Emily Jane Gallager, Division of Endocrinology, Diabetes and Bone Disease, Mount Sinai School of Medicine, One Gustave L. Levy Place, New York City, New York 10029, USA. Tel: +1 212 241 6500 Fax: +1 212 861 7429 Email: Emily.Gallagher@mssm.edu
Hemoglobin HbA1c (A1c) has been used clinically since the 1980s as a test of glycemic control in individuals with diabetes. The Diabetes Control and Complications Trial (DCCT) demonstrated that tight glycemic control, quantified by lower blood glucose and A1c levels, reduced the risk of the development of complications from diabetes. Subsequently, standardization of A1c measurement was introduced in different countries to ensure accuracy in A1c results. Recently, the International Federation of Clinical Chemists (IFCC) introduced a more precise measurement of A1c, which has gained international acceptance. However, if the IFCC A1c result is expressed as a percentage, it is lower than the current DCCT-aligned A1c result, which may lead to confusion and deterioration in diabetic control. Alternative methods of reporting have been proposed, including A1c-derived average glucose (ADAG), which derives an average glucose from the A1c result. Herein, we review A1c, the components involved in A1c formation, and the interindividual and assay variations that can lead to differences in A1c results, despite comparable glycemic control. We discuss the proposed introduction of ADAG as a surrogate for A1c reporting, review imprecisions that may result, and suggest alternative clinical approaches.
Hemoglobin A1c (A1c) was initially identified as an “unusual hemoglobin in patients with diabetes” by Rahbar et al. in the 1960s.1 Around the same time, there was a strong suspicion that hyperglycemia was related to the vascular complications observed in individuals with diabetes, but the association was difficult to prove due to a lack of objective markers of glucose control.2 As stated by Knowles in a paper reviewing the subject in 1964, “The most doubtful measurement of all is that of control…for it is impossible to determine with certainty the chemical state of patients during their day-to-day life and activity”.3 From the discovery of A1c, multiple small studies were conducted correlating it to blood glucose measurements, with the hypothesis that it could be used as an objective measure of glucose control.2,4 It was introduced into widespread clinical use in the 1980s and subsequently became a cornerstone of clinical practice.5 The Diabetes Control and Complications Trial (DCCT) and UK Prospective Diabetes Study (UKPDS) demonstrated that in Type 1 and Type 2 diabetes, respectively, intensive glucose control, reflected in blood glucose and A1c measurements, decreased the risk of complications.6,7 Furthermore, data on the correlation between blood glucose levels and A1c from these studies were used to derive regression equations for calculating average blood glucose (the mean of the seven daily capillary glucose recordings for the preceding 120 days, as described in the DCCT) from the A1c, aiding in the formation of current diabetes management guidelines.8–11
To improve diabetes control worldwide by introducing a global standardization for A1c measurement, changes in the methods of reporting A1c have been proposed. Given the new methodologies and proposed changes, it appears appropriate at this time to reappraise the use of A1c in the management of diabetes.12–16 As part of the standardization of A1c, the National Glycohemoglobin Standardization Program (NGSP) has proposed that “A1c” be adopted as the preferred term for hemoglobin A1c, and so is used throughout the present review.
