The formation of glycohemoglobin, especially the hemoglobin A_1c (HbA_1c) fraction, occurs when glucose becomes coupled with the amino acid valine in the β-chain of Hb; this reaction is dependent on the plasma concentration of glucose. Since the early 1970s it has been known that diabetics display higher values OF HbA_1C because they have elevated blood glucose concentrations. Thus HbA_1c has acquired a very important role in the treatment and diagnosis of diabetes mellitus. After the introduction of the first quantitative measurement OF HbA_1C, numerous methods for glycohemoglobin have been introduced with different assay principles: From a simple mini-column technique to the very accurate automated high-pressure chromatography and lastly to many automated immunochemical or enzymatic assays. In early days, the results of the quality control reports for HbA_1c varied extensively between laboratories, therefore in United States and Canada working groups (WG) of the Diabetes Controls and Complications Trial (DCCT) were set up to standardize the HbA_1c assays against the DCCT/National Glycohemoglobin Standardization Program reference method based on liquid chromatography. In the 1990s, the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) appointed a new WG to plan a reference preparation and method for the HBA_1c measurement. When the reference procedures were established, in 2004 IFCC recommended that all manufacturers for equipment used in HbA_1c assays should calibrate their methods to their proposals. This led to an improvement in the coefficient of variation (CV%) associated with the assay. In this review, we describe the glycation of Hb, methods, standardization of the HbA_1c assays, analytical problems, problems with the units in which HbA_1c values are expressed, reference values, quality control aspects, target requirements for HbA_1c, and the relationship of the plasma glucose values to HbA_1c concentrations. We also note that the acceptance of the mmol/mol system for HbA_1c as recommended by IFCC, i.e., the new unit and reference ranges, are becoming only slowly accepted outside of Europe where it seems that expressing HbA_1c values either only in per cent units or with parallel reporting of percent and mmol/mol will continue. We believe that these issues should be resolved in the future and that it would avoid confusion if mmol/mol unit for HbA_1c were to gain worldwide acceptance.

Fisher[1] synthetized a molecule named fructosamine in 1889. In the 1950s and 60s, it was reported that carbohydrate residues could become attached to hemoglobin (Hb)[2,3]. In 1967 Holmquist and Schroeder[4] observed that glucose bound to Hb and Rahbar[5] revealed that the electrophoretic fraction containing glucose was higher in blood samples from diabetic subjects than in healthy subjects. Subsequently, Shapiro et al[6] described that many carbohydrate components including glucose could be bound to Hb and that glucose was most efficiently bound to the β-chain of Hb. After Trivelli et al[7] published their quantitative measurement of Hb fractions in human blood, the analysis OF HbA_1C was recognized as being a very important parameter in the assessment of diabetic patients.

In the early years, the results of HbA_1c analysis varied extensively between methods and laboratories[8-12]. Therefore during the 1980s, in United States and Canada working groups (WG) of the Diabetes Controls and Complications Trial (DCCT) were set up; these were originally incorporated within a multicenter, randomized clinical trial designed to compare treatments of insulin dependent diabetes mellitus in the National Glycohemoglobin Standardization Program (NGSP), but subsequently this activity was expanded to standardize the HbA_1c assays to the DCCT/NGSP reference method of liquid chromatography[9].

Later in the 1990s, the International Federation of Clinical Chemistry (IFCC) decided to develop a reference system for the international standardization of HbA_1c/glycohemoglobin measurements[13], which became the basis of the reference laboratory network for HbA_1c. In addition, IFCC set up another WG on Standardization for HbA_1c (1994/1995) in order to develop a primary standard[14] and reference method[15] for worldwide use in the HbA_1c measurement. These were the reasons why from the 1990s, IFCC organized WG to promote the standardization of all types of assays for HbA_1c. The adoption of these recommendations achieved substantial improvements in the analytical aspects and clinical significance of HbA_1c[16,17] so that today reference standards and standardized methods are in widespread use.

