| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ORIGINAL RESEARCH COMMUNICATION |
1 From the Human Nutrition Unit, University of Sydney, Sydney, Australia (AWB, JM-P, VMF, and JCB-M); the Department of Statistics, Macquarie University, Sydney, Australia (PP and TP); the Department of Ophthalmology, Centre for Vision Research, Westmead Millennium Institute, Westmead Hospital, University of Sydney, Sydney, Australia (VMF and PM); and the NSW Centre for Public Health Nutrition, Human Nutrition Unit, University of Sydney, Sydney, Australia (VMF)
2 Address reprint requests and correspondence to JC Brand-Miller, Human Nutrition Unit, University of Sydney, Sydney, NSW, Australia 2006. E-mail: j.brandmiller{at}mmb.usyd.edu.au.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Objective:The objective was to evaluate the association between GI, GL, and chronic disease risk with the use of meta-analysis techniques.
Design:A systematic review of published reports identified a total of 37 prospective cohort studies of GI and GL and chronic disease risk. Studies were stratified further according to the validity of the tools used to assess dietary intake. Rate ratios (RRs) were estimated in a Cox proportional hazards model and combined by using a random-effects model.
Results:From 4 to 20 y of follow-up across studies, a total of 40 129 incident cases were identified. For the comparison between the highest and lowest quantiles of GI and GL, significant positive associations were found in fully adjusted models of validated studies for type 2 diabetes (GI RR = 1.40, 95% CI: 1.23, 1.59; GL RR = 1.27, 95% CI: 1.12, 1.45), coronary heart disease (GI RR = 1.25, 95% CI: 1.00, 1.56), gallbladder disease (GI RR = 1.26, 95% CI: 1.13, 1.40; GL RR = 1.41, 95% CI: 1.25, 1.60), breast cancer (GI RR = 1.08, 95% CI: 1.02, 1.16), and all diseases combined (GI RR = 1.14, 95% CI: 1.09, 1.19; GL RR = 1.09, 95% CI: 1.04, 1.15).
Conclusions:Low-GI and/or low-GL diets are independently associated with a reduced risk of certain chronic diseases. In diabetes and heart disease, the protection is comparable with that seen for whole grain and high fiber intakes. The findings support the hypothesis that higher postprandial glycemia is a universal mechanism for disease progression.
Key Words: Glycemic index • glycemic load • dietary carbohydrates • epidemiology
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
60% of all deaths, and the proportion is predicted to increase to 75% by the year 2020 (1, 2). Habitual diet is the major modifiable risk factor, and the identification of simple, cost-effective strategies for prevention and management is a matter of urgency. Although changes in the quantity and quality of fat have received considerable attention, the role of carbohydrates is less clear (2). Increases in refined sugar intake have been accompanied by more subtle changes in starchy foods, eg, processed cereal products have replaced more traditionally processed grains. Because carbohydrate is the main dietary component affecting insulin secretion and postprandial glycemia (3), it is implicated in the etiology of many chronic diseases. Both the amount and type of carbohydrate consumed have an effect on both insulin secretion and postprandial glycemia, with differences not explained by glucose chain length (4). In 1981, the concept of the glycemic index (GI) was introduced by Jenkins et al (5) to quantify the glycemic response to carbohydrates in different foods. Glycemic load (GL), the mathematical product of the GI of a food and its carbohydrate content, has been proposed as a global indicator of the glucose response and insulin demand induced by a serving of food (6).
The results of studies that investigated the association between overall dietary GI, GL, and disease risk have been inconsistent. With respect to diabetes, a positive association was documented in 6 large cohort studies (6-11), but no association was seen in 2 others (12, 13). In cardiovascular disease, 2 studies reported a positive association (14, 15), whereas 1 found no relation (16). Most of the studies that have investigated cancer risk have reported no associations (11, 17-29), but there are notable exceptions (30-37). Two studies that investigated the risk of gallbladder disease showed positive associations (38, 39). Finally, 2 studies (40, 41) reported an association with eye disease, whereas a third found no association (42).