Hemoglobin (Hb) is made up of two globin dimers, each with an associated heme moiety. In most adults, of the total Hb, HbA (α2, β2) comprises 97%, A2 (α2, δ2) comprises 1.5–3.5%, and fetal hemoglobin (HbF; α2, γ2) forms <2%. These percentages may change with certain hemoglobinopathies. For example, HbF levels are increased in the presence of hereditary persistence of HbF, β-thalassemia, sickle cell disease, pregnancy, anemia, and certain leukemias. Levels may also be increased in hospitalized patients.17
The components of HbA were identified by charge separation on cation exchange resin and named according to their order of elution as follows: A0, A1a, A1b, and A1c.18 A1c is the Hb component that is composed chiefly, albeit to a variable degree, of glycohemoglobin. In the setting of hyperglycemia, the highly permeable erythrocyte cell membrane allows exposure of Hb to elevated intracellular glucose levels. Glycohemoglobin is formed by the non-enzymatic glycation of the N-terminal valine on the β chain of Hb in a two-step Maillard reaction. First, glucose forms a labile and readily reversible aldamine (Schiff base) with the N-terminal valine on the β chain. The aldamine then undergoes an Amadori rearrangement to form a stable ketoamine. At greater levels of hyperglycemia, glycation of N-terminal lysines and α chains may also occur to a variable extent.19 The degree to which deglycation occurs in vivo remains unclear.20
On average, erythrocytes survive 117 days in men and 106 days in women. At any given time, a given blood sample contains erythrocytes of varying ages, with different degrees of exposure to hyperglycemia. Although older erythrocytes are likely to have more exposure to hyperglycemia, younger erythrocytes are more numerous.19 Blood glucose levels from the preceding 30 days contribute approximately 50% to A1c, whereas those from the period 90–120 days earlier contribute only approximately 10%.21
A1c is affected by a number of genetic, hematologic, and illness-related factors (Fig. 1). Studies demonstrate that individuals without diabetes and with comparable glucose tolerance have a range of A1c levels within the normal parameters, even when factors known to alter A1c (outlined below) are excluded.22 Twin studies suggest that 69% of this interindividual A1c variance can be attributed to genetic factors, whereas the remainder is related to age and the environment.23 Approximately one-third of this inherited variance is related to the “glycation gap”, the difference between the A1c predicted by the glycation of serum proteins and the actual A1c. Variance in the glycation gap is reported to be related, in part, to differences in the erythrocyte transmembrane gradient, suggesting variations in the degree of glucose entry into the erythrocyte, as well as 2,3-diphosphoglycerate and pH levels within the erythrocyte, although other influences have yet to be determined.22,24,25
Changes in erythrocyte lifespan can affect A1c, because increasing the mean age of erythrocytes will increase A1c. Although certain disease states will alter the erythrocyte lifespan, there appears to be significant interindividual variation in mean erythrocyte age in those without known hematological disorders, potentially accounting for some of the variation in A1c in individuals without diabetes.26 An increase in the mean age of erythrocytes will occur in the setting of decreased erythropoiesis, such as in iron and vitamin B12 deficiency, due to a lack of erythropoietin in renal failure, and due to bone marrow suppression in pregnancy and alcoholism.27–32 Conversely, increased numbers of reticulocytes in the circulation decrease the erythrocyte mean age and will decrease A1c. This is seen with hemolytic anemia, after administration of erythropoietin in patients with renal failure, and after repletion of iron and vitamin B12 stores.27,29,30 Increased reticulocytes and a lower A1c are also seen in chronic liver disease, even in the absence of cirrhosis and splenomegaly, but the mechanism responsible is uncertain.33 Increased rates of hemolysis from splenomegaly, rheumatoid arthritis, or drugs such as antiretrovirals, ribavirin, and dapsone can lead to decreased A1c.34–37 Splenectomy increases A1c as a result of increased erythrocyte survival.38
Numerous hemoglobinopathies have been identified that can influence A1c. The most common hemoglobinopathies seen in the US are: (i) HbAS (sickle cell trait) and HbAC, present in populations of African descent; (ii) HbE, common in Asian populations; and (iii) HbD, which occurs in Indian populations. Although homozygous HbSS and HbCC prevent the measurement of A1c due to a lack of β chains, HbAS and HbAC can lead to alterations in A1c measurement due to interference with some assays. HbE and D can also interfere with certain assays. Some hemoglobinopathies, such as Hb Raleigh, can interfere with glycation of Hb.17,39 HbF at levels of <5% generally does not interfere with A1c assays; however, at higher levels, it can cause problems.17,40 Elevated levels of HbF can be induced by the drug hydroxyurea, which is sometimes used to treat sickle cell disease and certain hematological malignancies.41 The presence of methemoglobin can lead to abnormal results with some assays.34