Further in 2010, the American Diabetes Association (ADA) emphasized the important role of HbA_1c in the diagnosis of diabetes, setting an analytical cutoff limit value of 6.5% for HbA_1c values[18] if they were expressed in percentage terms. Recently Hanas and John[19] reported the conclusions of the International Consensus Committee 2013 Update that HbA_1c results should be reported by clinical laboratories worldwide in Système Internationale (SI) units (mmol/mol - no decimals) and the corresponding NGSP units (% - one decimal) and recommended strongly to the editors of scientific journals that submitted manuscripts should report HbA_1c values in both SI (IFCC) and NGSP/DCCT units.

This review provides a brief summary of the reaction of glucose with Hb and the possible techniques available for analyzing HbA_1c. It also reviews the quality control, requirements of the target limits and considers issues related to the units and cutoff limits for HbA_1c in relation to the recommendations of IFCC and ADA.

Glucose is the major soluble carbohydrate and it can combine with different protein molecules in blood and tissues depending on the glucose concentration. In blood, Hb is one of these proteins and the β-chain of Hb is the main target of glucose[6]. Glucose is initially non-enzymatically bound to the amino acid valine on the β-chain of Hb via the formation of a reversible aldimine moiety, which then becomes rearranged into the irreversible ketoamine form. The structure of this ketoamine is similar to that of fructosamine[6]. This phenomenon occurs during the whole lifetime of erythrocytes (120 d) and thus the content of the ketoamine in Hb correlates with the age of the erythrocytes and the value of HbA_1c normally represents a mean value reflecting a time period starting from about three months before sampling.

By analyzing Hb components in blood, Trivelli et al[7] demonstrated that the small fraction A_1c was glycated more than the other small fractions A_1a and A_1b and that the amounts of this A_1c fraction were clearly higher in the blood of diabetic patients than in the normal subjects. The fractionation of Hb by a Mono S™ HR5/5 cation exchange column from Pharmacia with 0.01 mol/L malonate buffer for normal and diabetic human blood samples is illustrated in Figure 1[20].

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Fractionation of blood hemoglobin by an automated Mono S™ HR5/5 cation exchange column from Pharmacia with malonate buffer (0. 01 mol/L, pH 5.7, 11 min, room temperature) for normal and diabetic human blood samples is illustrated as a chromatogram (Penttilä et al[10]).

The first methods for the HbA_1c measurements were simple electrophoretic or ion-exchange mini-column chromatography assays, but they were rather soon replaced by many automated techniques, e.g., different liquid chromatography techniques [ion-exchange chromatography (HPLC or FPLC) or affinity chromatography][8-12,20-22] as seen in Tables 1 and 2. The NGSP/DCCT organizations in United States and Canada selected the automated Bio-Rex 70 liquid chromatography assay as the reference method for the HbA_1c measurement[9]. It should be stressed that this equipment has not been available for many years. However, in addition to the NGSP/DCCT system, there were other different assay techniques used for standardization and these had their own standards, e.g., in Japan (NGH) and in Sweden (FPLC with a Mono S column). These methods could be compared to each other and to the new IFCC procedure by using the appropriate master equations[23].

In 1988, an automated immunoassay method for the epitope assays of proteins was developed[24] and then utilized in the assay of HbA_1c. These immunoassay or enzymatic methods have replaced the chromatography methods previously used in clinical laboratories so that today only about 20% to 40% of methods for HbA_1c are based on liquid chromatographic techniques[16,25]. The numbers of immunoassays for HbA_1c based on many different principles are continuously increasing but these techniques usually have a higher inter-laboratory CV% than can be attained with liquid chromatography or enzymatic methods[16,22,25-27]; some of the most important immunoassay methods are listed in Tables 1 and 2.

As noted before, there were marked inter-laboratory differences in the quality assurance results before standardization procedures were adopted[8-12]. It was also reported that by incorporating an extra sample, against which the primary results could be recalculated, and significantly smaller CV% values could be obtained[9,16,17].