Of concern, 5 (13%) (22, 25, 27, 31, 33) of the 37 prospective studies that investigated the relation between dietary carbohydrates, GI, GL, and chronic disease risk did not validate carbohydrate intake, and an additional 5 (13%) (12, 13, 20, 36, 37) showed correlation coefficients for total carbohydrate of <0.5. Another 2 (5%) studies (29, 32) appear to have been validated, but the validation study has not been published, and 2 others (5%) were validated in a population markedly different from the cohort under investigation (8, 21).
To help resolve these inconsistencies, we used meta-analysis techniques to evaluate the association between GI, GL, and risk of certain chronic diseases by reanalyzing the primary data. In the first instance, we examined all prospective cohort studies and then reanalyzed only those studies that were appropriately validated in a comparable population.
|
SUBJECTS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The contents of 274 abstracts or full-text manuscripts identified through the literature search were reviewed independently by 2 investigators (AWB, JM-P) in duplicate to determine whether they met the eligibility criteria for inclusion. Studies were eligible for inclusion if the design was that of a prospective cohort study and the final outcome was occurrence of a chronic disease, but not its risk factors. The flow of studies in this analysis is depicted in Figure 1
. We included a total of 37 prospective cohort studies representing data from 1 950 198 participants (6-42). All of these prospective cohort studies were conducted in free-living individuals with no history of the disease under investigation.
|
Statistical analysis
Some studies included in our meta-analysis differed in the units used to report levels of GI and GL (glucose = 100 versus bread = 100). We therefore converted those using the bread scale to the glucose = 100 scale, using a conversion rate of 0.71 (ie, bread Scale x 0.71 = glucose scale).
Rate ratios (RRs) for a comparison of the highest with the lowest quantiles were used as the measure of association between GI, GL, and risk of developing a chronic disease. Statistical testing for heterogeneity indicated that there was significantly different variability between studies; hence, although both fixed- and random-effects models yielded similar findings in many cases, results from the random-effects models are presented here. This also takes into account the different disease conditions, study duration, and dietary GIs and GLs that were reported from the original prospective cohort studies. To assess the potential for publication bias, we constructed funnel plots for each outcome in which the log RR was plotted against their SE (43). We also conducted a sensitivity analysis in which each prospective cohort study was excluded in turn to evaluate the influence of that prospective cohort study on the estimate. All analyses were conducted in parallel by 2 investigators (PP and TP) using 2 different meta-analysis packages: Comprehensive Meta Analysis version 2.2 (Biostat, Englewood, NJ; Internet: http://www.meta-analysis.com/) and R version 2.4.1 with its contributed package rmeta version 2.14 (R Foundation for Statistical Computing, Vienna, Austria; Internet: http://cran-rproject.org/).
Most studies included in our meta-analysis reported correlation coefficients for total carbohydrates, and, occasionally, additional carbohydrate fractions, to assess the ability of their food-frequency questionnaire (FFQ) to rank individuals according to GI and GL. Brunner et al (44) suggests that correlation coefficients above "... 0.5 for most nutrients and 0.8 for alcohol between methods is good evidence that the FFQ has the ability to rank individuals ... according to nutrient intake." Because the GI is a characteristic of the carbohydrate in different foods, we hypothesized that the studies with low correlation coefficients for total carbohydrate and/or carbohydrate fractions would most likely be unable to rank individuals according to GI and/or GL. As a consequence, we conducted additional analyses, which included only those studies with correlation coefficients for total carbohydrate 
0.5, in validation studies that had been conducted in representative samples.
