点击显示 收起
Department of Human Nutrition
Centre for Advanced Food Studies
The Royal Veterinary and Agricultural University
30 Rolighedsvej
DK-1958 Frederiksberg C
Denmark
E-mail: bsl{at}kvl.dk
Department of Applied Nutrition & Food Chemistry
Lund University
Sweden
Dear Sir:
We were surprised to read that Brand-Miller finds our title and conclusion too decisive, because we merely stated the findings of our study (both the pros and the cons of low-glycemic-index diets) and specifically pointed out the time frame of 10 wk. Furthermore, we stated that more studies are needed to substantiate our findings.
We deal with the questions raised by Brand-Miller as follows:
We found a significant 10% decrease in LDL cholesterol in the low-glycemic-index group compared with a 2% increase in the high-glycemic-index group. This difference cannot be explained by changes in fat mass, as suggested by Brand-Miller, because we observed no correlation between changes in fat mass and changes in LDL cholesterol (total fat mass: r = 0.17, P = 0.27; trunk fat: r = 0.09, P = 0.56). Nor does inclusion of changes in total or trunk fat mass in the covariance analysis change the significance of the diet effect on LDL cholesterol.
Regarding the power in the study, these calculations were made before the study on the basis of body weight. A postexperimental calculation of power as suggested by Brand-Miller is not appropriate according to an article published in American Statistician (1). As stated in our discussion, we agree with Brand-Miller that a longer study period or inclusion of more subjects might have resulted in a significant difference in body weight. Nevertheless, we maintain that our finding of no significant difference in body weight loss after 10 wk with 22–23 subjects per diet group does at least question the clinical relevance of the glycemic index in body weight control. Furthermore, existing evidence that low-glycemic-index diets are an effective tool for achieving weight reduction is, at present, contradictory (2).
In response to Brand-Miller’s request for data more indicative of visceral fat changes, we reanalyzed our dual-energy X-ray absorptiometry measurements on trunk fat. We found no significant difference between the high-glycemic-index and low-glycemic-index group in trunk fat mass decreases from week 0 to week 10 (low-glycemic-index: –0.45 ± 0.18 kg; high-glycemic-index: –0.18 ± 0.20 kg; P = 0.65). Nor did we find any significant difference between groups in waist circumference changes (low-glycemic-index: –1.9 ± 0.7 cm; high-glycemic-index: –1.1 ± 0.7 cm; P = 0.47).
Brand-Miller also advocates the use of in vivo measurements of glycemic index, but if the glycemic index concept is to make any sense as a tool for consumers, in vivo analysis cannot be a prerequisite for composing a low-glycemic-index diet. Moreover, this requirement is not consistent with Brand-Miller’s own practice. For example, in vivo analysis was not an inclusion criteria in her previously published meta-analysis (3), in which not even in vitro measurements of glycemic index were a requirement. A hydrolysis index of the test foods was the basis for the glycemic index calculation in our study. This method correlates well with in vivo glycemic index determination, especially for starch-rich products such as those used in our study (4, 5). Another laboratory also analyzed the carbohydrate quality of our test foods by using the Englyst method, and the results from this analysis support earlier findings with use of the hydrolysis index method. We also calculated the mean glycemic index value of the 2 groups of test foods with the use of the latest international tables of glycemic index and glycemic load values and found a mean difference of 34.5 units. However, we do agree that it is difficult to predict the glycemic index of a mixed diet from the glycemic index values of the individual food items from table values. Last, we argue that the significant difference in LDL cholesterol between the 2 groups is a good marker for a true difference between the 2 groups of test foods, because the study was designed so that everything other than the glycemic index was similar in the 2 groups.
As reported in our paper, some subjects found the amounts of test food too high (not "far too high" as stated by Brand-Miller). This was not a complaint that continued throughout the study period, but a statement made mostly during the first week of the study period, and for most subjects, it was only addressed when specifically requested. It is not unusual for overweight subjects to complain about quantities of food when they are served low-fat, high-carbohydrate diets that have a high dietary fiber content because these diets are probably more satiating than the subjects’ habitual diets (6).
The dietary instructions were necessarily strict because we wanted to control the glycemic index and macronutrient composition of the subjects’ diets to effectively compare the effects on energy intake of a low-fat, high-carbohydrate diet with either a low or a high glycemic index. However, energy intake was not restricted; we merely supplied the subjects with carbohydrate-rich test foods comprising 49% of their estimated total energy intake. Dietary record data from weeks 5 and 10 clearly show that the test foods provided were only part of the subjects’ diets, as intended.
Finally, Brand-Miller suggests performing an intention-to-treat analysis on body weight changes. As could be expected, this analysis only reduces the difference between the 2 groups and increases the P value (low-glycemic-index, –1.55 ± 0.44; high-glycemic-index, –1.19 ± 0.29; P = 0.57).
ACKNOWLEDGMENTS
The original study was supported by Danone Vitapole, France. Rice was donated by Masterfoods as, Denmark, and Euryza GmbH, Germany, and rye bread by Cerealia R&D, Schulstad Brød A/S, Denmark. None of the authors had any conflicts of interest.
REFERENCES