Effect of Extruded Cereals Enriched With β-glucan on Postprandial Plasma Glucose and Serum Insulin Responses

Overview

β-glucan (BG) is a viscous dietary fiber present in barley and oats that, when added to carbohydrate containing foods, reduces the postprandial glycemic response (PPGR), which is considered to be beneficial to health. In 2011, the European Food Safety Authority (EFSA) concluded that 4g of either oat BG (OBG) or barley BG (BBG) per 30g available carbohydrate (avCHO) is required to obtain a reduction in PPGR, however this is an impractically large amount of BG to incorporate into palatable foods. A recent systematic review and meta-analysis found that doses of OBG well below 4g/30g avCHO can significantly reduce PPGR. It is therefore hypothesized that 25g avCHO portions of cereal containing 1.4-1.5g of either OBG or BBG will elicit a statistically significantly lower PPGR compared to a control cereal containing no BG.

Statistical Analysis: Analysis populations: The intention-to-treat (ITT) population includes all subjects who consumed at least 1 of the 3 test meals in that phase. The per-protocol (PP) population for each phase includes all subjects who completed all 3 tests in that phase with no serious protocol violations. There may also be a group, termed PPDO (dropouts), defined as the PP population plus subjects who competed 1 or 2 of the 3 tests with no serious protocol violations. Conclusions will be based on the glucose and insulin endpoints for the PP population. To assess the potential impact of excluding those subjects who dropped out after completing 1 or 2 tests, a secondary analysis will be performed for the PP+DO population where the missing values are imputed using methods described by Snedecor and Cochran. Distribution of data: the methods of analyzing continuous data assume that the data are normally distributed. Normality of distribution will be tested by comparing the actual distribution of the Z scores to the expected distribution using the chi-square test with the data taken to be normally distributed if p\>0.05. Non-normally distributed data will be transformed by (in order of preference) log-transformation, square-root-transformation, some other method, until the transformed data are normally distributed. Primary endpoint: the primary endpoint will be assessed as follows: serum glucose concentrations will be subjected to repeated-measures analysis of variance (ANOVA) using the linear model analyzing for the main effects of time and treatment and the time×treatment interaction. After demonstrating a significant time×treatment interaction (indicating that the pattern of glucose response differs significantly among treatments), the values for iAUC 0-2h will be analyzed by ANOVA for the main effect of treatment; after demonstrating significant heterogeneity, the mean for each of the 2 test cereals will be compared only to that of each respective control cereal using Dunnett's Test to adjust for multiple comparisons. The 2 test cereals will not be compared to each other. The criterion for significance will be two-tailed p\<0.05. Secondary endpoints: the results for insulin iAUC will be assessed as described for the primary endpoint. Serum glucose and insulin concentrations and increments at each time point will be subjected to repeated-measures ANOVA. Upon demonstration of significant heterogeneity, individual means of the test and matched control cereals will be compared using Dunnett's Test to adjust for multiple comparisons. The 2 test cereals will not be compared to each other. The criterion for significance will be two-tailed p\<0.05. Statistical Power: The average within-individual CV of glucose iAUC elicited by repeated tests of oral glucose is within the range of 22-25% in our previous studies in healthy individuals. To demonstrate that both test cereals differ significantly from the control, each comparison must have 90% power to detect a significant difference in order to obtain 80% power to detect both significant differences (since 0.902 = 0.81). Assuming a SD of 25% and using the normal distribution, n=40 subjects provide 95% power to detect a significant difference of 20% between test and control cereals.