P values <0.05 were considered to indicate statistical significance. RR and 95% CIs are shown. We conformed as much as practicable to MOOSE (Meta-analyses of Observational Studies in Epidemiology) guidelines in the report of this meta-analysis (45).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
and Table 2, respectively. Of the 37 prospective cohort studies, 25 were conducted in the United States, 5 in Canada, 2 in Australia, and 5 in European countries. The number of participants ranged from 526 in the study by Chiu et al (40) to 124 907 in the Cancer Society Cancer Prevention Study II reported by Patel et al (11). Most of the participants in the studies were women (90.4%). Of the 37 studies, 7 reported diabetes events, 7 reported breast cancer events, 6 reported colorectal cancer events, 3 reported cardiovascular disease events, 3 reported eye disease events, 4 reported pancreatic cancer events, 3 reported endometrial cancer events, 2 reported gallbladder disease events, 1 reported ovarian cancer events, and 1 reported gastric cancer events. Diagnosis was primarily through the assessment of the participant's medical records (65%) or biochemical (eg, fasting plasma glucose or oral glucose tolerance) tests (11%). However, diagnosis was based on self-report in 5 of the studies and on physical examination in 4. The participants ranged in age from 24 to 76 y, and their average BMI (kg/m2) ranged from 23.5 to 29.0. The median dietary GI of the groups was 54 (glucose = 100; range: 50–61), and the median dietary GL was 112 units (range: 80–140 units). The median GI value for the highest quantile was 58 and for the lowest quantile was 49. The median GL value in the highest quantile was 142 and in the lowest quantile was 92. FFQs were used in all studies, except for in the study by van Dam et al (16), who used the diet history method for assessing typical food and nutrient intakes. Correlation coefficients for total carbohydrates between the FFQs and weighed food records ranged from a low of 0.45 in the study by Stevens et al (13) to a high of 0.78 in the studies by Hodge et al (8) and Giles et al (21). However, the validation studies of Hodge et al (8), Giles et al (21), and Stevens et al (13) were not conducted in a population representative of their cohorts (46, 47), and for the purposes of this meta-analysis, were not considered to be valid estimates. No study directly assessed the ability of their FFQ to measure GI or GL. The greater proportion (46%) of studies only assessed food and nutrient intakes at baseline, although a reasonable proportion assessed intakes twice (20%) and 5 times (16%) and the remainder from 3 (10%) to 4 (8%) times throughout the duration of the study. The duration of follow-up ranged from 4 to 20 y.
|
|
Listed in Table 3 are the RRs and 95% CIs for the fully adjusted random-effects models that investigated the association between GI and GL and the risk of developing type 2 diabetes, heart disease, stroke, breast cancer, colorectal cancer, pancreatic cancer, endometrial cancer, gastric cancer, gallbladder disease, eye disease, and all diseases combined. When all of the studies were analyzed, there were significant positive associations between GI or GL and RR for type 2 diabetes (GI RR = 1.20, 95% CI: 1. 04, 1.38; GL RR = 1.16, 95% CI: 1.04, 1.28), heart disease (GI RR = 1.25; 95% CI: 1.00, 1.56), colorectal cancer (GI RR = 1.10; 95% CI: 1.00, 1.21), endometrial cancer (GL RR = 1.40; 95% CI: 1.05, 1.87), gallbladder disease (GI RR = 1.26, 95% CI: 1.13, 1.40; GL RR = 1.42, 95% CI: 1.25, 1.60), and all diseases combined (GI RR = 1.12, 95% CI: 1.08, 1.17; GL RR = 1.09, 95% CI: 1.05, 1.14). In an analysis stratified by the 2 major chronic disease groups, cancer and cardiovascular disease, there were significant positive associations between GI and risk of all cancers (RR = 1.08; 95% CI: 1.02, 1.14) and between GL and risk of all cardiovascular diseases (RR = 1.41; 95% CI: 1.18, 1.69).
