Evaluation of glycemia is used for the diagnosis and management of patients with diabetes. Glucose and hemoglobin A1c (HbA1c) provide complementary information and both are used to assess an individual’s glycemic status. The concentration of glucose in the blood indicates the subject’s glycemia at the time of blood sampling. However, blood glucose concentrations are modified by numerous factors, ranging from food ingestion and exercise to stress and medication (1). By contrast, the concentration of HbA1c in the blood reflects the average glucose over the preceding 8–12 weeks. Thus, HbA1c provides an additional criterion for assessing glucose control that is free of the wide diurnal fluctuations that occur with blood glucose. HbA1c has several additional attributes, which render it valuable in the setting of diabetes. These include, but are not limited to, the following: the subject does not need to be fasting, blood can be sampled any time of the day, the sample is stable, and there is very little biological variability (1). These factors, in conjunction with the documentation that HbA1c predicts the development of microvascular (2,3) and macrovascular (4) complications of diabetes, have led to the widespread adoption of HbA1c as integral to the management of patients with diabetes. Guidelines from several prominent clinical organizations recommend that HbA1c be measured at regular intervals in all patients with diabetes (5,6).
The quality of analytical methods for HbA1c initially lagged considerably behind the evidence of its clinical value. Early assays lacked standardization, substantially limiting the use of HbA1c in patient care. Considerable effort was invested to effect standardization, with schemes developed in the 1990s in Japan (7), Sweden (8), and the U.S. (9). All HbA1c results were reported as a percentage of hemoglobin. The subsequent development, almost a decade later, of a reference method for HbA1c (10,11) led to different units (12). This situation has generated considerable controversy as to how HbA1c should be reported. In this review, the background leading up to HbA1c standardization and the development of different units are summarized. The formula for converting patient results from one set of units to the other is provided, and the current state of HbA1c reporting in several countries is indicated.
Identification of HbA1c
In normal adults, hemoglobin usually contains HbA (∼97% of the total) (Table 1), HbA2 (∼2.5%), and HbF (∼0.5%). HbA is made up of four polypeptide chains, two α- and two β-chains. Several posttranslational modifications of hemoglobin have been observed. These include carbamylation, acetylation, sulfation, and glycation. Glycation is the nonenzymatic attachment of a sugar to amino groups of proteins. The phenomenon of glycation, also termed “browning” or the “Maillard reaction,” has been known for over 100 years (13). Early evidence documenting that hemoglobin is glycated was the demonstration in 1955 that small amounts of hemoglobin could be separated from HbA by their migration on electrophoresis in a starch slab (14). Three years later, three minor heme proteins, termed HbA1a, HbA1b, and HbA1c on the basis of their elution, were observed to be resolved when normal human adult hemoglobin was subjected to cation-exchange chromatography (15). (In this review, the term “glycated hemoglobin” is used to refer to the set of all glycated hemoglobins, and “HbA1c” is used to refer to a specific molecular form as described in the text.) The clinical significance of this finding remained obscure for 10 years until Rahbar (16) detected an unusual hemoglobin on electrophoresis of blood from patients with diabetes. This hemoglobin, which was identified as HbA1c, was found to be increased twofold in patients with diabetes compared with healthy individuals (17). At essentially the same time, analysis revealed that HbA1c has a hexose attached covalently to the NH2-terminal valine residue of the β-chain of HbA (Table 2) (18). Several years later, HbA1c was defined by the International Union of Pure and Applied Chemistry as the fraction of the β-chains of hemoglobin that has a stable hexose adduct on the NH2-terminal amino acid valine (19).
It is important to emphasize that glycation of hemoglobin may also occur at sites other than the end of the β-chain, such as the NH2-terminal valine residue of the α-chain as well as lysine residues on the α-chain or β-chain (20). These glycated hemoglobins are referred to as glycated HbA0 or total glycated hemoglobin (GHb) (Table 1). The components of other forms of HbA have been identified. HbA1a1 and HbA1a2, which make up HbA1a, have fructose 1,6-diphosphate and glucose 6-phosphate, respectively, attached to the NH2 terminus of the β-chain (Table 2). The structure of HbA1b, solved by mass spectrometry, contains pyruvic acid linked to the NH2-terminal valine of the β-chain, probably by a ketimine bond (21).