After the appearance of the IFCC WG for HbA_1c, a reference system was soon organized in 1996[13], with the goal of achieving standardization of HbA_1c analysis; this formed the basis for the worldwide reference laboratory network to help clinical laboratories in their HbA_1c measurements. The reference preparations which represented the primary and secondary standards of HbA_1c and HbA0 were produced in 1998[14] followed in 2002 by the reference method for the specific measurement of HbA_1c[15]. The final measurements in the reference method were based on the assay of N1-deoxyfructosyl-hemoglobin[28]. The assay consisted of a primary fractionation of the sample by affinity chromatography followed by analysis utilizing either HPLC/electrospray mass spectrometry or HPLC/capillary electrophoresis. The primary and secondary standards and the reference methods and the guidelines about names and units were then introduced to be adopted worldwide by all manufacturers making equipment for analysis of HbA_1c and also for clinical laboratories[13,23,25,29].

The standardization protocols about HbA_1c have been published by many societies of laboratory medicine in their own languages, e.g., ADA in United States[18,29], DGKL in Germany[30], NEQAS in United Kingdom[31], SiBioC in Italy[32], EQUALIS in Sweden[33], Finnish Society of Clinical Chemistry (FSCC) in Finland[34], etc. With respect to the standardization procedures, it should be noted that according to Weykamp et al[35] the units and standardization protocols for the NGSP/DCCT and IFCC procedures are different for NGSP/DCCT and IFCC results and thus the reference values and quality control requirements cannot be of the same size.

The typical problems[36] which interfere with the HbA_1c assays, are attributable to hyperbilirubinemia, hypertriglyceridemia, leukocytosis and many Hb variants. In addition, certain physiological and pathological characteristics such as gestational stage, age, race, pre-symptomatic type 1 diabetes, malaria, iron deficiency, bleedings, transfusions, splenectomy, kidney failure, alcohol abuse, and some drugs may affect the HbA_1c results. The HbA_1c results may be too high in some cases and too low in others (hemolysis, pregnancy).

If erythrocytes have a short life-time, e.g., in hemolytic anemias, this may decrease the HbA_1c results. Some abnormal forms of Hb cause erroneous results (HbF, HbS, HbC, HbD, HbE, etc.) depending on the assay type[37-39]. Furthermore, iron deficiency anemia as well as iron deficiency without anemia may induce elevated HbA_1c values compared with controls even though blood glucose levels are normal[39].

Many of these errors can be very difficult to detect, especially when using immunochemical assays. With some systems such as with liquid chromatography, the erroneous results can be visualized in the chromatograms, and these assays are commonly used for comparison[23,35-40]. However, errors can also be due to the problems in the action or response of insulin which are not directly related to the methods being used to assay HbA_1c[36].

In 2009, prior to the use of the new IFCC system with the accepted name and unit for HbA_1c, questionnaires about which units should be used were sent to the European societies of laboratory medicine and some other societies outside of Europe (mainly in clinical chemistry)[40]. Germany was the first country to adopt mmol/mol exclusively in the daily laboratory practice (Deutsche Vereinte Gesellschaft für Klinische Chemie und Laboratoriums-medizin); this was inaugurated at the start of 1.1.2010. Germany was followed later in 2011 by The Netherlands, Sweden and the United Kingdom. Then gradually the number of mmol/mol reports of HbA_1c increased up to 13 in 2014, representing 25% of all replies from 51 queries. During these years, there was also an increase in the numbers of laboratories (from 9 to 12) reporting HbA_1c values in parallel units, i.e., in both % and mmol/mol. By 2014, a minority of responders (24%) stated that they were using only the mmol/mol whereas nearly every other respondent (49%) stated that values were being expressed not only in mmol/mol but also as % values (Table 3). However, by the end of 2014, ten societies had not responded to the questionnaire, although the e-mail, telefax and mail addresses were taken from the annual catalogues of IFCC.

On the other hand, in the period from 2010 to 2015, the users of % values in the HbA_1c surveys conducted by Labquality Ltd.[41] have gradually but significantly decreased (from 54% to 40%).

