This investigation provides preliminary evidence suggesting that a standard liquid meal tolerance test may be used to assess indices of insulin sensitivity and secretion across the spectrum of glucose tolerance. The results conformed to the expected relationships between insulin secretion and sensitivity, as indicated by the AUCins/glu and the Matsuda index, respectively. The product of these two indices, which is analogous to the disposition index (acute insulin response * insulin sensitivity index) described by Kahn et al. [8, 9] was able to discriminate between the three categories of fasting glucose tolerance, even with relatively small numbers of subjects. This product may therefore provide a useful measure of the ability of the pancreas to match insulin secretion to the prevailing degree of insulin resistance.
Intravenous glucose tolerance tests and glycemic clamp procedures are not practical for many clinical and epidemiologic investigations because of the cost and high level of sophistication required to conduct such tests. The oral glucose tolerance test provides a possible alternative for use in such instances. However, a meal tolerance test has a number of potential advantages compared to the oral glucose tolerance test. Because insulin resistant states are associated not only with disturbances in carbohydrate metabolism, but also with fasting and postprandial lipid levels, it may be desirable to simultaneously assess post-load glucose and lipid responses, without the need for tests on separate days. A mixed meal tolerance test also provides a more physiologic stimulus for assessing incretin responses . Some subjects experience nausea or reactive hypoglycemia after a load consisting of glucose only. Neither of these issues was experienced by any of the 73 subjects in our sample, although subjects with a history of signs or symptoms of hypoglycemia were excluded from participation. In addition, the liquid meal used in our study is a commercially available product that is easily reproducible and requires no food preparation, thus can be employed easily in a standardized fashion across multiple research sites.
Other investigators have studied liquid or solid mixed meal loads for assessing insulin secretion or sensitivity [12–18]. In particular, Aloulou and colleagues  evaluated the ability to estimate insulin sensitivity using a standard breakfast test with various methods, most of which were originally validated using oral glucose tolerance test data. Using the minimal model-derived insulin sensitivity index from intravenous glucose tolerance test results as their standard for comparison, they found that the Matsuda index was the strongest correlate of insulin sensitivity (r = 0.656, p < 0.001). Retnakaran et al.  also recently reported that the product of the Matsuda index and the AUCins/glu was the best indicator of the appropriateness of insulin secretion for the level of insulin sensitivity among several measures they evaluated. These findings prompted the authors to select the Matsuda index and the AUCins/glu as the two measures for evaluation in the present study. In addition, these indices can be easily calculated using a spreadsheet, without the need for sophisticated mathematical modeling software. The liquid test meal used in the present investigation was similar to those employed previously in studies of insulin secretagogue medications [15, 24].
The current study did not have gold standard indices of insulin secretion or sensitivity against which to compare the liquid meal tolerance test derived indices. The strong inverse correlations between the HOMA2-IR and Matsuda indices in all three fasting glucose tolerance groups (r = -0.914 to -0.940) suggest that both are measuring the same underlying physiologic process. Means for both indices showed the expected order, in that subjects with diabetes had the greatest degree of insulin resistance, subjects with NFG had the least, and subjects with IFG had intermediate values. Some caution is warranted regarding the correlation between these measures because the HOMA2-IR value is based on the product of fasting insulin and glucose levels, whereas the Matsuda index contains this product as an element of the calculation. However, both have also been shown to correlate well with minimal model and euglycemic clamp assessments of insulin sensitivity [4, 25].
Using a minimal model analysis of data from meal tolerance tests, Steil et al.  showed that the insulin sensitivity index from a meal tolerance test correlated with that from a euglycemic clamp (r = 0.760, p < 0.001) to a degree similar to that derived from an intravenous glucose tolerance test (r = 0.717, p = 0.001), but that the index derived from the meal tolerance test was consistently higher than that from the intravenous glucose tolerance test by a factor of ~2.3. These results are in agreement with those from Caumo and colleagues  who also found a strong correlation between insulin sensitivity indices derived from minimal model analyses of meal and intravenous glucose tolerance tests, but with 2.2-fold higher values from the meal tolerance test (Spearman r = 0.89, p < 0.01). Thus, the available data suggest that a meal tolerance test (liquid or solid) can provide an appropriate stimulus for assessing insulin sensitivity, although absolute values from such tests may not be directly comparable to those derived from the euglycemic clamp or intravenous glucose tolerance test.
