QUICKI Is Not a Useful and Accurate Index of Insulin Sensitivity following Exercise Training

Extract

To the editor:

We appreciate the comments by Dr. Quon regarding our study (1). The major criticism is based on a perceived problem with the interpretation of the data due to “not having a reference standard for insulin sensitivity to compare with Quantitative Insulin Sensitivity Check Index (QUICKI) and SI_MM.” Although our use of an unmodified frequently sampled iv glucose tolerance test and a reduced sampling schedule might be, as Dr. Quon states, “suboptimal for minimal model analysis,” any error introduced by this technique would be constant, and, therefore, it is unbiased with respect to determining changes in SI_MM and QUICKI that are induced by exercise training. With this in mind, our study showed that changes in SI_MM were not correlated with changes in QUICKI (r = 0.24, P = 0.39) (1).

Dr. Quon remarks that “when compared with a direct measure of insulin sensitivity such as that derived from the glucose clamp, the correlation of QUICKI with SI_Clamp is significantly stronger than the correlation between SI_MM and SI_Clamp.” However, our inspection of the data (2) on which this statement is based yields a different interpretation. In this study, Katz et al. (2) reported that the overall correlation between QUICKI and SI_Clamp is r = 0.78, whereas the overall correlation between SI_MM and SI_Clamp is r = 0.57, when groups of nonobese, obese, and diabetic subjects are combined. The subgroup correlations between QUICKI and SI_Clamp are r = 0.49 for nonobese, r = 0.89 for obese, and r = 0.70 for diabetics. The subgroup correlations between SI_MM and SI_Clamp are essentially identical for the nonobese (r = 0.48) and obese (r = 0.82) groups, however, the low correlation between SI_MM and SI_Clamp for diabetics (r = 0.51) attenuates the overall correlation between these measures. This is because the insulin response to iv glucose is diminished in diabetics, which mitigates the use of SI_MM for this group. Indeed, 7 of the 15 diabetic subjects studied by Katz et al. (2) were eliminated from the analysis because of nonsensical SI_MM values. Thus, the significantly stronger correlation of QUICKI and SI_Clamp is due to the low correlation of SI_MM and SI_Clamp in the diabetic group. Our study did not suffer from this flaw because all of our subjects were nondiabetic and had an adequate insulin response to iv glucose.