点击显示 收起
Division of Epidemiology and Community Health
University of Minnesota
1300 South Second Street, Suite 300
Minneapolis, MN 55454
E-mail: pereira{at}epi.umn.edu
Dear Sir:
Mayer-Davis feels that further methodologic research is needed to better understand the glycemic index (GI) before moving forward with more definitive randomized trials on GI and body weight regulation and related health outcomes. As evidence that GI has questionable clinical importance in free-living settings, she refers to her null findings on associations between GI derived from a food-frequency questionnaire (FFQ) and measures of fasting and postprandial glycemia in a relatively small study of a multiethnic cohort (1). However, the use of glycemic endpoints, as attractive as they are for characterizing diabetes risk, does nothing to overcome the short-comings of the FFQ that I pointed out in my editorial (2) on the study by Hare-Bruun et al (3). Whether the endpoint is body weight, serum glucose concentration, glycosylated hemoglobin, or frank diabetes or any other malady, if dietary intake is assessed with an instrument of dubious validity, a good chance exists that the finding will be null and uninformative. Indeed, the validation study of the FFQ used in the study described by Mayer-Davis revealed correlation coefficients for total carbohydrate, between the FFQ and eight 24-h dietary recalls, ranging from 0.25 to 0.64 across ethnic groups. Overall, a validity correlation of 0.37 was noted for energy-adjusted, log-transformed carbohydrate. That is, the FFQ-derived carbohydrate intake explained only 14% of the variance in the 24-h recall-derived carbohydrate intake (4). Bias toward the null and negative publication bias are rampant in secondary data analysis of existing cohort studies. Certainly, future studies should aim to test the validity of GI and glycemic load (GL) from FFQ- and diet history–based data. My strong suspicion is that the validity correlation will be similar to that for total carbohydrate—somewhere on the order of 0.4 to 0.6.
Several fruitful, well-controlled intervention studies on the topic of GI and body weight regulation have been conducted in the recent past (5-9). Although not entirely consistent in their findings, these studies laid the groundwork and provided the preliminary evidence in support of larger and longer studies. Whereas more contributions may come from basic food chemistry studies that attempt to learn more about the effect of specific starches, sugars, and fibers on postprandial glycemia, it would not be prudent to lose the momentum we have gained from the more applied, clinical intervention studies. A recent randomized controlled feeding trial conducted by McMillan-Price et al (5) examined daily variations in blood glucose and insulin from morning to evening in 129 subjects randomly assigned to high- and low-GI diets at different carbohydrate and protein concentrations. GI had a direct and important effect on the 10-h blood glucose and insulin profiles, independent of amount of carbohydrate and protein in the diet. The study of McMillan-Price et al provided evidence in support of Mayer-Davis's concern that the effect of GI on glycemia may be attenuated later in the day, likely as a result of the effect of previous meals and variations in physical activity. However, large and statistically significant differences in the 10-h area under the curve for glucose and insulin were observed, which documented the important aggregate effect of a GI that was based on several meals and snacks fed from morning to evening in a manner mimicking free-living conditions (5). Over the course of the intervention, the low-GI diet was observed to have beneficial effects on body fat changes and on some risk factors for cardiovascular disease (5). Similar effects on changes in disease risk factors were noted in a weight-loss intervention isocalorically comparing high- and low-GL diets (6). In addition, although the expected 10% decrease in resting energy expenditure (the main outcome of the study) was observed during 10% weight loss with the high-GL diet, that change was attenuated to 5.5% with the low-GL weight-loss treatment (P < 0.05 for treatment effect; 6). Collectively, these studies clearly show that GI and GL are valid concepts in mixed meals in real-world, free-living settings. Some larger and longer, hypothesis-driven, controlled intervention studies of the effect of GI and GL on body weight regulation and related chronic disease risks are ready and waiting to be conducted, and some likely are in progress.
ACKNOWLEDGMENTS
The author had no personal or financial conflict of interest.
REFERENCES