The authors were surprised that HOMA2-%B did not discriminate between the three fasting glucose tolerance groups. The subjects with diabetes generally had good glycemic control as indicated by an average HbA1c concentration <7.0% and were all taking sulfonylureas, which may have contributed to higher values than might have been the case otherwise. However, very similar values were present in subjects with diabetes, IFG and NFG. In contrast, AUCins/glu did discriminate between diabetes and the other categories, although not between the NFG and IFG subjects. The product of the Matsuda index and AUCins/glu discriminated between all three groups, consistent with the view that greater insulin secretion was partially compensating for the somewhat higher level of insulin resistance present in the subjects with IFG, and underlining the fact that the appropriateness of the pancreatic beta-cell insulin response can only be interpreted in relation to the prevailing degree of insulin resistance. Recently, Festa and colleagues  reported that HOMA-%B underestimated the degree of beta-cell dysfunction in subjects with impaired glucose tolerance and early-stage, asymptomatic type 2 diabetes, thereby supporting our results and confirming that post-load measures of beta-cell function are more sensitive indicators of beta-cell dysfunction than those derived from fasting values.
The relationship between insulin sensitivity (Matsuda index) and insulin secretion (AUCins/glu) in the present study was consistent with the expectation that progressive worsening of glucose tolerance status from NFG to IFG to diabetes would be associated with shifts downward and to the left in the regression curves. However, slope of the regression line for the equation loge AUCins/glu = intercept * slope (loge Matsuda index) did not approximate -1.0 in any of the three fasting glucose tolerance groups, as would be expected based on results for the oral glucose tolerance test described by Retnakaran et al. . Values for the slope of this regression line ranged from -0.734 to -0.548. The reason for this difference between our results from liquid meal tolerance tests and theirs from oral glucose tolerance tests is unclear and will need to be investigated further. As discussed above, findings from other investigators have suggested that insulin sensitivity indices from mixed meal tests correlate with those from methods that use an isolated glucose stimulus, but are not always directly comparable in absolute terms. One potential explanation for the apparent difference is that differences in incretin responses elicited by a mixed meal, compared with an oral or intravenous glucose stimulus, alter the Matsuda index to AUCins/glu relationship .
The present investigation was intended as an initial evaluation of particular measures of insulin secretion (AUCins/glu) and sensitivity (Matsuda index) derived from a standard liquid meal tolerance test. The results are encouraging, but should be considered only a first step. Additional studies will be needed to compare this approach directly to established methods such as minimal model analyses, as well as euglycemic and hyperglycemic clamp methods for assessing insulin sensitivity and secretion. Moreover, studies will be required to assess the test-retest coefficient of variation in order to facilitate sample size calculations for clinical intervention trials. While it can be argued that the use of systemic insulin concentration as a surrogate marker of beta-cell function was flawed because it reflects both pancreatic secretion and hepatic extraction, the objective of this trial was to assess the utility of methods previously evaluated using oral glucose or mixed solid meal tolerance test data, but substituting a mixed liquid meal as the stimulus. Those previous examinations collected pre- and post-load insulin and glucose data, and we followed the same approach.
Some additional limitations of the present study should also be considered. The sample was relatively small and selected for participation in a clinical trial. Subjects with diabetes had good average glycemic control and were all taking sulfonylurea medications. The subjects with IFG did not cover the full range of fasting glucose levels for this category due to the exclusion criteria established for participation in the trial. Because no oral glucose tolerance test was completed, it was not possible to categorize glucose tolerance in subjects without diabetes beyond the classifications of NFG and IFG. Several investigators have shown that subsets of subjects with pre-diabetes such as those with isolated IFG, isolated impaired glucose tolerance, or both, differ regarding the degree to which peripheral insulin resistance, hepatic insulin resistance and beta-cell dysfunction contribute to their type and degree of glucose intolerance [3, 27–29]. In the present study, some of the subjects with NFG and IFG might have been shown to have impaired glucose tolerance if oral glucose tolerance test results had been available. Therefore, testing in a broader sample across the entire spectrum of fasting and post-load glucose tolerance will be needed to demonstrate the generalizability of our results. Simultaneous measurement of incretins and parameters related to lipid metabolism would help to establish the utility of the standard liquid meal tolerance test for evaluation of multiple metabolic responses.