|
0.5 in a group representative of the cohort; Table 4), there were significant positive associations between GI or GL and risk of type 2 diabetes (GI RR = 1.40, 95% CI: 1. 23, 1.59; GL RR = 1.27, 95% CI: 1.12, 1.45), heart disease (GI RR = 1.25; 95% CI: 1.00, 1.56), gallbladder disease (GI RR = 1.26, 95% CI: 1.13, 1.40; GL RR = 1.41, 95% CI: 1.25, 1.60), breast cancer (GI RR = 1.08, 95% CI: 1.02, 1.16), and for all diseases combined (GI RR = 1.14, 95% CI: 1.09, 1.19; GL RR = 1.09, 95% CI: 1.04, 1.15). There was a nonsignificant positive association between GI and colorectal cancer (RR = 1.11; 95% CI: 0.99, 1.24; P = 0.059). Results are shown in Figure 2 and Figure 3 for GI and GL, respectively. When stratified by the 2 major chronic disease groups and validity of study FFQ, significant positive associations remained between GI and risk of all cancers (GI RR = 1.08; 95% CI: 1.03, 1.14) and GL and risk of all cardiovascular diseases (RR = 1.41; 95% CI: 1.00, 1.97). When the 5 studies that relied on self-diagnosis were removed from the analysis, the RR remained essentially the same for GI (RR = 1.12; 95% CI: 1.05, 1.16; P < 0.0001), but increased slightly for GL (RR = 1.11; 95% CI: 1.05, 1.17; P < 0.0001).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This meta-analysis has notable strengths. There were many large studies with a correspondingly high number of incident cases, which improved the statistical power to detect significant differences. We found no evidence of publication bias on testing, and our sensitivity analysis showed minimal influence on the combined results for any single study: the heterogeneity of variance between studies was allowed for by using random-effects models. A limitation, however, was that 90% of participants were female; therefore, the findings may not be generalizable to men. We made an a priori assumption that validation of carbohydrate intake was an important attribute of more reliable studies on GI and GL. A significant limitation, however, was the fact that no study actually validated the assessment of GI or GL using another dietary method or against an objective standard. The assignment of GI values to foods in a nutrient database is to some extent subjective and may be unreliable when extrapolating from one country to another. It is likely in any case that any misclassification of GI or GL would lead to a bias toward the null hypothesis and diminish the observed effect size. Despite comprehensive measurement and adjustment strategies, the uncontrolled or residual confounding in observational studies of dietary intake is always a concern. Healthy lifestyle effects associated with dietary intake cannot be completely adjusted for in observational studies. Therefore, a meta-analysis of intervention studies looking at "hard" clinical endpoints, not chronic-disease risk factors, may be warranted, when sufficient data have accumulated.
Our findings support the hypothesis that postprandial hyperglycemia, in individuals without diabetes, contributes to chronic disease. Higher glucose concentrations are thought to play a direct pathogenic role in the disease process. The DECODE study, a meta-analysis of 13 studies involving 25 000 individuals, found an almost 2-fold increased risk of all-cause mortality in individuals with an elevated 2-h postchallenge blood glucose (53). Similarly, a meta-analysis of 38 studies involving 194 000 individuals found a progressive relation between hyperglycemia and cardiovascular disease risk (54). Cancer risk is also elevated in those with preexisting hyperglycemia. Larsson et al (55) reported a 30% increase in risk of colorectal cancer in a meta-analyses of persons with type 2 diabetes; similarly, Huxley et al (56) found an 82% increase in risk of pancreatic cancer.
There are plausible mechanisms linking the development of certain chronic diseases with high-GI diets. Specifically, 2 major pathways have been proposed to explain the association with type 2 diabetes risk (57). First, the same amount of carbohydrate from high-GI foods produces higher blood glucose concentrations and a greater demand for insulin. The chronically increased insulin demand may eventually result in pancreatic β cell failure, and, as a consequence, impaired glucose tolerance. Second, there is evidence that high-GI diets may directly increase insulin resistance through their effect on glycemia, free fatty acids, and counter-regulatory hormone secretion. High glucose and insulin concentrations are associated with increased risk profiles for cardiovascular disease, including decreased concentrations of HDL cholesterol, increased glycosylated proteins, oxidative status, hemostatic variables, and poor endothelial function (58). The mitogenic action of insulin-like growth factors suggests a role in the etiology of various cancers. Insulin itself stimulates a rise in free insulin-like growth factors, which are necessary for the cell to progress from the G1 to the S phase of the cell cycle (58).
