Genuth, Lachin, and Nathan (1) pose several questions regarding the methodology and conclusions of our recent report (2). We are responding in order to clarify the validity of our approach, results, and conclusions regarding the important relationship between biological variation (BV) in HbA1c and mean blood glucose (MBG) and the development of microvascular complications of diabetes.

A key consideration in our hypothesis is the existence of consistent between-patient directional differences over time in HbA1c not accounted for by the associated MBG levels (2). BV in biochemical metabolites is a well-recognized phenomenon in clinical chemistry (3). BV in HbA1c levels has been reported in individuals without diabetes (4), where MBG has little impact on HbA1c levels. In patients with diabetes, the strong correlation between HbA1c and MBG must be taken into account when assessing the impact of BV on HbA1c levels. The hemoglobin glycation index (HGI) was devised for this purpose (2,5). We have used the HGI to demonstrate the presence of directional BV in HbA1c in longitudinal data from patients with diabetes in our own clinic population (5) as well as participants in the Diabetes Control and Complications Trial (DCCT) (2). In repeated measures obtained over the course of both studies, we found that large numbers of patients had consistently nonrandom, large positive or negative deviations (high or low HGIs, respectively) of their observed HbA1c from the predicted HbA1c based on their associated MBG (2,5). Thus, HbA1c levels from patients with diabetes can be shown to be determined by idiosyncratic patient-specific factors as well as by MBG.

Genuth, Lachin, and Nathan (1) do not directly question the existence of between-patient BV in HbA1c, but they do offer an alternative explanation for the findings. They suggest that DCCT participants who usually have poor glycemic control and high HbA1c levels may have been motivated to deliberately improve their glycemic control just on the day they collected their glucose profile set. This would cause lower glucose levels just for that day and result in a misleadingly low MBG. These patients would then appear to have a much higher HbA1c level than would be predicted for their MBG and thus exhibit a large positive HGI. However, it should be kept in mind that there were as many DCCT participants with low HGI as there were with high HGI. By the reasoning of Genuth, Lachin, and Nathan, a similarly large number of patients who usually have good glycemic control with low HbA1c levels deliberately worsened their glycemic control just for the day of the profile set. Thus, leading to HbA1c levels unexpectedly lower than would be predicted from their MBGs and giving a large negative HGI. We believe that this explanation is not plausible. It is unlikely that two-thirds of the DCCT participants consistently and deliberately either improved or worsened their glycemic control just for 1 day each quarter for 7–9 years. Nor is it a plausible explanation for BV in HbA1c levels reported in clinic patients where MBG was estimated from multiple glucose samples obtained over a period of 30 days (5) or in individuals without diabetes (4). We have considered other possible explanations for BV in HbA1c within the discussion of our report (2). Ultimately, the possibility that BV in HbA1c is a manifestation of individual differences in metabolic processes, other than MBG, that influence hemoglobin glycation and development of complications is a hypothesis that deserves further attention (2,5).

Genuth, Lachin, and Nathan point out that HbA1c is a good index of glycemia over a period of 4–12 weeks, while the DCCT MBG is derived from a seven-sample, 1-day glucose profile set and may not provide as good an estimate of the true MBG over the same period of time. To obtain a better estimate of HGI, Genuth, Lachin, and Nathan recommend using an average of MBGs and HbA1c levels over the entire DCCT follow- up. For the purpose of predicting complications, the HGI was indeed computed as an average value for each subject over the course of the DCCT follow-up (2), as Genuth, Lachin, and Nathan recommended. When computed in this way, the HGI represents a measure of each individual’s average directional divergence in HbA1c from that predicted over dozens of quarterly measurements; it is not merely a single regression residual.

Furthermore, we note that the association of HbA1c levels with glucose levels is so strong that consistent regression relationships have been demonstrated between HbA1c and an MBG derived from multiple glucose samples obtained over 30 days (5), seven glucose samples obtained over the course of 1 day (6), or even one glucose sample (7). It stands to reason that an MBG calculated from a 1-day, 7-point glucose profile set may not estimate the true MBG as well as an MBG calculated from a more extensive sampling protocol. However, the 1-day DCCT MBG has been shown by DCCT investigators to provide a valid regression relationship with HbA1c (6), and this regression model has been extensively promoted as a way to estimate MBG based on HbA1c assays standardized to the DCCT high-performance liquid chromatography methodology (8). There is no evidence that the 1-day MBG estimates themselves are biased. Furthermore, MBG calculated from the DCCT 1-day profile sets have themselves been successfully used to predict the development of diabetic retinopathy (9). Collectively, this information suggests that the MBG estimates from the DCCT are quite adequate for analysis of BV in HbA1c and risk of complications.

Our main hypothesis tested in the report was that BV in HbA1c, after accounting for the influence of MBG, is predictive of diabetic retinopathy and nephropathy (2). The simple approach of including both HbA1c and MBG together in a statistical model for prediction of complications is undesirable, since HbA1c is highly correlated with MBG. Genuth, Lachin, and Nathan point out that HGI and HbA1c are also highly correlated. However, HGI is not correlated with MBG (5) and can therefore be reliably included as an independent variable in a statistical model together with MBG for the prediction of nephropathy and retinopathy. This statistical approach allowed us to successfully demonstrate that BV in HbA1c, as measured by HGI, is predictive of retinopathy and nephropathy, independent of the important effect of MBG.

The findings of our report should in no fashion be construed as detracting or contradicting the important results and conclusions of the DCCT. Nor should they be taken to suggest that HGI should replace HbA1c or deter the clinical use of HbA1c for the assessment of patients with diabetes. If anything, our findings suggest that in addition to reflecting the effects of preceding MBG, HbA1c also contains important information about other processes that influence hemoglobin glycation and the development of complications. We hope that this discussion will promote further study of these processes and their impact on diabetes complications.

1.
Genuth S, Lachin JM, Nathan DM: Biological variation in HbA1c predicts risk of retinopathy and nephropathy in type 1 diabetes (Letter).
Diabetes Care
28
:
233
,
2005
2.
McCarter RJ, Hempe JM, Gomez R, Chalew SA: Biological variation in HbA1c predicts risk of retinopathy and nephropathy in type 1 diabetes.
Diabetes Care
27
:
1259
–1264,
2004
3.
Fraser CG:
Biological Variation: From Principles to Practice
. Washington, D.C., American Association of Clinical Chemistry Press, 2001
4.
Kilpatrick ES, Maylor PW, Keevil BG: Biological variation of glycated hemoglobin: implications for diabetes screening and monitoring.
Diabetes Care
21
:
261
–264,
1998
5.
Hempe JM, Gomez R, McCarter RJ Jr, Chalew SA: High and low hemoglobin glycation phenotypes in type 1 diabetes: a challenge for interpretation of glycemic control.
J Diabetes Complications
16
:
313
–320,
2002
6.
The DCCT Research Group: Diabetes Control and Complications Trial (DCCT): results of a feasibility study.
Diabetes Care
10
:
1
–19,
1987
7.
Rohlfing CL, Wiedmeyer H, Little RR, England JD, Tennill A, Goldstein DE: Defining the relationship between plasma glucose and HbA1c.
Diabetes Care
25
:
275
–278,
2002
8.
National Glycohemoglobin Standardization Program
. Available at www.missouri.edu/~diabetes/ngsp/. Accessed 17 November 2004
9.
Service FJ, O’Brien PC: The relation of glycaemia to the risk of development and progression of retinopathy in the Diabetic Control and Complications Trial.
Diabetologia
44
:
1215
–1220,
2001