**Insulin sensitivity based upon the OGTT**

The previous discussion suggests that “quick and dirty” indices from fasting values may not be accurate, and it is often true that laboratory procedures are impractical. Recently, it has been suggested that indices calculated from the OGTT may be useful for population studies. For decades the OGTT has been used as a screening test and a diagnostic tool for diabetes and IGT [139,140]. It might be ideal if one could classify the state of carbohydrate tolerance and simultaneously measure insulin sensitivity and β-cell function. As long ago as 1981, Voors et al. suggested the mathematical product of the 1-h glucose and the 1-h insulin concentration as “peripheral insulin resistance” [141]. Ten years later, Cederholm and Wibell expressed sensitivity as the ratio of the glucose metabolic clearance rate to the logarithmic transformed mean serum insulin concentration [141]. This latter index was later modified by Gutt et al. and shown to correlate with insulin sensitivity values derived from the euglycemic clamp [142]. More recently, Matsuda and DeFronzo applied a square root conversion of the products of fasting OGTT glucose and insulin to correct for the nonlinearity of the values [143]. Stumvoll et al. performed multiple regression of several demographic parameters (body mass index, sex, waist to hip ratio, OGTT glucose and insulin) upon sensitivity from the euglycemic hyperinsulinemic clamp test [144]. Thus, their approach included as independent variables metabolic characteristics which themselves can alter insulin action [145 – 148].

The OGTT-derived estimates of insulin sensitivity mentioned earlier correlated with insulin sensitivity values obtained by the euglycemic clamp technique (Table 15.2). Several of the indices were additionally tested in different ethnic groups [149 – 151]. Kanauchi et al. showed that Matsuda’s (r = 0.45), Gutt’s (r = 0.64), and Stumvoll’s (r = 0.53) formulas correlated modestly in a Japanese population with insulin sensitivity data from euglycemic clamps [151]. Kuo et al. reported the Matsuda index and the homeostatic model assessment equally applicable for estimation of insulin sensitivity in Chinese diabetic patients and their offspring [150]. In contrast, Chiu et al. [149] reported that the estimated indices (Matsuda and Stumvoll) correlate with clamp-derived insulin sensitivity in a wide spectrum of glucose-tolerant healthy subjects, but that they are less likely to detect differences in insulin sensitivity among different ethnic groups (Asian Americans, African Americans, Caucasian Americans, and Mexican Americans). However, the estimations of insulin sensitivity and β-cell function by OGTT-derived methods failed to reproduce the hyperbolic relation [152], which is well established for the relation between β-cell function and insulin sensitivity when measured by more established methods [106].

It is clear that a correlation between proposed OGTT-derived indices and clamp-derived insulin sensitivity may exist. However, the OGTT-derived indices contain endogenous insulin response as an important term in their calculation and it cannot be excluded that reported correlations reflect islet cell response rather than insulin sensitivity per se. If it is true that OGTT-based methods reflect secretion rather than sensitivity, application of the OGTT methods to subjects with impaired β-cell secretory capacity (e.g. those with impaired glucose tolerance, gestational diabetes, T2DM) may underestimate insulin resistance, as postload hyperinsulinemia would be reduced and they may not be accurate for diabetic patients. Confirming this latter weakness, correlations between OGTT-based indices and clamp-based sensitivity [143,153] were lower when the indices are employed in subjects with T2DM.

Poor reproducibility of the 75 g OGTT also limits the value of the OGTT [154]. The overall test-to-test reproducibility (coeffi- cient of variation) of the OGTT is no more than 65% [155]. The inherent variability is likely due to high day-to-day variability in gastrointestinal function (gastric emptying, absorption, incretin effect). It is well established that a delay in gastric emptying or reduced glucose absorption will result in lower glucose excursions as well as lower insulin levels [84,156,157]. In fact it has been demonstrated that gastric emptying alone accounts for 35% of the variance in peak plasma glucose after a 75 g OGTT in both healthy volunteers and patients with T2DM [84].

We have investigated the possible effects of variability in islet cell function as well as variability in intestinal glucose absorption on glucose and insulin patterns and OGTT-based indices using a computer model [158]. The glucose and insulin time series simulated for a normal healthy subject closely match published data [159]. The OGTT data simulated under these conditions indicate that ±50% changes in β-cell sensitivity or glucose absorption have a far greater impact on the pattern of glucose and insulin than alterations of insulin sensitivity itself. An imposed 50% reduction of insulin sensitivity caused only very minor changes in sensitivity calculated by those indices (Table 15.3). In contrast, isolated alterations of β-cell sensitivity and glucose absorption resulted in considerably higher or lower apparent values of insulin sensitivity from almost all indices. While these results are from a simulation study, they suggest strongly that one must be very careful in interpreting differences in OGTT values in terms of insulin sensitivity per se because differences in metabolic parameters other than insulin sensitivity may have greater effects on OGTT. Therefore alterations in OGTT cannot be readily interpreted to reflect changes in insulin sensitivity alone, but may reflect changes in gastric emptying, insulin secretion, glucose effectiveness, and so on. Clearly, however, the OGTT (or meal tolerance tests) are far superior for assessment of insulin sensitivity than measures based upon fasting glucose and insulin alone.

**Conclusions**

The objective to accurately measure insulin sensitivity (i.e., “insulin resistance”) in patients remains an important goal. Insulin resistance is a risk factor for a plethora of chronic diseases (diabetes, cardiovascular disease, hypertension, colon cancer). Extensive efforts are underway to find genetic variants underlying insulin resistance. To study these diseases, and to intervene early in their pathogenesis, requires that we quantify insulin resistance as one important risk factor. To measure sensitivity accurately it is best to use a method that observes in some fashion the metabolic effect of insulin given intravenously. Thus, the glucose clamp and the minimal model are accurate and reproducible methods. While one must remain vigilant regarding limitations of these methods, they may be applied with confidence under a wide variety of conditions.

Unfortunately it is not possible to be equally sanguine in recommending simpler (i.e., surrogate) methods. While it may be appropriate to use fasting insulin to reflect insulin resistance in patients without any degree of β-cell failure, it is usually not possible to determine a priori whether any such defect exists. Thus, one can report fasting insulin simply as a qualitative reflection of insulin resistance. Indices that also include glucose appear to add little to the fasting insulin itself. There is strong evidence that the HOMA index does not measure insulin resistance accurately, and in fact may be a better index of β-cell response. Because the purpose of using surrogate methods is to differentiate these two physiologic functions—insulin sensitivity versus β-cell response—it is not possible to recommend the use of surrogate methods based upon fasting values, or at least be highly sensitive to the limitations of using them. Thus, if a genetic variant related to HOMA (or its logarithmic cousin, QUICKI) is identified, one cannot say with confidence the phenotypic trait related to the measured value. Caution in interpretation is the watchword.

Insulin sensitivity measures based upon oral glucose or meals are superior to fasting measure-based surrogates. One must be cognizant, however, of the possible involvement of alterations in nutrient absorption as well as gastrointestinal hormones in the values which emerge. Hormones such as GLP-1 alter insulin action [160], and such effects will alter the sensitivity measures emanating from the oral tests.

Clearly more work is justified to identify new and novel methods to obtain accurate and precise measures of insulin action *in vivo *from simple-to-perform tests. Among others, we continue this quest in our own Institute.

**Click here to view all Chapter 15 references.**