Glycated haemoglobin values: problems in assessing blood glucose control in diabetes mellitus

BMJ 1994; 309 doi: (Published 15 October 1994) Cite this as: BMJ 1994;309:983
  1. E S Kilpatrick,
  2. A G Rumley,
  3. M H Dominiczak,
  4. M Small
  1. Department of Pathological Biochemistry, Gartnavel General Hospital, Glasgow G12 OYN Diabetic Unit, Gartnavel General Hospital, Glasgow G12 OYN.
  1. Correspondence to: Dr Kilpatrick, lipid and diabetes laboratory
  • Accepted 9 August 1994


Objective: To see whether two measures of glycated haemoglobin concentration - the haemoglobin A1 (HbA1) value and the haemoglobin A1c (HbA1c) value - assess blood glucose control differently in diabetes.

Design: Diabetic patients had glycaemic control assessed on the basis of HbA1 and HbA1c values measured by the same high performance liquid chromatography instrument and on the basis of HbA1 measured by electrophoresis.

Setting: A diabetic outpatient clinic.

Subjects: 208 diabetic patients and 106 non-diabetic controls.

Main outcome measures: Glycated haemoglobin concentrations classified according to European guidelines as representing good, borderline, or poor glycaemic control by using standard deviations from a reference mean.

Results: Fewer patients were in good control (25;12%) and more poorly controlled (157;75%) as assessed by the HbA1c value compared with both HbA1 assays (39 (19%) and 130 (63%) respectively when using high performance liquid chromatography; 63 (30%) and 74 (36%) when using electrophoresis). The median patient value was 8.0 SD from the reference mean when using HbA1c, 5.9 when using HbA1 measured by the same high performance liquid chromatography method, and 4.1 when using HbA1 measured by electrophoresis.

Conclusions: Large differences exist between HbA1 and HbA1c in the classification of glycaemic control in diabetic patients. The HbA1c value may suggest a patient is at a high risk of long term diabetic complications when the HbA1 value may not. Better standardisation of glycated haemoglobin measurements is advisable.

Clinical implications

  • Clinical implications

  • Improved glucose control in diabetic patients reduces the risk of long term microvascular complications

  • HbA1 and HbA1c are glycated haemoglobins commonly measured to give an indication of glycaemic control over the preceding six to eight weeks

  • In this series HbA1 measurement classified fewer patients as poorly controlled and more as well controlled in comparison with HbA1c

  • Patients may thus appear to be at less risk of long term complications when HbA1 concentration rather than the more specific HbA1c concentration is measured

  • until standardisation to HbA1c measurement occurs doctors should be aware that care is required with the interpretation of glycated haemoglobin measurements


Over the past decade measurement of glycated haemoglobin concentration has brought a major advance in the assessment of glycaemic control in diabetes mellitus by providing an objective indication of a patient's overall blood glucose control for the preceding six to eight weeks.1 The term glycated haemoglobin encompasses both haemoglobin A1 (HbA1) and haemoglobin A1c (HbA1c). HbA1 refers to the non-enzymatic binding of several species of carbohydrate to haemoglobin, whereas in HbA1c the carbohydrate is specifically glucose.2

The desirability of good glycaemic control in insulin dependent diabetes mellitus has been reinforced by the diabetes control and complications trial, which showed an impressive reduction in microvascular complications in intensively treated patients when compared with a group treated conventionally.3 Though self blood glucose monitoring is an important safety check for patients, technical problems and lack of appeal often make it unreliable as an indicator of glycaemic control.4,5 Therefore, in routine clinical practice more emphasis is placed on glycated haemoglobin measurement. Hence it is important that the classification of a patient's glycaemic control by using this measurement should be both accurate and reproducible. However, because of differing methods of analysis the stated reference ranges for glycated haemoglobin measurements may vary substantially. This means it is difficult to find a common set of target values for glycated haemoglobin which will be applicable to all analyses. In an attempt to account for this, recent guidelines have been set for patients with insulin dependent and non-insulin dependent disease which define categories of glycaemic control as a HbA1 or HbA1c concentration so many standard deviations from a particular method's non-diabetic population mean.6

When evaluating the original recommendations for non-insulin dependent diabetes published in 19887 we found that measurement of HbA1c by agglutination inhibition placed significantly more patients in the poorly controlled group than HbA1 measured by electrophoresis.8 The aim of this study was to ascertain whether a discrepancy remained when both HbA1 and HbA1c were measured simultaneously on one instrument by the same method (high performance liquid chromatography). If such a disparity existed, then the interchangeable use of HbA1 and HbA1c for the measurement of glycaemic control would require reappraisal.

