Reply to R Mendosa

¹ Human Nutrition Unit, School of Molecular and Microbial Biosciences, University of Sydney, New South Wales 2006, Australia. E-mail: j.brandmiller{at}biochem.usyd.edu.
² Department of Mathematical Sciences, University of Technology, New South Wales 2007, Australia.

Dear Sir:

Recently, the American Diabetes Association published the following statement: "with regard to the glycemic effects of carbohydrates, the total amount of carbohydrate in meals and snacks is more important than the source or type" (1). However, the revised glycemic index (GI) table published in the July 2002 issue of the Journal (2) indicates that even when foods contain the same amount of carbohydrate (ie, carbohydrate exchanges), there are up to 10-fold differences in the glycemic effect among them. Moreover, prospective observational studies found that the risk of developing diseases of affluence (eg, type 2 diabetes, cardiovascular disease, and some cancers) is independently related to the overall dietary GI and glycemic load (GL), but not to the total carbohydrate content (3). GL—the product of the carbohydrate content of a serving of food and its GI—is a measure of both the quantity and quality of carbohydrate and the interaction between them. Thus, in the 2002 table (2), we included not only the GI values of the foods but the carbohydrate content per nominal serving and its calculated GL value. In the table, each unit of GL is the glycemic equivalent of 1 g of pure glucose.

The question remains, however, which is the stronger determinant of GL: the amount of carbohydrate per serving or the GI? At Mendosa’s suggestion, we used the data published in the revised tables to explore the statistical relations between GI, GL, and carbohydrate content.

As the first step, we excluded data for foods in which carbohydrate was either a very minor component (high-protein meal-replacement shakes) or was only partially absorbed (sugar alcohols and bulking agents), producing a final data set with a total of 1058 entries. Among these, the amount of carbohydrate per serving varied from 3 to 75 g ( ± SD: 25 ± 12 g) and the GI values ranged from 5 to 139 (54 ± 19). Subsequently, we performed a regression analysis to determine how much of the variation in GL values was explained by the amount of carbohydrate in nominal portions of foods versus the corresponding GI values. Obviously, both factors will be directly related to GL, because the values were calculated by using the following formula: GL = (amount of carbohydrate in the portion of food x GI value for the food)/100 ( The data were transformed onto a log scale, because GL values are a product of the multiplication and not the addition of the carbohydrate content and the GI. Linear regression analysis with the use of the log scale indicated that the carbohydrate content alone explained 68% of the variance in GL values, whereas the GI value alone explained 49% of the variance in GL values. The results were similar when the data were analyzed on the ordinary scale, as shown in Figure1. Hence, although carbohydrate is indeed the greater determinant of GL, it makes no sense to ignore a factor that accounts for almost 50% of the variation in GL on its own and an additional 32% of the variation when carbohydrate is also included in the linear regression analysis

View larger version (24K):
FIGURE 1. . Scattergrams of the correlation between the glycemic load (GL) values for the foods listed in the international tables and their corresponding glycemic index (GI) values and carbohydrate contents (n = 1058) on the ordinary scale.

 
Mendosa’s second question relates to appropriate cutoffs for high, medium, and low GLs. Currently, there is little experience with the use of GL values in a practical dietetic setting. As a starting point, we suggest that the preliminary cutoffs be 10 for a low GL and 20 for a high GL. Hence, a GL between 11 and 19 represents a medium GI. These cutoffs should be reevaluated in the future.

It remains to be shown, of course, that the concept of GL makes physiologic sense. Currently, 2 assumptions need to be tested in vivo: 1) that food portions calculated to have the same GL produce similar glycemic responses and 2) that stepwise increases in GL produce proportionate increases in glycemia and insulinemia. Such relations need to be proven in a variety of subjects (eg, lean and obese and insulin-sensitive and insulin-resistant). Until then, it makes sense to consider both the carbohydrate content and the GI of foods in the research and management of postprandial glycemia.

REFERENCES

1. Franz MJ, Bantle JP, Beebe CA, et al. Evidence-based nutrition principles and recommendations for the treatment and prevention of diabetes and related complications. Diabetes Care 2002;25:148–98.
2. Foster-Powell K, Holt SHA, Brand-Miller JC. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr 2002;76:5–56.
3. Ludwig DS. The glycemic index—physiological mechanisms relating to obesity, diabetes, and cardiovascular disease. JAMA 2002;287:2414–23.