Measurement of HbA1c
Commercial assays to measure HbA1c became available in 1978 (22,23), and the test gained popularity during the 1980s. The first mention of glycated hemoglobin by the World Health Organization was in 1985 when the potential value of its measurement in diabetes was indicated (24). In 1988, the American Diabetes Association (ADA) recommended in its Standards of Medical Care that HbA1c determination should be performed at least semiannually for routine monitoring of patients with diabetes (25). The commercialization of HbA1c assays led to the development of a plethora of methods to measure glycated hemoglobin. The general concept underlying these methods is to separate the glycated from the nonglycated hemoglobin and quantify the amount of each. Techniques that have been used to achieve this separation include those based on charge differences (ion-exchange chromatography, high-performance liquid chromatography [HPLC], electrophoresis, and isoelectric focusing), structural differences (affinity chromatography and immunoassay), or chemical analysis (photometry and spectrophotometry) (26). Analysis by electrophoresis or chemical techniques has become obsolete in most countries. The methods most commonly used today are HPLC and immunoassays (27).
Unfortunately, the early glycated hemoglobin assays suffered from several deficiencies, most notably a lack of standardization. The diverse methods, coupled with the different forms of glycated hemoglobin that were measured, produced a very wide variation in results. For example, in 1993 only 50% of clinical laboratories in the U.S. were reporting glycated hemoglobin as HbA1c (28). The remaining laboratories were measuring, and reporting, HbA1 (21%) or total GHb (29%). One study compared seven glycated hemoglobin methods and observed that results for a single sample varied from 4.0 to 8.1% among the different methods (29). Many people in the diabetes community were unaware of these differences in reporting. The publication of the Diabetes Control and Complications Trial (DCCT) (2) in 1993 provided the impetus necessary to initiate a resolution to this problem.
Documentation of the clinical value of HbA1c
The DCCT evaluated the effect of intensive insulin therapy (compared with conventional insulin therapy) in patients with type 1 diabetes (2). The study documented that maintaining lower blood glucose concentrations (assessed by HbA1c) resulted in a delayed onset and reduced the rate of progression of microvascular complications. (All of the glycated hemoglobin measurements in the DCCT were performed in a single laboratory by an HPLC assay that measured HbA1c. This approach obviated the issue of test variability and established HbA1c as the species of glycated hemoglobin that should be reported.) The risk of retinopathy increased continuously with increasing HbA1c, and a single measure of HbA1c predicted the progression of retinopathy 4 years later. Further analysis of the DCCT data revealed that the mean HbA1c was the dominant predictor of retinopathy progression, and a 10% lower HbA1c concentration (e.g., from 9 to 8.1%) was associated with a 45% lower risk (30). Extended follow-up demonstrated that the incidence of cardiovascular disease was reduced by 42% in patients with lower HbA1c (4). Thus, the DCCT unequivocally established the value of measuring HbA1c in patients with diabetes.
Evidence that lowering HbA1c in patients with type 2 diabetes reduces complications was provided in 1998 with the publication of the UK Prospective Diabetes Study (UKPDS) (3). To ensure that HbA1c results in the UKPDS were comparable to those in the DCCT, an ion-exchange HPLC method calibrated to the DCCT was used. Mean HbA1c was 7.0% in the intensive group compared with 7.9% in the conventional group (3). Notwithstanding the seemingly small difference in HbA1c concentrations between the two groups, significant differences in the rate of complications were found. Analogous to the DCCT, the UKPDS showed that intensive blood glucose control reduced the risk of microvascular complications. Risk reductions of 37% for microvascular disease, 21% for deaths related to diabetes, and 14% for myocardial infarction were observed for each 1% reduction in HbA1c (e.g., from 9 to 8%) (31). Ten-year follow-up demonstrated that the risk of myocardial infarction was significantly lower in patients who had lower HbA1c at the end of the UKPDS (32). Thus, both the UKPDS and DCCT (large, prospective, multicenter, clinical studies) documented that a small change in HbA1c values translates into a large alteration in the risk of diabetes complications in patients with type 1 or type 2 diabetes. It was evident that the state of measurement of HbA1c in routine patient samples was untenable, and accurate, standardized HbA1c testing was essential.
