Chalew et al. (1) have commented on our reassessment (2) of their hemoglobin glycation index (HGI), in which we show that the HGI is largely a function of A1C and that it does not—and in fact cannot—contribute independently to the risk of complications. Although we readily acknowledge that A1C only explains a modest fraction of the risk of diabetes complications, we continue to question whether the HGI alone, or in combination with mean blood glucose (MBG), represents a biological process that merits consideration.
The original premise underlying the derivation of the HGI is the putative existence of “substantial differences in the level of A1C between patients at any given MBG” that are attributed to differences in the propensity to glycation as measured by the HGI (1). Apart from the flawed analyses previously conducted, there is in our opinion little biological evidence to support this proposition.
Chalew et al. (1) also propose that in conjunction with the MBG, the HGI provides an alternative mechanism to describe the effects of hyperglycemia on the risks of complications. To support this view, they cite the imperfect correlation of MBG derived from the Diabetes Control and Complications Trial (DCCT) seven-point profile and from a second study with an average of 89 measurements per subject during a 30-day period, with the corresponding A1C levels (R2 = 0.67 and 0.49, respectively) (3,4). They choose to attribute the remaining 33–51% of unexplained variation in A1C to differences among subjects in HGI levels (3,4). However, an equally plausible explanation for the imperfect correlation between MBG and A1C levels is the imprecision or inaccuracy of the measurement of the MBG over the ∼120 days that a single A1C value reflects.
The DCCT capillary glucose profile was not designed to capture MBG over this 120-day interval; it measured glucose levels only at seven time-points on a single day and only once every 3 months. The relatively large interindividual variation of these seven-point profiles (30%), compared with the variation in A1C (∼10%), suggests a large degree of variability in the mean glucose profiles so measured on a single day quarterly (5) and the potential for sampling errors. Of note, when more frequent glucose measures have been used to capture mean blood glucose levels more completely, the correlations between calculated MBG and A1C have been substantially stronger, ranging from an R2 of 0.84 to 0.9 (6–8).
Thus, the variability in the relationship between MBG and A1C that lead Chalew et al. (1) to develop the HGI concept decreases substantially when the measure of MBG is more complete and reliable. Whether the ∼15% remaining variance in the relationship between MBG and A1C is attributable to biological factors such as HGI or, more likely, is related to the imprecision of calculated MBG needs to be determined.
Even if biological factors that lead to the purported variation measured by the HGI do exist, such factors have not been shown to contribute to the risk of complications. Owing to the weakness of the DCCT measure of MBG over time, these questions cannot be definitively addressed by the DCCT.
