₁₀.

The number of different assays used for HbA1c in Sweden (with a population of 10 million) is limited, compared to the number of assays in use worldwide. The data presented here do only represent the performance of the laboratories participating in the scheme and should not be taken as representative for the assays as such outside the scheme.

Other studies show good agreement between the HbA1c methods used on the European market.¹⁵ A recent comparison also confirms that the results from Swedish laboratories agree well with results from laboratories in other European countries.¹⁶ The global standardization of HbA1c with traceable results to the IFCC RMP thus seems to have been successful.

Decreasing mean HbA1c values for persons in NDR might easily be interpreted as a consequence of improved treatment, while the reverse is true when the mean HbA1c values increase. From 1996 to 2007 the mean HbA1c for patients in NDR decreased. During this period the treatment of diabetes was intensified as an outcome of the NGSP and DCCT-studies, that most likely resulted in a decreased mean HbA1c for patients. However, the increase between 2007and 2012 of the mean HbA1c by 2.0 mmol/mol (0.18 NGSP%) has no obvious reasons, and is not the result of a known change in treatment policy (Figure 3). Several explanations can be speculated on; the number of persons in the registry increased during this period from 170 000 to 350 000 with a different selection of patients, the mean body weight of the persons increased, and the number of patients with non-European ethnicity increased. After 2012 the mean HbA1c has again decreased, with no obvious change in treatment policy.

The increase of mean HbA1c observed for the persons in Diabetes Registry between 2007 and 2012 co-insides in time with the introduction of the IFCC reference measurement procedure for HbA1c worldwide. Such an effect on the HbA1c level was unexpected, after the careful studies of the reference measurement procedure that were made in advance.

Knowing that a majority of clinical HbA1c results have been produced by the two methods Bio-Rad Variant II Turbo and Tosoh G7-G8, both of which methods showed significant positive bias in relation to target values in EQA after 2007, it is likely that the major part of the increase of mean HbA1c for subjects with diabetes population was due to method bias, and not due to any true change in mean HbA1c for the patients. In some way the implementation of the traceability to IFCC RMP affected some of the methods already in 2007, which also affected the results traceability to the Mono S procedure in Sweden. A positive bias for the Bio Rad Variant and Tosoh HPLC assays during this period was also noted in CAP surveys,¹⁷ as well as in other studies.^18,19

The decrease in mean HbA1c for persons with T2D in the registry between 2012 and 2017 might reflect a correction of a bias introduced 2007-2012. During this period the positive bias in relation to target values in the EQA scheme also decreased (Tables 2a, 2b). However, the mean HbA1c level 2017 for persons with T2D is still 1 mmol/mol (0.1 NGSP%) higher than the value 2007. Some of the methods show a positive bias in EQA for 2017, and it is therefore possible that the mean HbA1-level as measured today (2017) still might have a positive bias of 1-2 mmol/mol (0.1-0.2 NGSP%) compared to the level measured 2007. Further studies, for example by reanalysis of stored or biobanked material may shed some light on this.

Table 2b.

Mean Deviation Expressed in NGSP Units (±95% Confidence Interval Limits) From HbA1c-Target Per Year and Method Group.