Standardization to DCCT/NGSP numbers
There are >100 methods currently available to measure glycated hemoglobin, and it is vital that they are standardized to report the same (or at least a very similar) result for a single blood sample. Moreover, as mentioned earlier, both the DCCT and UKPDS measured exclusively HbA1c and not other forms of glycated hemoglobin. As a result of the DCCT, the American Association for Clinical Chemistry (AACC) established a committee in 1993 to standardize glycated hemoglobin testing (9). The NGSP (previously called the National Glycohemoglobin Standardization Program) was created 3 years later to execute the protocol developed by the AACC committee. The goal of the NGSP is to standardize glycated hemoglobin test results to those of the DCCT and UKPDS, which established direct relationships between HbA1c concentrations and outcome risks in patients with diabetes. The concept is that all clinical laboratories that measure patient samples should report an HbA1c value equivalent to that reported in the DCCT and UKPDS. A network of laboratories, located in the U.S., Europe, and Japan, has been established to achieve this standardization (28). A brief description of the process follows.
The Central Primary Reference Laboratory (CPRL) measures HbA1c with a Bio-Rex 70 cation-exchange HPLC, which is the method used in the DCCT (28). Three primary reference laboratories (PRLs), which use the same method, serve as backup to the CPRL. The eight second reference laboratories (SRLs) assist manufacturers with calibrating their assays so they will report a value equivalent to that measured in the DCCT. (A calibrator is a material of known concentration that is used to adjust a measurement procedure.) Thus, an HbA1c result of 7.0% in a patient’s blood performed by that assay in a routine clinical laboratory would be essentially identical to a result of 7.0% in the DCCT or UKPDS. If an HbA1c method meets strict accuracy criteria, the manufacturer receives a certificate that is valid for 1 year (28). In 2011, there were 112 methods that had NGSP certification. The ADA recommends that laboratories use only HbA1c assays that are certified by NGSP as traceable to the DCCT reference (33). These assays are listed on the NGSP website (http://www.ngsp.org) and are updated at least annually. The HPLC method used in the CPRL and PRLs is not suitable for routine measurement of patient samples. By contract, the SRLs all use commercially available methods, identical to those used in clinical laboratories. National standardization schemes were developed in Japan and Sweden (7,8). The Japanese and Swedish values were rarely adopted by other countries. Due to its link to the DCCT and UKPDS, the NGSP system was, by an overwhelming margin, the most popular global HbA1c standardization system. NGSP-certified methods are currently used worldwide in the clinical laboratories that measure patient samples, and these results are directly traceable to the DCCT and UKPDS.
The efforts of the NGSP resulted in a considerable improvement in the performance of HbA1c measurement by routine clinical laboratories (28). The fraction of laboratories reporting glycated hemoglobin measurements as HbA1c increased from 50% in 1993 to 80% in 1996 to ∼99% in 2004. This was accompanied by a concurrent improvement in accuracy and reduced variability among laboratories (28).
A working group on HbA1c standardization was established by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) in 1995. The strategy adopted by this committee was different from that of the NGSP. Instead of standardizing to a comparison method, the primary objective of the IFCC committee was to develop a true reference method for HbA1c. This goal was achieved (10). In brief, the approach is to digest hemoglobin with an enzyme (termed endoproteinase Glu-C) that cleaves a hexapeptide off the NH2 terminus of the β-chain of hemoglobin. The glycated hexapeptides are separated from the nonglycated hexapeptides by reverse-phase HPLC. The peptides are then quantified by either mass spectrometry or capillary electrophoresis. HbA1c is measured as the ratio of glycated to nonglycated hexapeptide. Results were initially reported as a percentage. The IFCC Working Group also developed primary reference materials of pure HbA1c and HbA0 (10). These are purified from human whole blood, blended together, and used to calibrate the primary reference measurement system. Secondary reference materials are also produced, and manufacturers use these to calibrate their instruments. A network of reference laboratories (15 at the time of this report) works together to form an IFCC reference system (34). The IFCC reference method is technically demanding, time consuming, and very expensive and is not designed for routine analysis of patient samples. It serves as a reference measurement procedure with well-defined metrologic traceability to a “higher-order method.”