The reason that low-GI diets may offer greater protection than low-GL diets may be due to the fact that low-GL diets are more heterogeneous and can include either low-GI, high-carbohydrate foods or low-carbohydrate foods (eg, meat and cheese) (59). Although both diets will reduce postprandial glycemia, it is likely that the 2 dietary patterns will have very different metabolic effects, including differences in β cell function (60), triacylglycerol concentrations (60), free fatty acid concentrations (61), and effects on satiety (62). These factors would, in turn, affect the risk of developing chronic disease in genetically susceptible individuals.
This meta-analysis provides high-level evidence that diets with a high GI, high GL, or both, independently of known confounders, including fiber intake, increase the risk of chronic lifestyle-related diseases. The effect was modest overall (GI RR = 1.14; GL RR = 1.09) but more pronounced for type 2 diabetes (GI RR = 1.40; GL RR = 1.27), heart disease (GI RR = 1.25), and gallbladder disease (GI RR = 1.26; GL RR = 1.41). Overall, the GI had a more powerful effect than did the GL (the product of carbohydrate and GI), with more positive associations between GI and chronic disease risk, and associations of greater magnitude, which suggests that, irrespective of the level of carbohydrate intake, the GI of contributing carbohydrate foods is important. The findings indicate that the judicious choice of low-GI foods offers a similar or higher level of protection as whole-grain foods or high fiber intake in the prevention of chronic lifestyle-related disease. The observation that a subgroup (the lowest quantile) of people in developed countries self-selects a low-GI or a low-GL diet is evidence that such diets can be sustained over the longer term. Because most participants were women, there is a need to confirm the findings in men.
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONSUBJECTS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. Galeone, C. Pelucchi, L. D. Maso, E. Negri, R. Talamini, M. Montella, V. Ramazzotti, R. Bellocco, S. Franceschi, and C. La Vecchia Glycemic index, glycemic load and renal cell carcinoma risk Ann. Onc., November 1, 2009; 20(11): 1881 - 1885. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Bao, V. de Jong, F. Atkinson, P. Petocz, and J. C Brand-Miller Food insulin index: physiologic basis for predicting insulin demand evoked by composite meals Am. J. Clinical Nutrition, October 1, 2009; 90(4): 986 - 992. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Lagiou, M. Rossi, A. Tzonou, C. Georgila, D. Trichopoulos, and C. La Vecchia Glycemic load in relation to hepatocellular carcinoma among patients with chronic hepatitis infection Ann. Onc., October 1, 2009; 20(10): 1741 - 1745. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Rossi, L. Lipworth, L. D. Maso, R. Talamini, M. Montella, J. Polesel, J. K. McLaughlin, M. Parpinel, S. Franceschi, P. Lagiou, et al. Dietary glycemic load and hepatocellular carcinoma with or without chronic hepatitis infection Ann. Onc., October 1, 2009; 20(10): 1736 - 1740. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. E Buyken, A. L. Gunther, A. Barclay, J. Brand-Miller, and M. B Schulze Glycemic index in overweight development: distinguishing limited evidence from limits in evidence Am. J. Clinical Nutrition, July 1, 2009; 90(1): 243 - 244. [Full Text] [PDF]
|
|
|
|
|
|
M. A Mendez and H. Schroder Reply to AE Buyken et al Am. J. Clinical Nutrition, July 1, 2009; 90(1): 244 - 246. [Full Text] [PDF]
|
|
|
|
 |
|
