Biological Variation in HbA_1c Predicts Risk of Retinopathy and Nephropathy in Type 1 Diabetes: Response to McCarter et al.

We read the article by McCarter et al. (1) with interest. Technically, all nonanalytical variation, irrespective of its source, is biological variation. Thus, mean blood glucose (MBG)-associated changes are included in biological variation. It must also be stressed that all population regression equations have confidence limits that need to be taken into account when comparing values from individuals to the population study mean.

However, such points do not take away from the message of McCarter et al. (1) that non-MBG–related biological variation may be an important prognostic indicator. The real question is how health care professionals are to identify this variation in routine clinical practice. MBG has many problems such as a large variation, which is common when many independent analytes are measured (2), bias due to calibration issues (3), or the time taken for separation (4). Most importantly, it is rarely used in routine clinical practice. In addition, HbA_1c also has its problems (5). Accordingly, the calculation of the hemoglobin glycation index is problematic in routine clinical practice. Furthermore, calculated indexes will suffer from the propagation of error, contributing to misclassification and inaccurate prediction of complications (6).

We previously recommended the use of a rolling mean to reduce the effect of analytical and biological variation (7). The associated SD in stable patients would reflect the total variation for HbA_1c. Since the majority of the total variation is nonanalytical, use of the SD would easily identify those patients in routine clinical practice with large non-MBG–related biological variation. As well as being easier to perform in routine clinical practice, it would also be a more valid way of identifying within-patient HbA_1c variability. In addition, the use of a rolling mean and its associated SD makes the detection of critical changes in HbA_1c levels easier and more objective (5). Accordingly, we recommend the use of a rolling mean and its associated SD for the investigation of non-MBG–related biological variation.