Subjects and methods

Two methods of glycated haemoglobin analysis were used. HbA1 and HbA1c were measured individually by high performance liquid chromatography (HiAutoAlc, Model 8121, Kyoto Daiichi Kagakiu, Japan); HbA1 was additionally measured by an electrophoretic method (Ciba Corning Diagnostics, Halstead, Essex). Between batch imprecision (coefficient of variation) was less than 4.5% for each analysis at a mean HbA1c concentration of 8.2%.

A locally derived reference range (mean with 2 SD) for the high performance liquid chromatography and electrophoretic methods was compiled by studying 106 non-diabetic subjects (42 male, 64 female; median age 36 (range 16-82) years), comprising hospital staff and families.

During the same period 208 samples from consecutive patients (114 male, 94 female; 90 insulin treated, 118 non-insulin treated; median age 60 (range 13-94) years) attending the diabetic outpatient clinic were analysed by both high performance liquid chromatography and electrophoresis. All samples were analysed within three days of collection.

Diabetic patient samples were categorised according to European guidelines for insulin dependent diabetes mellitus. These define good glycaemic control as a HbA1 or HbA1c value less than 3 SD from a method's non-diabetic population mean. Borderline control is between 3 and 5 SD and poor control is above these limits.6

Statistical analysis was by the McNemar test for paired samples and the X2 test for unpaired proportions. The Gaussian distribution of the reference samples was verified by Kolmogorov-Smirnov one way analysis. STATGRAPHICS software (Statistical Graphics System, Rockville, Maryland; Statistical Graphics Corporation, 1986) was used throughout.


Table I shows the results of glycated haemoglobin measurements obtained from the reference population for each analysis with their respective good, borderline, and poor control limits. The spread (SD) of each assay's reference values is also expressed as a percentage of the method mean (sample coefficient of variation).


Reference population statistics and derived glycaemic control categories (n=106)

View this table:

Figure 1 shows good table 5 correlation between the two HbA1 assays and HbA1c. HbA1 measured by electrophoresis also correlated with HbA1 measured by high performance liquid chromatography (y=0.90x+1.61; r=0.888).

Fig 1
Fig 1

Left: Relation between HbA1 and HbA1c concentrations (as measured by high performance liquid chromatography) in diabetic patients (y=1.20x + 1.01; r=0.982). Right: Relation between HbA1 concentrations (as measured by electrophoresis) and HbA1c concentrations (as measured by high performance liquid chromatography) (y=1.10x+2.37; r=0.891). Dashed lines represent 3 and 5 SD limits

Table II, however, shows significantly fewer patients classified in good control and more as poorly controlled with HbA1c (high performance liquid chromatography) compared with both HbA1 assays. This was true for both insulin treated and non-insulin treated patients (no significant difference). The clinic patient median HbA1c (high performance liquid chromatography) value was proportionally higher than HbA1 (both by high performance liquid chromatography and electrophoresis) when compared with values in the reference population. Constituents of HbA1 other than HbA1c - that is, HbA1a1, HbA1a2, and HbA1b - were estimated by subtracting the HbA1c value from HbA1. The median diabetic patient value for this was 2.30%, which represented a 24% increase above the non-diabetic mean of 1.86%.


Diabetic patient statistics and glycaemic control according to European guidelines for patients with insulin dependent diabetes mellitus (n=208)

View this table:

Figure 2 shows why more patient were classified as poorly controlled when HbA1c was measured. The distribution of diabetic samples in which HbA1c (high performance liquid chromatography), HbA1 (high performance liquid chromatography), and HbA1 (electrophoresis) values were measured is shown as a function of the SD from their respective method means.