The Finnish experience is an example of the difficulty to obtain acceptance of the molar unit recommended by IFCC for expressing HbA_1c. In 2009, the FSCC recommended that clinical laboratories should report the HbA_1c results in parallel, i.e., as mmol/mol and % values according to the recommendations[23]. In 2011, the FSCC proposed that the HbA_1c values should be provided only in mmol/mol. Nonetheless, the issue has not fully resolved due to the opposition from physicians treating diabetics. Thus the following compromise was agreed: In the future, the laboratories will report the HbA_1c values in parallel units, i.e., as mmol/mol and % although the measurements in many laboratories have originally been calculated in mmol/mol[40].

In Finland, one out of five university hospital districts (Pirkanmaa and Kanta-Häme) stopped reporting the % values for HbA_1c from April 2014; in that hospital the HbA_1c values are only being reported in mmol/mol[40]. It is possible that some other Finnish hospital districts will come to the same decision in the near future. In September 2014, the board of FSCC established a new working group to discuss with physicians and with non-laboratory societies and as well as with other interested parties about the issues related to the determination of HbA_1c. During the summer of 2015, FSCC decided to recommend that from 1.1.2016 the HbA_1c values should no longer be expressed as % units; hopefully this recommendation will prove acceptable.

The QC surveys utilize common statistical methods such as the coefficient of variation (CV%)[16-18]. On the basis of these and the standard deviation (SD) values, the new target limits have been calculated for the QC-surveys of HbA_1c around the mean and the reference values expressed as either % or mmol/mol values. This means that within the range mean ± 2SD 95.6% of all results are within the acceptable limits and these correspond to the recent findings of Hyltoft Petersen and Lee[42]. Furthermore, the new report of Nielsen et al[43] about the value of HbA_1c in the classification of diabetes highlights the importance of the exact and precise measurement of HbA_1c.

As mentioned earlier, there was a remarkably high variation in the values being obtained with the different methods being used in laboratories all around the world[8-12]. Weykamp et al[11] were one of the first groups who described the dramatic improvement that could be obtained in the CV% of the HbA_1c results of either normal or diabetic subjects when the primary results were recalculated by incorporating an extra sample with a known HbA_1c value. They had reviewed the corrected results from 110 laboratories using 21 different methods.

In Finland, the glycohemoglobin quality assurance surveys of Labquality started in 1986 with two fresh native EDTA-blood samples, one at the normal level and the second at the diabetic level, which were sent in each survey to all participants[41]. The mean values of the surveys were used as the target values and the acceptable ranges were ± 10% around the mean values. Figure 2A displays the significant improvement in the HbA_1c % values (r = 0.896) from 1994 to 2015 expressed as CV% values, they correspond well to the earlier publications[16,17]. Similarly in Figure 2B from 2009 when the results were reported in mmol/mol units, a further improvement in the CV%s was observed (for results expressed as mmol/mol) (r = 0.642). The mean ± SD values and CV% have been calculated from the surveys of Labquality.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 The total variations of the glycohemoglobin results have been presented both in % and in mmol/mol units from the surveys of Labquallity Ltd. from 1994 to 2015. In the part A have been presented the total variations of the glycohemoglobin in % units and in the part B the total variations of HbA_1c in mmol/mol units.

In the most recent survey (2015) the range mean ± 2SD indicates that 95% of all results are within these limits[9,44,45]. For example, in December 2014, at the HbA_1c level of 6.77%, near the cut-off limit of ADA of 6.5%[18], the SD value was 0.21 for % units and that for mmol/mol units at the level of 50.8 mmol/mol the corresponding SD value was 2.1 mmol/mol. Thus the calculated acceptable limits with the old range of mean HbA_1c 6.77% ± 10% were from 6.09% to 7.48% and with the new narrower limit dating from 1.1.2015 the limits with mean ± 6% were from 6.36% to 7.18%. The latter range corresponds well with that of 6.77% ± 0.21% with the CV of 3.26% for the survey (0.21/6.77 × 100%). If one takes the mean ± 2SD of all results, then the range would be 6.35% to 7.19% in that survey and these would be clearly inside the old range of 10% and similar to the new limits from Labquality. Correspondingly in the same survey of December 2014 for mmol/mol values, the mean HbA_1c value was 50.8 ± 2.1 mmol/mol and the range for mean ± 2SD extended from 46.6 to 55.0 mmol/mol being very close to the new mmol/mol limit from Labquality as mean ± 8%, i.e., from 46.7 to 54.9 mmol/mol with the CV% of 4.13% (2.1/50.8 × 100%). In the future, the target limits for HbA_1c of Labquality may be made more demanding. The latest Finnish results that in 2015 the CV% values for HbA_1c have reached a sufficiently low level (Figure 2) to be comparable to others in the literature both for % and mmol/mol results[9,16-18,46,47].