The NGSP and IFCC networks have complementary roles in the HbA1c standardization process. The IFCC provides manufacturers with traceability to a standard of higher metrologic order, and the NGSP defines the limits of acceptability for method performance. The two networks together form a solid basis to establish the accuracy and reliability of HbA1c measurement in a patient’s sample performed in a clinical laboratory anywhere in the world.
Comparison between networks
Analyses using pooled whole-blood samples were conducted to compare the values of HbA1c obtained by the IFCC and NGSP networks. A linear relationship was observed between HbA1c results of the IFCC reference method and the NGSP network (11). The calculated regression equation is NGSP = 0.09148(IFCC) + 2.152. This equation is termed the “master equation” and permits conversion between the two sets of values. Importantly, the HbA1c values measured by the IFCC method are significantly lower than the NGSP values. Moreover, the difference is not constant. For example, an NGSP value of 6.0% is 4.2% in IFCC numbers (difference of 1.8), and 10.0% NGSP is 8.6% IFCC (difference of 1.4). Comparisons of the IFCC network with the Swedish and Japanese standardization schemes revealed a unique regression equation for each network, with lower values obtained by the IFCC method in all cases (11,34). The most likely explanation for the higher values with the HPLC-based standardization schemes is that the HbA1c peak on the chromatogram contains other substances in addition to HbA1c. These observations introduced a conundrum and have generated considerable controversy as to how HbA1c should be reported. The major arguments promulgated for and against each set of units are summarized in Table 3.
Three preeminent clinical diabetes organizations, the ADA, the European Association for the Study of Diabetes (EASD), and the International Diabetes Federation (IDF), attempted to resolve the dispute by agreeing to consider reporting HbA1c with estimated average glucose (eAG) (35,36). A prospective multinational study documented a linear relationship between HbA1c and mean blood glucose (37). Nevertheless, many experts concluded that the large variability between HbA1c and mean blood glucose precludes the use of eAG (38–42). By contrast, others believe that eAG is useful in communicating the extent of glycemic control with patients (43–45). In response to a questionnaire included in the College of American Pathologists HbA1c proficiency survey in April 2012, 1,153 (35.7%) of 3,233 participating clinical laboratories (∼90% of which are located in the U.S.) indicated that they report eAG together with HbA1c results.
Another factor has exerted considerably more influence on how HbA1c is reported. A decision was made in 2007 to report IFCC results in the International System of Units (SI) rather than percent (12,19). Thus, IFCC values are now expressed as millimoles of HbA1c per mole of HbA0. The Committee on Nomenclature, Properties, and Units of the IFCC proposed a new term for HbA1c, namely Haemoglobin beta chain(Blood-N-(1-deoxyfructos-1-yl)haemoglobin beta chain; substance fraction) (19). This term has not gained wide acceptance. However, the SI units have (see “Current reporting” section below). A considerable advantage of reporting IFCC values in SI units (as opposed to reporting IFCC values as percent) is that it will avoid the confusion that would almost certainly have ensued from using the same units (namely percent) but different reference intervals. For example, NGSP-certified HbA1c concentrations of 6.5 and 7.0% (previously 4.8 and 5.3% in the original IFCC reporting system) now correspond to 48 and 53 mmol/mol, respectively (Table 4).
In 2004, a working group with representatives from the ADA, EASD, and IDF was established to harmonize HbA1c reporting (35,36). There was unanimous agreement among the members of the group, termed the ADA/EASD/IDF Working Group of the HbA1c Assay, that the same HbA1c values should be reported throughout the world. The initial report was followed in 2007 by a consensus statement on the worldwide standardization of HbA1c (46). The publication called for HbA1c results to be reported worldwide in IFCC units (mmol/mol) and derived NGSP units (%), using the IFCC-NGSP master equation. A subsequent publication appeared 3 years later (47) and reiterated that HbA1c results should be reported by clinical laboratories worldwide in SI and derived NGSP units. Regrettably, this ideal has not been realized.
