Buyken and Liese (1) raised the relevant question of whether low variability in glycemic index and glycemic load could explain the lack of association with insulin resistance (2). The complete ranges (medians) of glycemic index and glycemic load in our study are 16–105 (79) and 0–1,208 (197), respectively. This is in accordance with previous studies (3), and thus, it is unlikely that this explains the lack of association.
We disagree that our article should have provided data on associations of glycemic index and glycemic load with other (dietary) variables because this would have expanded the extent of the article considerably and furthermore blurred the focus of the article.
We are aware of the methodological problems related to dietary assessment methods including estimation of glycemic index (2). Unfortunately, we cannot change the fact that information on intake of soft drinks and selected sweet products were not available in our study. Soft drinks may not, however, contribute substantially to the daily intake of glycemic index–inducing carbohydrates (4), despite the high–glycemic index value of sucrose. Additionally, the intake of sucrose from sucrose-containing foods and soft drinks is not consumed in large amounts in the general Danish population (25–65 years) (5). Thus, the lack of data on soft drinks and selected sweet products may not be a major concern.
Bias introduced in all dietary studies with underreporting cannot be excluded (2). It is, however, impossible to estimate the exact degree of underreporting. Therefore, we do not have a rational basis for a sensitivity analysis. Hence, we would have to make up a set of assumptions regarding the magnitude of underreporting intake of socially undesirable sucrose-containing foods. These assumptions would never be reliable. Thus, we recommended caution when interpretation of results from this part of our study.
The last point raised by Buyken and Liese was that alcohol intake may predict glycemic index of the diet and hence affect the estimates of glycemic index and glycemic load. The reported associations are relative changes in the continuous variable homeostasis model assessment of insulin resistance (HOMA-IR) expressed as HOMA ratios with corresponding 95% CIs. After adjustment for alcohol in the multiple analyses, glycemic index demonstrated an insignificant association with HOMA-IR (HOMA ratio 0.99 [95% CI 0.97–1.02]; P = 0.60) and glycemic load demonstrated a significant inverse association with HOMA-IR (HOMA ratio 0.99 [0.98–1.00]; P < 0.01).