From the survey conducted in 2002, the HbA_1c values of the EDTA blood samples for the % results were also analyzed by the European Reference Laboratory for Glycohemoglobin (ERLGH)[48]. It was found that there was an almost perfect correlation between the mean HbA_1c % values of Labquality and the values HbA_1c % values of ERLGH (r = 0.997). The same phenomenon was also seen for the mmol/mol results from 2010 when comparing the mean values of Labquality and those of ERLGH (r = 0.973). Since the mean values of the surveys conducted by Labquality are in practice around the same size as the ERLGH values, thus one may be utilized as target values for mmol/mol The findings correspond well with the earlier reports[43,44] despite the fact that there are differences in both units and standardization programs between the NGSP/DCCT and IFCC procedures[35]. The new target limits of the HbA_1c results in 2015 for % results are ± 6% and for mmol/mol results ± 8% around the target value and agree well to those reported by Little et al[17], Weykamp et al[49] and Lindblad and Nordin[50]. In addition, on the basis of CAP surveys, Little et al[17] have reported that at the normal HbA_1c levels (6%-7%) the target value described in % units and with ± 6% around the reference value may be good enough for the diagnosis and follow-up of the treatment of diabetes. Lindblad and Nordin[50] reported that when near to the critical level of HbA_1c of 48 mmol/mol, the maximum allowable difference from the target value should be less than 3.5 mmol/mol, which corresponds to a CV% of 7.3%.

The Uppsala Meeting for Quality Specifications in Clinical Laboratories in early 1990s had one session devoted to the quality of HbA_1c measurements. The Danish clinical chemists[51,52] proposed that a change of 1.0% from the measured % HbA_1c value might suggest that the treatment could be necessary whereas a change of 2.0% would demand that treatment should be initiated. This degree of accuracy is necessary that clinicians could feel confident with their decisions to initiate what could well be life-time therapy. These findings are in accordance with the earlier reports[8,16,20] and with our present findings. On the other hand, recently in 2015, Weykamp et al[45] reported that when using mmol/mol units, the calculated total allowable error could be 4.2 mmol/mol. The HbA_1c results of current Labquality surveys conducted in 2015, i.e., the mean values of 50.8 ± 2.1 mmol/mol are within this value.

On the other hand, a single laboratory should be able to perform HbA_1c % measurements with a total analytical variation (within and between series) from between 1.4% to 3.0% when using liquid chromatography[10,20,21,26], this variation due to the analytical procedure is about 4.5 times lower than the measured biological variation of this parameter[42,53]. Furthermore, also the differences (errors) between frequent measurements should be as small as possible to ensure reliable follow-up of the treatment of those patients in a stable diabetic state[16,17,23,31]. After the proposal from ADA[18] to use a diagnostic cutoff limit of HbA_1c for diabetes, this change demanded that the accuracy of the analytical performance of HbA_1c had to be better than before the setting of this fixed limit, a fact that has also been criticized.

The commonly used reference intervals for HbA_1c in % DCCT/NGHP units have been from 4.0% to 6.0%[16,17,20,49,54]; the corresponding values for mmol/mol have been from 20 to 42 mmol/mol[18,50] as presented in Table 4. The exact values somewhat vary around those in the table but nonetheless are rather close to those published earlier[16,17,20,49,50,54]. In addition, the commonly applied reference values of HbA_1c in published reports expressed as per cent and mmmol/mol are normally stated in the laboratory manuals all around the world as are also the important limits for diagnosis and treatment of diabetes; these manuals also provide the master equations to convert between the system in use and the procedure of IFCC[23,30-35].

