Background: Glucose effectiveness (S(g)) is an important component in glucose tolerance. Values of S(g) using "open loop" glucose kinetic computer programs are usually higher compared to closed loop method (CLM) programs that incorporate insulin secretion modeling. We aimed to test whether these differences are caused by (1) inclusion of insulin secretion modeling or (2) the method of representing plasma insulin values in the first few minutes of the frequently sampled intravenous glucose tolerance test (FSIGT).
Methods: FSIGTs without insulin supplementation were performed in six healthy volunteers, and the Bergman minimal model was fitted to the data using the simulation and modeling program SAAM.
Results: The CLM, which represents the insulin data in the first few minutes by a best-fit curve extrapolated to the y-axis, yielded a significantly lower S(g) than the approach similar to the computer program MINMOD, where the first few minutes of insulin data are represented by a line joining the basal to the peak values (1.55 +/- 0.28 vs. 1.97 +/- 0.27 [SE] x 10(-2)/min, P < 0.05). This second analysis was then repeated while forcing the program to represent the insulin data after the insulin peak in the same way as in the CLM, obtaining an almost identical result for S(g) (1.99 +/- 0.29). Insulin sensitivity was not significantly affected.
Conclusions: The higher S(g) estimates are caused by the method of representing the first few minutes of insulin data rather than by the incorporation of insulin secretion modeling. It is, therefore, important to know how the early insulin data are represented when comparing results from different computer modeling programs.