In 1998, the European Union introduced a directive on in vitro diagnostics requiring that laboratory tests be traceable to a “higher-order method” (48). This factor, in conjunction with the use of SI units for reporting almost all other laboratory tests in many countries, has resulted in the adoption of SI units for HbA1c by some countries in Europe and the Antipodes (Table 5). Most of the countries that have decided to convert to SI units have had a period of dual reporting where both NGSP and IFCC units have been used before switching to single reporting of SI units. The two countries that had their own national standardization schemes, namely Sweden and Japan, both chose to discontinue reporting results using these numbers. Sweden elected to adopt SI units exclusively, whereas Japan is currently reporting both NGSP and Japan Diabetes Society numbers, before switching to only NGSP numbers in 2013 (Table 5). After considerable deliberation, Canada has recently decided to continue using NGSP values. Although no formal statement has been issued, the U.S., which continues to use “traditional” (non-SI) units for all blood tests, is extremely unlikely to switch to SI units for HbA1c. Most countries have not yet made a decision whether to adopt SI units, but it is likely that at least some will convert. A concern has been expressed by some experts from both less-developed countries and newly industrialized countries that efforts, and limited resources, should be directed toward adopting higher-quality (and standardized) analytical methods, expanding the use of HbA1c in patient care, and educating health care providers. The concern is that introducing SI units will both divert resources from these necessary tasks and generate confusion.
Implications of different HbA1c units
The hope expressed by members of the ADA/EASD/IDF Working Group of the HbA1c Assay that the same HbA1c values be reported globally (36) has clearly not been realized (Table 5). This situation is most unfortunate and has consequences for all those in the field of diabetes. It is essential that countries that choose to report HbA1c in SI units introduce extensive education programs that will adequately explain the new units to all health care providers. Because management of diabetes requires the active participation of the patient (49), the change in reporting must also be clearly communicated to patients. It is important that if HbA1c units are altered, the modification should be coordinated so that it is instituted throughout the entire country to avoid having different units used in a single country. Medical and scientific journals should require that authors provide both SI and DCCT units for all HbA1c results. This dual reporting will enable readers to evaluate results and compare them to prior publications.
Physicians and investigators need to be aware that the conversion of HbA1c results between DCCT/NGSP (%) and IFCC units is not as simple as changing glucose values from traditional to SI units or vice versa. The glucose conversion is based on the molecular mass of glucose (C6H12O6), which is 180.16 g/mol. Therefore, to change glucose from mg/dL to mmol/L, one divides by 18.016 (usually rounded off to 18), and values are multiplied by 18 to switch from mmol/L to mg/dL. For example, 126 mg/dL is equivalent to 7.0 mmol/L, and 40 mg/dL is 2.2 mmol/L. By contrast, the conversion of HbA1c results is more complex, partly because it is the conversion of a ratio of two measurements made in different assays rather than of a single concentration. At an NGSP HbA1c of 4%, the IFCC values are fivefold higher (20 mmol/mol), whereas at 12%, the IFCC results are ninefold greater (108 mmol/mol) (Table 4), precluding the use of a simple multiplication or division factor to transform values. Inspection of Fig. 1 reveals that although the relationship between HbA1c measured in NGSP units and IFCC units is a straight line, that line has a slope that differs significantly from 1 and an intercept that differs from 0. Therefore, conversion between NGSP and IFCC units requires a simple linear equation, the master equation (NGSP = 0.09148(IFCC) + 2.152 or IFCC = 10.93(NGSP) – 23.50). It is reassuring that the master equation has been shown to be stable for >11 years (34), and comparisons between the NGSP and IFCC networks continue to be conducted twice a year. These ongoing assessments, which validate the stability and reliability of the networks, will ensure that results can be converted from DCCT/NGSP units to SI units and vice versa. Tables and/or calculators that convert units should be readily accessible to health care providers and patients. To assist in this endeavor, the NGSP has posted both a table and a calculator on its website (http://www.ngsp.org/convert1.asp). A calculator is also available at http://www.hba1c.nu/eng2.html. For countries that select SI units for reporting HbA1c, it is imperative that the tenet “primum non nocere” (first do no harm) is adhered to and that care of patients with diabetes is not compromised.
This work was supported by the Intramural Research Program of the National Institutes of Health.
No potential conflicts of interest relevant to this article were reported.
The author thanks Drs. Carla Siebelder and Cas Weykamp (Queen Beatrix Hospital, Winterswijk, the Netherlands) for kindly providing the figure. This article is dedicated to the memory of Mervyn C. Berman (1934–2012), scholar, mentor, role model, and friend.