Recently, Lachin et al. (1) published a reassessment of the role of the hemoglobin glycation index (HGI) for prediction of microvascular complications in the Diabetes Control and Complications Trial (DCCT). They conclude that HGI is not a useful predictor because it is not statistically independent of A1C. Their analysis was based on work from a prior publication of ours, in which we came to the conclusion that between-patient biological variation in A1C, as quantified by HGI, along with mean blood glucose (MBG) were both important predictors of complications (2). We believe that the difference in conclusions is due to a difference in the research question being addressed by our two respective groups.
Our group has previously shown that there are substantial differences in the level of A1C between patients at any given MBG that are nonrandom and consistent over time (2–3). The impact of these between-patient differences on the variance of A1C is fairly large, second only to the effect of MBG (4). Thus, we have provided evidence for two main components that influence the measured level of A1C: 1) MBG and 2) idiosyncratic differences between patients in the level of A1C at the same MBG (2–4). To facilitate analysis, we developed the HGI, which quantifies the consistent between-patient variation in A1C after the MBG effect on A1C is removed (2–4). HGI is statistically independent of MBG. As both HGI and MBG contribute to the level of A1C, they are highly correlated with A1C.
We used DCCT data to test whether risk for diabetes complications is influenced by the MBG component of A1C alone or whether factors underlying between-patient variation in A1C, as measured by HGI, can also contribute to risk. A1C was not included as a covariate in our statistical analysis, as we were specifically seeking to test the influence of the two important A1C subcomponents. We found that both MBG and HGI were independent predictors for development of complications (2). Simply stated, at any given level of MBG during the DCCT, patients with higher HGI, and thus higher A1C levels at the same MBG, have higher risk for complications.
In contrast, Lachin et al. (1) chose to test whether HGI is statistically independent of A1C as a predictor for complications. As would be expected, HGI, a component of A1C, is statistically “overpowered” by placing A1C together with it as a covariate in the model, as Lachin et al. have done. But their analysis does not address the original question of whether determinants of A1C other than MBG influence the risk of complications. Of interest, however, Lachin et al. did find that even after removing the effect of MBG, A1C still remains an important predictor of microvascular complications (1). Thus, at any given level of MBG, patients with higher A1C levels have higher risk for complications. Their result clearly agrees with ours on the existence of other factor(s) besides MBG influencing patient A1C levels and also predicting the development of complications. We suggest that the factor(s) underlying between-patient biological variation in A1C contributes to this source of risk.