Fig 2
Fig 2

Distribution of diabetic patient samples as function of SDs from respective method means.


The results of the diabetes control and complications trial have provided the best objective guide for desirable glycaemic control limits to prevent microvascular complications in insulin dependent diabetes mellitus. In that study the median HbA1c concentration in intensively treated patients was 4 SD from the mean non-diabetic value whereas the median in the conventionally treated group was 8 SD from the mean.3 In our study the median diabetic value was 4.1 SD from the non-diabetic mean when using electrophoretically measured HbA1, 8.0 when using HbA1c (measured by high performance liquid chromatography), and 5.9 when using HbA1 measured by high performance liquid chromatography. Thus, depending on the method of measuring glycated haemoglobin, our diabetic clinic patients could be described as equivalent to either the intensively treated group, the conventionally treated group, or about midway between. Though our study included non-insulin dependent patients, it has been suggested that the results are likely to be equally applicable to this group.9

In the United Kingdom the variety of assays for glycated haemoglobin is shown by the submission of samples to the national external quality assessment scheme from participating laboratories. In October 1993 the scheme reported four different instruments for HbA1 measurement (Corning electrophoresis being the commonest) together with eight instruments for HbA1c analysis (high performance liquid chromatography being the commonest). This study has clearly shown that there is considerable discrepancy in the classification of glycaemic control when comparing the electrophoretic HbA1 method with HbA1c high performance liquid chromatography. As assessed with European guidelines, 74 (36%) of our patients were poorly controlled (>5 SD) when electrophoresis was used as compared with 157 (75%) when HbA1c was measured by high performance liquid chromatography. This is consistent with our previous findings when comparing electrophoretically measured HbA1 values with HbA1c measured by an agglutination inhibition method.8

The discrepancy is not confined only to the electrophoretic HbA1 assay. This study has also shown that a substantial disparity between HbA1 and HbA1c categorisation remained even when patient specimens were measured by using the same high performance liquid chromatography instrument, time of analysis, and reference range samples. Significantly more patients had poor control as assessed by HbA1c values than by HbA1 (75% v 63%). The reasons for this appear twofold. Firstly, the spread (SD) of the non-diabetic HbA1 reference population results was relatively greater than that of HbA1c (7.8% v 7.1% of the mean reference value). Thus more diabetic samples fell within 3 and 5 SD when HbA1 was measured rather than HbA1c. Secondly, in comparison with non-diabetic values, patient HbA1c values were proportionally higher than HbA1 (median value 57% v 46% greater than the reference mean). The implication is that this was due to the concentration of glycated analytes of HbA1 other than HbA1c (HbA1a1), HbA1a2, and HbA1b rising less rapidly than the concentration of HbA1c itself.

There remained a significant difference in the classification of glucose control between the two HbA1 methods. A total of 130 (63%) patients were poorly controlled when high performance liquid chromatography was used and 74 (36%) when electrophoresis was used. This was due to the electrophoretic assay exhibiting a comparatively higher reference range SD (11.9% of the mean v 7.8%). This disparity was likely to be due in part to the fact that, unlike the chosen high performance liquid chromatography method, both glycated and non- glycated fetal haemoglobin comigrates with HbA1 in the electrophoretic and some other assays and so is included in the HbA1 result.10 Nearly half of patients with insulin dependent diabetes mellitus have fetal haemoglobin concentrations exceeding 0.5%.11 Not only may this lead to spuriously raised HbA1 values in some patients but it may also lead to an increase in the imprecision (and therefore reference range) of the assay as a whole.11

In conclusion, we have found substantial differences in the classification of glycaemic control in diabetic patients when using HbA1 measurement rather than HbA1c. In relation to the diabetes control and complications trial this inconsistency may have considerable consequences for the long term wellbeing of diabetic patients and may also influence the allocation of resources towards their treatment. Therefore, this study reinforces the need for more standardisation in the methods used for measuring glycated haemoglobin values. Adopting a standardised HbA1c would allow the development of clear guidelines for clinicians based on both recent and subsequent complications trials.

We express our gratitude to Professor I W Percy-Robb for his support and to Biomen (UK) Ltd for the loan of equipment.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
View Abstract