 P. Hujoel Dietary Carbohydrates and Dental-Systemic Diseases Journal of Dental Research, June 1, 2009; 88(6): 490 - 502. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. C. N. Chan, V. Malik, W. Jia, T. Kadowaki, C. S. Yajnik, K.-H. Yoon, and F. B. Hu Diabetes in Asia: Epidemiology, Risk Factors, and Pathophysiology JAMA, May 27, 2009; 301(20): 2129 - 2140. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Hlebowicz, A. Hlebowicz, S. Lindstedt, O. Bjorgell, P. Hoglund, J. J Holst, G. Darwiche, and L.-O. Almer Effects of 1 and 3 g cinnamon on gastric emptying, satiety, and postprandial blood glucose, insulin, glucose-dependent insulinotropic polypeptide, glucagon-like peptide 1, and ghrelin concentrations in healthy subjects Am. J. Clinical Nutrition, March 1, 2009; 89(3): 815 - 821. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. M. E. van Bakel, N. Slimani, E. J. M. Feskens, H. Du, J. W. J. Beulens, Y. T. van der Schouw, F. Brighenti, J. Halkjaer, A. E. Cust, P. Ferrari, et al. Methodological Challenges in the Application of the Glycemic Index in Epidemiological Studies Using Data from the European Prospective Investigation into Cancer and Nutrition J. Nutr., March 1, 2009; 139(3): 568 - 575. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. G Mulholland, L. J Murray, C. R Cardwell, and M. M Cantwell Glycemic index, glycemic load, and risk of digestive tract neoplasms: a systematic review and meta-analysis Am. J. Clinical Nutrition, February 1, 2009; 89(2): 568 - 576. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Puntoni, B. Bonanni, and A. Decensi Dietary Changes After Breast Cancer in Women Without Hot Flashes: A Simple and Inexpensive Way to Target Tumor and Host? J. Clin. Oncol., January 20, 2009; 27(3): 323 - 325. [Full Text] [PDF]
|
|
|
|
|
|
M. A Mendez, M. I. Covas, J. Marrugat, J. Vila, H. Schroder, and on behalf of the REGICOR and HERMES investigators Glycemic load, glycemic index, and body mass index in Spanish adults Am. J. Clinical Nutrition, January 1, 2009; 89(1): 316 - 322. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Livesey and H. Tagami Interventions to lower the glycemic response to carbohydrate foods with a low-viscosity fiber (resistant maltodextrin): meta-analysis of randomized controlled trials Am. J. Clinical Nutrition, January 1, 2009; 89(1): 114 - 125. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. W.C. Kendall, A. Esfahani, A. J. Hoffman, A. Evans, L. M. Sanders, A. R. Josse, E. Vidgen, and S. M. Potter Effect of Novel Maize-based Dietary Fibers on Postprandial Glycemia and Insulinemia J. Am. Coll. Nutr., December 1, 2008; 27(6): 711 - 718. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Livesey and R. Taylor Fructose consumption and consequences for glycation, plasma triacylglycerol, and body weight: meta-analyses and meta-regression models of intervention studies Am. J. Clinical Nutrition, November 1, 2008; 88(5): 1419 - 1437. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. G Mulholland, L. J Murray, and M. M Cantwell Glycemic index, glycemic load, and chronic disease risk Am. J. Clinical Nutrition, August 1, 2008; 88(2): 475 - 476. [Full Text] [PDF]
|
|
|
|
|
|
A. W Barclay, V. M Flood, J. C Brand-Miller, and T. Prvan Reply to HG Mulholland et al Am. J. Clinical Nutrition, August 1, 2008; 88(2): 476 - 477. [Full Text] [PDF]
|
|
|
|
|
|
T.-P. Tuomainen, J. Mursu, and S. Voutilainen Bold conclusions from inadequate evidence Am. J. Clinical Nutrition, August 1, 2008; 88(2): 477 - 478. [Full Text] [PDF]
|
|
|
|
|
|
A. W Barclay and J. C Brand-Miller Reply to T-P Tuomainen et al Am. J. Clinical Nutrition, August 1, 2008; 88(2): 478 - 479. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
