We read with interest the correspondence between Drs. van Haeften (1) and Matsumoto et al.(2) in a recent issue of Diabetes Care about the correct formula for insulin resistance calculated by homeostasis model assessment (HOMA). The HOMA model (3) is a structural computer model of the glucoseinsulin feedback system in the homeostatic (overnight-fasted) state. The model consists of a number of nonlinear empirical equations describing the functions of organs and tissues involved in glucose regulation. These are solved numerically to predict glucose, insulin, and C-peptide concentrations in the fasting steady state for any combination of pancreatic beta-cell function and insulin sensitivity (or resistance). These predictions allow the deduction of beta-cell function (% beta) and insulin sensitivity (% S) from pairs of fasting glucose and insulin (or Cpeptide) measurements. The nonlinearity of the model precludes an exact algebraic solution, but estimations are possible either graphically or by using simple mathematical approximations, as presented in Matthews et al.(3): R (which is the inverse of% S)=(insulin X glucose)/22.5 and% beta= 20 x insulin/(glucose-3.5). The apparent redundancy of the expression in question was due to the removal of terms from an original, more complex, expression. Two developments have taken place since 1985.

First, the physiological basis of the model has been developed, both in terms of insulin secretion (4) arid glucose metabolism (5), and further modifications have included a model of proinsulin secretion, allowing HOMA to be used with either immunoreactive insulin or specific insulin assays, and a model of renal glucose losses, allowing its use in more hyperglycemic states. These modifications provide a more accurate representation of physiology and successfully predict the homeostatic responses to an intravenous glucose infusion. The use of the current HOMA model performs well in comparison with several tests of insulin sensitivity, including the intravenous glucose tolerance test, and with minimal model analysis and the short insulin tolerance test of Bonora (6) and tests of beta-cell function (7), including the hyperglycemic clamp (8), the oral glucose tolerance test (9), and the frequently sampled intravenous glucose tolerance test (FSIVGTT)(7).