Intended for healthcare professionals


Authors' reply

BMJ 1995; 310 doi: (Published 18 March 1995) Cite this as: BMJ 1995;310:741
  1. Eric S Kilpatrick,
  2. Alan G Rumley,
  3. Marek H Dominiczak,
  4. Michael Small
  1. Career registrar Principal biochemist Consultant biochemist Consultant physician Department of Pathological Biochemistry, Gartnavel General Hospital, Glasgow G12 0YN

    EDITOR,—Susan Standing and Richard Taylor suggest that use of a fixed biological variation to express SD may help in the comparison of methods of measuring haemoglobin A1 (HbA1) that have different imprecisions. This may be simplified further as the imprecision of the assay affects only the spread of results: it does not affect either the mean glycated haemoglobin concentration in a reference population or the median value in a diabetic population. Thus it would seem more useful to avoid problems with the imprecision of assays by eschewing the use of the SD in favour of comparison of a diabetic patient's result with the mean non-diabetic value.

    In a situation analogous to that used in screening for Down's syndrome and for neural tube defects, a diabetic patient's result could be expressed as a multiple of the mean value (MoM) in non-diabetic people. If this is applied to our study the median diabetic HbA1 value measured by high performance liquid chromatography was 1.46 MoM—that is, 46% higher than the non-diabetic mean, which is similar to the value of 1.48 MoM found when electrophoresis was used. As with Standing and Taylor's suggestion, however, this method of comparison still leads to discrepancies when HbA1 is compared with haemoglobin A1c (HbA1c): even when the same patients and high performance liquid chromatography instrument were used to measure both HbA1 and HbA1c the median HbA1c value implied poorer glycaemic control at 1.57 MoM. As a guide, the group who were intensively treated in the diabetes control and complications trial had a median HbA1c value of approximately 1.40 MoM while the value in the conventionally treated group was 1.80 MoM.1

    Thus, while comparisons of two methods of measuring either HbA1 or HbA1c that use the MoM may be valid, comparisons of HbA1 with HbA1c remain problematic. S Bulusu reinforces the need, recognised by the British Diabetic Association, for more standardisation in measurements of glycated haemoglobin. The errors in the accuracy and imprecision of assays introduced by both haemoglobin variants and abnormal fetal haemoglobin concentrations are well known,2 3 but most methods of analysis remain affected.


    1. 1.
    2. 2.
    3. 3.
    View Abstract