It has been reported that chronic ingestion of aspirin and high doses of vitamin C and E, as well as other antioxidants, can lead to inhibition of hemoglobin glycation, thereby lowering A1c, although the degree to which this occurs clinically is debated.42–45 Recent reports on Hb glycation in chronic renal failure indicate that lipid peroxidation of Hb may increase glycation.46 The relationship between chronic renal failure and A1c is complex: patients with chronic renal failure have decreased levels of erythropoietin, potentially increased glycation, higher levels of carbamylated Hb, and variable exposure to exogenous glucose through dialysate fluid, which could elevate A1c, whereas decreased erythrocyte lifespan would lower A1c.46–48 Alcohol ingestion resulting in the formation of acetaldehyde has been reported to increase A1c measurement by its effect on certain assays.49,50 In addition, very high levels of triglycerides may lead to artefactually low A1c measurements, whereas elevated bilirubin has been reported to artefactually increase A1c results.51,52 Chronic opiate use has been reported to lead to elevated A1c levels, but the underlying mechanism remains unclear.53
The current assays for measuring A1c are based on two principles, namely charge and structural differences between Hb components, and are summarized in Table 1. Glycated Hb acquires an extra negative charge when glucose attaches to the β chain N-terminal valine. The lower isoelectric point is exploited by ion exchange chromatography, because the glycated Hb accelerates faster in a cation-exchange resin. The concentration of Hb is measured using a spectrophotometer and quantified by calculating the area under each peak of the chromatogram compared with a calibrated chromogram.54 Cation exchange performed by high-performance liquid chromatography (HPLC) is currently the most widely used assay method. The BioRad Diamat (HPLC cation exchange using Bio-Rex 70 resin; BioRad, Hercules, CA, USA) was the reference method used for the DCCT; the MonoS assay (Pharmacia Biotechnology, Uppsala, Sweden) and KO500 (Japanese Society for Clinical Chemistry; see below) are also ion-exchange HPLC systems.55 Inaccuracies occur when an Hb variant or derivative cannot be separated from HbA or the A1c fragment. The interferences from HbAS, HbAC, HbE, HbD, HbF, and carbamylated Hb are variable with different ion exchange assays, although inspection of the chromatogram may allow identification of a hemoglobinopathy.17,40,48,56–58 Structural differences are exploited with affinity chromatography and immunoassay.21 The factors that interfere with the most widely used A1c assays can be found at http://www.ngsp.org
Table 1. A1c assays
Ion-exchange chromatography (HPLC)
Glycated Hb has a lower isoelectric point and migrates faster than other Hb components
Variable interference hemoglobinopathies, HbF, and carbamylated Hb
Can inspect chromatograms for Hb variants
Glucose binds to m-aminophenylboronic acid
Measures not only glycation of N-terminal valine on β chain, but also β chains glycated at other sites and glycated α chains
Minimal interference from hemoglobinopathies, HbF, and carbamylated Hb
Antibody binds to glucose and between 4 and 10 N-terminal amino acids on β chain
Affected by hemoglobinopathies with altered amino acids on binding sites; some interference from HbF
Not affected by HbE, HbD, or carbamylated Hb
Standardization of A1c assays
In 1992, a survey by the College of American Pathologists reported large discrepancies in results between A1c assays. For the same blood sample, A1c results ranging from 4% to 8.1% were reported.59 This survey, along with the results of the DCCT, highlighted the need to introduce standardization of A1c measurements in order to improve diabetes management. Some countries subsequently set about establishing national standardization programs.
In the US, the National Glycohemoglobin Standardization Program (NGSP) was established in 1996. The NGSP reference system used the DCCT Laboratory, which measured A1c for the duration of the DCCT and used the BioRad Labs BioRex 70 HPLC System as the “reference standard method”. Manufacturers and laboratories calibrate their assays to this reference method. The NGSP issues certificates to individual laboratories and manufacturers that meet their standardization. The NGSP is used in many centers around the world, with a DCCT-aligned non-diabetic reference range of 4–6%.55,60 In Sweden, a strong methylsulfonate cation exchanger, the MonoS HPLC, was used as the standard reference method to calibrate all hospital and point-of-care assays.60 The Japanese Diabetes Society initially established their standardization program in 1993. They calibrated their assays using the JDS Calibrator Lot 1 (Committee for Standardization of Glycohemoglobin, Japan Diabetes Society, Tokyo, Japan), formed from lyophilized blood samples. In 2000, they updated their program by developing a more accurate set of calibrators (JDS Calibrator Lot 2; Health Care Technology Foundation, Kawasaki, Japan) and introducing a more specific high-resolution cation-exchange HPLC analyzer, the KO500.61