The commonly used limit of plasma glucose of 7.0 mmol/L[8,9,17-19] has been widely accepted when there is a need to diagnose diabetes or to monitor the treatment of the patients. Many physicians treating diabetics like to compare the actual plasma glucose values to the HbA_1c values and consider that they are assessing some kind of balance of diabetes. However, it was not until 2002 when Rohlfing et al[55] collected enough published data to be able to devise an equation describing the well-established ratio between blood mean HbA_1c and plasma glucose, i.e., (PG/HbA_1c): Plasma glucose (mmol/L) = [1.98 × HbA_1c(%)] - 4.29 (n = 1439, r = 0.82). However, when inspecting their proposal, it is quite evident that at the same HbA_1c value, e.g., 8.0%, the plasma glucose concentration may maximally vary from 6 to 15 mmol/L and thus the results are not precise, especially at the higher blood HbA_1c levels[49]. In addition, in subjects with impaired glucose tolerance (plasma glucose value from 5.6 to 6.4 mmol/L or blood HbA_1c from 5.7 to 6.4 mmol/mol) the 2 h oral glucose tolerance test may be as good or better at revealing both the insulin sensitivity and disturbance in glucose metabolism than can be achieved with a single value of fasting plasma glucose[56-60]. These are the reasons why many laboratories do not calculate the mean PG/HbA_1c ratios but report glucose and HbA_1c results separately[41]. It has also argued that in fulminant (acute) type I diabetes, the specificity of the analysis of HbA_1c may be doubtful low when compared to the analysis of glycated albumin with a significantly higher specificity[59]. This may have a significant role to make a selection of the analyze type in diagnosing of acute diabetes or prediabetes. It is also important that plasma glucose and blood HbA_1c levels should be assessed in an accredited laboratory if the values are to be used for diagnosis or screening of diabetes[18,19,23,58-62].

Finally, it has been pointed out that in both men and women elevated levels of blood HbA_1c increase the risk of developing cardiovascular disease[63]. In addition, as described in the chapter of analytical problems, many other diseases may affect the HbA_1c results and cause difficulties for the clinicians. Some diseases other than cardiovascular are also associated with HbA_1c[62,63], e.g., diseases altering iron metabolism.

The point-of care instruments (POC) are being continuously introduced for the analysis of HAb_1c[25,64] but there are still many issues associated with their use[64-66]: (1) they have not been universally recommended for the diagnosis of diabetes according to the guidelines[17-19,30]; (2) many POC users are not participating in the quality control programs of their home country; and (3) the POC analyses are difficult to standardize. In the future, in conjunction with the further development of these methods, especially with the adoption of reliable quality systems, POC analyses may achieve real breakthroughs[25].

This assessment of HbA_1c analytical procedures and values indicates that a considerable improvement has occurred during the past 30 years with respect to both their the precision and accuracy and these improvements are still on-going as reflected in reduced assay CV% values. The immunoassay techniques have replaced many chromatographic procedures during this time period.

During 2014/2015, the reports from quality assurance systems have confirmed the marked improvement for the quality of HbA_1c measurements irrespective of whether the results have been expressed in % or mmol/mol units. The acceptance of the mmol/mol system recommended by IFCC for HbA_1c and the new unit and reference ranges are only becoming slowly accepted outside of Europe where it seems that the parallel reporting for HbA_1c will continue. The use of a diagnostic cutoff limit for the HbA_1c value is still not finalized.

The reference analyses from the mean values of the survey results may be used as target values in both % and mmol/mol units assuming that the number of the participating laboratories is high enough to be statistically satisfactory.

The authors also hope that the use of the mmol/mol unit for HbA_1c can gain worldwide acceptance as this would make it much easier to compare results from different studies and remove the possibility of confusion when units are converted from one form to the other.

The authors cordially than all 41 societies which sent replies to the queries in 2009 to 2014. We also thank Labqualty Ltd. for the survey results of HbA_1c for our disposal and Dr. Ewen McDonald for correcting the English language.