	Year
Method	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017
Abbott Architect enzymatic											–0.04 (0.07)	0.01 (0.04)	–0.05 (0.06)
Alere Afinion	–0.20 (0.08)	–0.23 (0.05)	–0.23 (0.03)	–0.22 (0.02)	–0.28 (0.02)	–0.10 (0.01)	–0.04 (0.01)	–0.09 (0.01)	–0.27 (0.01)	–0.16 (0.01)	0.01 (0.01)	–0.01 (0.01)	–0.02 (0.00)
Bio-Rad D-10						–0.04 (0.73)	0.08 (0.08)	0.20 (0.08)	–0.02 (0.11)	0.04 (0.05)	0.16 (0.06)	0.12 (0.06)	0.06 (0.07)
Bio-Rad D-100												0.03 (0.08)	0.02 (0.06)
Bio-Rad Variant II	0.07 (0.03)	0.10 (0.02)	0.15 (0.03)	0.01 (0.05)	0.03 (0.05)	–0.03 (0.05)	0.33 (0.17)
Bio-Rad Variant II Turbo				0.17 (0.05)	0.05 (0.04)	0.14 (0.03)	0.20 (0.03)	0.15 (0.02)	0.03 (0.03)	–0.01 (0.02)	0.06 (0.02)	0.12 (0.02)	0.08 (0.02)
Bio-Rad Variant II Turbo*									0.05 (0.04)
Bio-Rad Variant II Turbo**							0.06 (0.03)
Mono S HPLC in house	0.00 (0.02)	0.00 (0.02)	0.00 (0.02)	0.00 (0.03)	0.00 (0.03)	0.00 (0.02)	0.06 (0.04)	0.09 (0.05)	0.03 (0.06)	0.04 (0.08)	–0.03 (0.04)	0.02 (0.02)	0.03 (0.04)
NycoCard-HbA1c	0.07 (0.07)	0.19 (0.08)	0.26 (0.21)
Roche Tina-quant			0.04 (0.06)	0.15 (0.04)	0.14 (0.03)	0.03 (0.03)	–0.00 (0.04)	–0.02 (0.04)	–0.07 (0.03)	–0.08 (0.03)	0.03 (0.03)	–0.01 (0.03)	–0.05 (0.03)
Roche Tinaquant/Unimate	–0.04 (0.06)	0.03 (0.05)	0.12 (0.09)
Sebia CAPILLARYS													0.07 (0.07)
Siemens DCA 2000	–0.17 (0.01)	–0.18 (0.01)	–0.19 (0.01)	–0.21 (0.01)	0.00 (0.02)	–0.11 (0.01)	0.01 (0.03)
Siemens DCA Vantage			–0.09 (0.07)	–0.06 (0.01)	–0.05 (0.01)	–0.13 (0.01)	–0.01 (0.01)	–0.01 (0.01)	0.00 (0.01)	–0.08 (0.01)	–0.04 (0.01)	0.13 (0.01)	0.19 (0.01)
Tosoh G7-G8	0.03 (0.04)	0.01 (0.03)	0.01 (0.03)	–0.05 (0.04)	0.03 (0.04)	–0.02 (0.03)	0.31 (0.04)	0.25 (0.04)	0.16 (0.06)	0.10 (0.03)	0.11 (0.03)	0.13 (0.03)	0.08 (0.03)
Tosoh G7-G8*									0.11 (0.03)	0.10 (0.03)	0.10 (0.03)	0.16 (0.04)	0.11 (0.04)
Tosoh G7-G8**							0.04 (0.03)	0.09 (0.03)	0.09 (0.03)	0.03 (0.03)	0.04 (0.03)	0.07 (0.02)	0.16 (0.05)
Tosoh G7-G8***								0.18 (0.04)
Source for target	Mono S	Mono S	Mono S	Mono S	Mono S	Mono S	IFCC SRM	IFCC SRM	IFCC SRM	IFCC SRM	Mono S exp	Mono S exp	Mono S exp

Cells with blue background show a negative bias of 0.1% or more, cells with red background show a positive bias of more than 0.1% or more, outside the confidence interval. The sources of the target values in the EQA surveys are explained on the last line of the table.

^*Measurements after calibration with calibrators from the MCA Laboratory in Netherlands. **Measurements with system calibrated with target values from a Mono S procedure. ***Measurements after calibration with calibrator material from Bio-Rad.

The impact of small variation of bias for methods is normally not a problem for the individual patient results. Such variation is included in the expected measurement uncertainty for all individual HbA1c-results. But even small variation in bias becomes a problem if data are used to calculate mean values of HbA1c for comparison between different patient cohorts. Given that variation of accuracy and trueness occurs over time for every HbA1c assay, it is necessary to monitor the performance of the assays continuously in EQA schemes. Unfortunately, the measurement method is rarely recorded together with the HbA1c result in patient records and registries. Such information should facilitate the comparison between clinical results and EQA data. The effect of method specific bias is otherwise cumbersome to investigate.²⁰

The worldwide standardization process has so far been a success, ⁸ even though there is a debate regarding the exact trueness for the IFCC primary reference procedures.²¹ Thus, we still don’t know exactly which HbA1c level that is the most correct. The potential remaining issues regarding trueness have minor impact on the uncertainty for individual clinical results but might be important when it comes to comparison of mean values between different patient cohorts or comparisons over time. EQA schemes is the key tool for the individual laboratory, for the manufacturing industry, as well as for clinical use of clinical HbA1c test results, to prove, and, when necessary, improve the accuracy.