Thus, the existing situation is unsatisfactory, with different assays used in different areas, giving different results, based on different calibrators, derived from different sets of human blood specimens. The International Federation of Clinical Chemists (IFCC) established a working group on A1c in an attempt to introduce an international standardization program. The purpose was to have one worldwide reference system, a single international reference method, and purified A1c calibrators. A definition of A1c was generated:63 Hb molecules with a stable adduct of glucose to the N-terminal valine of the Hb β chain (β-N(1-deoxy)fructosyl-Hb). Two reference methods were then formed with the combination of HPLC and electron spray mass spectrometry (MS) or, alternatively, a two-dimensional approach using HPLC and capillary electrophoresis (CE) with ultraviolet detection.63 Primary reference materials were also created, namely purified A0 and A1c, to calibrate these reference procedures. The IFCC references were accepted by the National Societies of Clinical Chemistry in 2001 and a network of laboratories was established using either the HPLC-MS or the HPLC-CE method. These methods show no interference from HbAS, HbAC, HbF, or carbamylated Hb. Owing to the expense of the IFCC assay methods, clinical laboratories continue to use their own assays, but they are now calibrated to these more accurate references. The target interlaboratory coefficient of variance is <2.5%. Studies comparing the NGSP, JDS, and Swedish A1c values have allowed the calculation of conversion equations for the various methods,60–62 as follows:
Follow-up studies of the IFCC standardization verify its performance on accuracy of results.64 The IFCC A1c is expressed in mmol/mol Hb. When given as a percentage, because it does not measure non-specific components of A1c, the IFCC reference range for non-diabetic patients is approximately 1.5–2% lower than DCCT values.63,64
Current controversies in A1c
Despite the greater accuracy and consistency of results with the IFCC standardization, concerns were raised that if the IFCC A1c was expressed as a percentage, the similar but lower A1c results compared with the current NGSP, Japanese, and Swedish ranges would result in deterioration of glycemic control. This phenomenon was reported in a Swedish study65 of changes in population A1c levels after the reference method was revised in 1992 and again in 1997 (Fig. 2). In 1992, the new reference range for A1c was approximately 1.4% higher than the previous range, leading to higher reported values. Patients’ A1c results were seen to initially increase as a result of the new scale, but were noted, after a period of time, to drift back down toward the old familiar reference range, reflecting improved glycemic control. However, in 1997, when the reference range was decreased by approximately 1.1%, the converse occurred; that is, A1c drifted up, reflecting a deterioration in glycemic control. This occurred despite concerted efforts to educate patients and practitioners on the new reference ranges and is hypothesized to have occurred owing to the psychological impact of the absolute numbers.65 Because the IFCC A1c reference range is smaller compared with the other reference systems when expressed as a percentage, this would mean that a smaller change in A1c would reflect a more significant change in average glucose levels, which may go unnoticed by patients and practitioners.
Several suggestions have been made for new approaches to A1c reporting based on the IFCC assay. One is to convert the IFCC units back to DCCT units for consistency and ease of understanding. The second is to express the A1c as mmol/mol Hb, with a non-diabetic range of 25–42. A third suggested approach has been to abandon the term A1c and to instead report an estimated average glucose or A1c-derived average glucose (ADAG).12
The proposed use of ADAG rather than A1c has led to renewed interest in investigating the relationship between blood glucose and A1c, and a large multinational study undertook to define this relationship.15 The regression equation currently used by the American Diabetes Association (ADA) was developed from the DCCT data. The equation
has an r2 of 0.82, with a rather marked degree of variability: an individual with a average plasma glucose of 180 mg/dL (10 mmol/L) could have an A1c of 6–11%.8 In children, this difference can be even greater, with data from the Diabetes Research in Children Network (DirecNet) showing that an A1c of 7% can reflect an average glucose of 138–189 mg/dL (7.5–9 mmol/L).66 Differences were noted in the DCCT between individuals in the intensively treated group compared with the conventional treatment group: an average glucose difference of 29 mg/dL (1.6 mmol/L) in those with an A1c of 7% compared with 52 mg/dL (2.8 mmol/L) in individuals with an A1c of 11%.67 The DCCT data are mostly from white individuals with Type 1 diabetes and although there is no clear reason why individuals with Type 2 diabetes would have a different A1c level for the same average blood glucose, the ADAG Study Group set out to investigate this possibility. They also aimed to gather data on different ethnic groups. The ADAG Study Group has reported data on 506 individuals aged between 18 and 70 years, incorporating 268 patients with Type 1 diabetes, 159 with Type 2 diabetes, and 80 without diabetes.15 Individuals had to have a stable A1c (<1% change) for the 6 months prior to recruitment. Of the population studied, 74% was white, 15% was African or African American, and 8% was Hispanic. Unfortunately, due to the withdrawal of one large Asian center, this population was under-represented. Individuals with hemoglobinopathies, anemia (hematocrit <39% in men and <36% in women), evidence of increased erythrocyte turnover, blood loss, transfusion, chronic renal or liver disease, those treated with high-dose vitamin C or epoetin, those on or with a condition possibly requiring steroid therapy, and those likely to become pregnant during the study period were excluded, in all excluding 15% of the potential population. Blood glucose was measured by a continuous glucose monitor (CGM) for 3 days out of every 4 weeks, along with capillary glucose monitoring (with the Hemocue Glucose 201 Plus; Hemocue, Angelholm, Sweden) eight times daily while using the CGM. For 3 days each week, while not using the CGM, capillary glucose monitoring was performed seven times daily with a different glucose monitor. A1c measurements were performed at baseline and then monthly for 3 months at a central laboratory using four different NGSP-approved assays, and the average was taken as the A1c measure. The investigators reported a linear relationship between A1c and average plasma glucose:
with 89.95% of samples falling within 15% of the calculated average glucose.15
The ADAG Study suggested that African and/or African Americans may have a tendency towards higher A1c values for a given average plasma glucose, but the numbers were too small to draw definite conclusions.15 In a US study of individuals with impaired glucose tolerance, the Diabetes Prevention Program (DPP) showed that A1c varied with race, despite comparable glucose levels: whites had the lowest A1c (5.78%), followed by Hispanics, Asians, and American Indians, with African Americans having the highest A1c (6.18%).68 Because individuals with hemoglobinopathies were excluded from the ADAG Study and because 7.8% of African Americans in the US and up to one-third of patients in sub-Saharan Africa exhibit HbAS, whereas 2.3% of African Americans and approximately 30% of the population of sub-Saharan Africa have HbAC, the applicability of the linear regression model from the ADAG trial in this group is questionable.17
Although some of the variability in the DCCT, DirecNet, ADAG, and DPP studies may be attributable to different modes and the accuracy of A1c and glucose measurement, the variability seen in studies may be related to interindividual variability in Hb glycation. Although it has been shown that, over time, a given individual’s average glucose correlates strongly with A1c, there are considerable differences between individuals. There may be genetically determined differences in Hb glycation rates that account for these differences in A1c.22–25,69 These considerations suggest that a nominal ‘average’ blood glucose derived from the A1c may produce inaccurate results, potentially confusing both patients and physicians who may well observe discrepancies between capillary glucose levels based on self-monitoring and the supposed average glucose.
Evidence continues to grow demonstrating the importance of tight glycemic control in diabetes to prevent the development of complications.6,7,70,71 Despite the advances in diabetes care since the 1960s, uncertainty still surrounds the issue of monitoring glycemic control. The use of patients’ recorded capillary glucose measurements to accurately reflect home glucose control has been considered unreliable.72,73 Therefore, A1c continues to be fundamental for the clinical management of diabetes and is likely to remain so for some time. The precision of the IFCC standardization is obviously advantageous, but has caused the current controversy. Clearly, expressing the IFCC A1c as a percentage has the potential to cause the most dangerous mistakes from misapprehension and miscalculation. Given the meticulous process undertaken to attain the IFCC A1c, it appears counterintuitive to convert such an accurate measurement to a value (ADAG) that is inherently inaccurate. Even in the ADAG study population, with stable glucose control and excluding any factors that are known to affect A1c, approximately 10% of values lay outside 15% of the predicted ADAG, leaving a degree of variability that makes it difficult to extrapolate this model to the general population.15
With the introduction of the IFCC A1c, we believe it is best to report it as mmol/mol, accompanied for a transition period by the DCCT (or national) A1c percentage to allow patients and medical staff a time to adjust to and understand the significance of the new units. Intensive education of physicians and patients with regard to the new units of measurement and their clinical significance would certainly be required, but a change to ADAG would also require re-education. Although the initial psychological impact of changing the units would be great, conversion cards could be made available for the easy conversion between the DCCT and the IFCC A1c akin to those provided when European countries changed their currencies to the Euro.
With the accurate IFCC units, studies could continue with the aim of gaining a greater understanding of the relationship between A1c and glycemia. Questions remain about this relationship in children and the older population, as well as in different ethnic groups. A clear understanding of genetic influences on A1c variance remains elusive. The effects of glucose fluctuations compared with those of chronic hyperglycemia on glucose transport into erythrocytes and the rates of glycation and deglycation remain to be determined. The significance of glycating sites other than the N-terminal valine of the β chain is uncertain. The exact mechanisms through which pregnancy, chronic liver disease, renal disease, and HIV affect A1c are incompletely understood. Therefore, further investigation and a new precise A1c may resolve the uncertainty of glucose control.
Supported, in part, by an unrestricted education grant from Daiichi Sanyko (Parsipanny, NJ, USA) to Mount Sinai Hospital, Division of Endocrinology, Diabetes and Bone Disease.