In most studies, the parameters are derived from modeling glucose, insulin and C-peptide kinetics during specific tests. The methodological aspects and the experimental and analytical methods for the correct assessment of insulin sensitivity and secretion have been subjects of several excellent reviews (21–23).

Mathematical handling of the fasting levels of glucose and insulin concentration has been used for the calculation of insulin sensitivity and secretion by introducing various indices such as HOMA (homeostasis model assessment) and QUICKI (quantitative insulin sensitivity check index) (24, 25). However, these simplistic approaches do not give any clue to the dynamic state of the relationship between insulin sensitivity and secretion.

The gold standard for measurement of insulin sensitivity is the euglycemic hyperinsulinemic clamp technique (26), which is reproducible and sensitive.

A drawback is, however, that this technique yields an estimate of insulin sensitivity calculated for only one of the infinite couples of the glucose/insulin space. To circumvent this drawback, different levels of insulin concentrations are required and the need for reaching and maintaining different steady states possibly within the same experiment makes it long, cumbersome and impossible to perform routinely. Another drawback is that insulin secretion cannot be evaluated from the euglycemic clamp. When using the clamp, therefore, it is necessary to carry out another experiment, such as a primed hyperglycemic glucose clamp, an i.v. glucose tolerance test (IVGTT), an arginine test or any other experiment where the beta cell is stimulated (not necessarily only by glucose) to release insulin (27).

An experimental procedure that includes determination of both insulin sensitivity and insulin secretion is the IVGTT (28, 29) with frequent sampling at the beginning (i.e. FSIGT). This allows measurement of glucose, insulin and more often also C-peptide during the highly dynamic phase that immediately follows the glucose injection. Advantages and disadvantages of this test as well as the different protocols and the relative performance in humans and experimental animals have been described in detail elsewhere (5, 18, 28, 30, 31). Briefly, advantages are that there is a known dose of glucose in the circulating blood and direct stimulation of the beta cell, without confounding effects of incretins or gastrointestinal hormones, and possible problems related to gastric emptying and to delay in glucose absorption do not interfere with the analysis. Disadvantages are that it is unphysiological, not easy to perform, requires the use of computers and sophisticated programming to be analyzed, and is hard to use in epidemiological studies. Once glucose and insulin concentration data are available for the whole duration of the experiment (from 3 to 4 h in humans to 50 min in mice), mathematical model analysis may be performed to calculate metabolic parameters not directly obtainable from simple combinations of the experimental data. The classic Bergman’s minimal model was developed to analyze glucose disappearance during FSIGT in dogs (5). The protocol in humans was then modified with the additional injection/infusion 20 min after that of glucose of exogenous insulin (28, 32, 33) (initially tolbutamide (34)) that augments the whole dynamics of the experiment, yielding a better identification of the parameters (34). In both cases, regular and modified FSIGT, the solution of a system of differential equations yields SI, which quantifies the ability of insulin to enhance glucose uptake by insulin-dependent tissues and to inhibit liver glucose production. Regardless of the protocol used, SI is a robust parameter and turned out to be the same with regular and insulinmodified FSIGTs when performed in the same subjects (29). As regards the analysis of the commonly used, simple and widespread oral glucose tolerance test (OGTT), recently several formulas have been published to calculate insulin sensitivity indices (35–37); their validation in various clinical settings has yet to be established.

There is no consensus on which to assess the reference method of insulin secretion. The simplest one is the calculation of the insulin AUC during the whole test or during specific intervals. This measurement gives an idea of the amount of insulin that acts on the tissues, but cannot add information on the dynamics of the hormone in terms of secretion and clearance. For the FSIGT, it has been proposed to use AIRG, i.e. the mean insulin concentration above basal during the first peak (in general from 2 to 10 min) (8, 9). Also mathematical modeling has been exploited to describe the main processes (secretion, extraction, clearance) during FSIGT (11, 12).

Another test for evaluating beta cell function is the glucose-dependent arginine stimulation test. This test was first developed byWard and collaborators in Seattle (38), and was later characterized in detail (39). This test characterizes the acute insulin secretion at three glucose levels and the glucose sensitivity of the beta cell secretion. The technique has also been used in combination with the euglycemic, hyperinsulinemic clamp, to illustrate the hyperbolic relationship of insulin sensitivity vs insulin secretion (14, 15).

In a recent study, it was also examined whether insulin sensitivity could be estimated from the glucosedependent arginine stimulation test (16). In this test, glucose is infused to raise and maintain glucose levels at 14 mmol/l, and by dividing the amount of glucose infused to reach this value by the resulting insulin level, an indirect measure of insulin sensitivity may be estimated. It was thereby shown that this measure correlated nicely with the measure of insulin sensitivity obtained by the gold standard of the euglycemic, hyperinsulinemic technique (16).