In their important analysis of data from the Inter99 Study on the relationship among glycemic index, glycemic load, and insulin resistance as estimated by the homeostasis model, Lau et al. (1) unfortunately do not provide adequate descriptive information on the distribution and variation in levels of glycemic index and glycemic load in their population. Additionally, the reader is left wondering about the associations of glycemic index and glycemic load with other (dietary) variables.
These types of information are critical for comparison and interpretation of the Inter99 Study to other studies. To date, the strongest association between dietary glycemic index and risk of type 2 diabetes was reported from the study with the largest variation in dietary glycemic index (2). In the absence of the respective data for the Inter99 Study, it is difficult to evaluate whether small variations in the levels of glycemic index and glycemic load could be responsible for the lack of an association with insulin resistance. A small variability can in turn be either inherent to the population or result from the dietary assessment method.
First, some indirect evidence for the latter comes from the fact that the authors used a total of only 57 glycemic index values to estimate the dietary glycemic index of all participants. Second, intakes of soft drinks and selected sweet products were not assessed; however, most of these foods have a high glycemic index and are highly predictive of the overall dietary glycemic index and glycemic load (3,4). In addition, the consumption of socially undesirable sucrose-containing foods may have been selectively underreported by the Inter99 participants, who were invited to partake in a health survey. Although most sucrose-containing foods have only intermediate glycemic index levels, they are often consumed in large amounts. A selective underassessment may thus affect the estimates of glycemic index, glycemic load, and sucrose without affecting estimates of dietary fiber intake. In this context, the discussion of reasons for the lack of an association between sucrose and the homeostasis model may need reconsideration given that sucrose has a glycemic index of 97 (white bread standard), which is very similar to the glycemic index of white bread, which is 100. Finally, alcohol intake was not considered in glycemic index and glycemic load estimation but has been shown to be highly predictive of glycemic index (3).
Thus, in conclusion, this discussion of the article by Lau et al. points out some of the challenges and complexities faced by applying the concept of glycemic index estimation to dietary data collected with a food frequency questionnaire.