| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Stanford Prevention Research Center and the Department of Medicine, Stanford University Medical School, Stanford
Department of Nutrition, Loma Linda University (M.M.), California
Cancer Research Center of Hawai'I, Honolulu, Hawaii (A.A.F.)
Address correspondence to: Christopher D. Gardner, Ph.D., Hoover Pavilion, N229, 211 Quarry Road, Stanford, CA 94305-5705. E-mail: cgardner{at}stanford.edu
 |
ABSTRACT |
|---|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
Design: Randomized clinical trial, cross-over design.
Subjects: Participants were 30–65 years of age, n = 28, with pre-study LDL-cholesterol (LDL-C) concentrations of 160–220 mg/dL, not on lipid lowering medications, and with an overall Framingham risk score of 
10%.
Intervention: Participants were required to consume sufficient milk to provide 25 g protein/d from each source. The protocol included three 4-week treatment phases, each separated from the next by a wash-out period of 
4 weeks.
Results: Mean LDL-C concentration at the end of each phase (± SD) was 161 ± 20, 161 ± 26 and 170 ± 24 mg/dL for the whole bean soy milk, the soy protein isolate milk, and the dairy milk, respectively (p = 0.9 between soy milks, p = 0.02 for each soy milk vs. dairy milk). No significant differences by type of milk were observed for HDL-cholesterol, triacylglycerols, insulin, or glucose.
Conclusion: A 25 g dose of daily soy protein from soy milk led to a modest 5% lowering of LDL-C relative to dairy milk among adults with elevated LDL-C. The effect did not differ by type of soy milk and neither soy milk significantly affected other lipid variables, insulin or glucose.
Key words: soy protein, soy milk, lipids, LDL-cholesterol, hypercholesterolemic adults
|
|
INTRODUCTION |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
In the last decade advances in soy research have identified several factors that could affect study outcomes in lipid lowering trials. Some evidence suggests the reduction in LDL-C is positively related to baseline LDL-C concentrations [3,8]. Another important factor may be the manner in which the protein is processed. For example, the 
’ subunit of the 7s fraction of one of the major soy storage proteins, beta-conglycinin, has been shown to upregulate hepatic LDL receptors and inhibit atherosclerosis in animal models [9,10]. Different soy protein processing techniques may have variable effects on the availability of this subunit. Another factor may be the isoflavone content of the soy protein although this point remains controversial [11,12]. In addition, some evidence suggests that individuals who possess intestinal bacteria capable of converting the isoflavone daidzein into equol (30–50% of the population) [13–15] experience a larger reduction in LDL-C in response to soy protein than those who do not [15,16]. Notably, over the past 30 years most trials, and especially those conducted in the United States, have used casein as the control protein and soy protein isolate (essentially fat-free and by definition at least 90% protein) as the source of soy protein [3,11,17]. Few clinical trials have evaluated the hypocholesterolemic effects of the more traditional, less processed Asian soy foods [18–22]. Variations in all of these factors may largely explain the inconsistencies in the literature regarding the effect of soy intake on blood lipids.
The objective of the current study was to examine the effect of soy milk intake on blood lipid variables as well as blood glucose and insulin concentrations among adults with particularly high LDL-C concentrations (160–220 mg/dL at eligibility screening) who were not on lipid lowering medications and were otherwise in general good health. Two types of soy milk, a whole soy bean (WB) beverage and a soy protein isolate (SPI) beverage were compared to dairy milk (DM) in order to address the possibility that differences in proteins, isoflavones, or processing would affect the results. The daily dose of protein was 25 g, consistent with the FDA health claim. Finally, the potential impact of equol producing status on study outcomes was assessed.
|
|
SUBJECTS AND METHODS |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
10% based on gender, age, LDL-C, HDL-C, blood pressure and diabetes [23]. Exclusion criteria were cigarette smoking, currently pregnant or lactating, actively on a weight loss program, diabetes, prevalent coronary artery disease or cancer, uncontrolled hypertension, body mass index higher than 35 or lower than 19 kg/m2, and daily average intake of 3 or more alcoholic drinks/day. All participants provided written informed consent and the study was approved annually by the Stanford University Human Subjects Committee.
Study Design
The study design was a randomized three-way cross-over. Prior to randomization the protocol included a 4-week run-in phase which involved three separate group sessions of instructions to follow a heart-healthy diet based on American Heart Association guidelines [24]. At this time participants were also instructed how to avoid soy products.
Participants willing to continue after the run-in phase were randomly assigned to one of three orders of the three study milks: WB-SPI-DM, SPI-DM-WB, or DM-WB-SPI. Randomization was done in blocks of 12 and occurred by having a blinded research technician select folded pieces of paper with order assignments from an opaque envelope such that four participants were evenly assigned to each order for every 12 participants. Participants were unblinded to the order of the milks because the beverages were provided in containers available at local markets. During each 4-week milk phase participants picked up study milks on a weekly basis. A wash-out period of at least 4 weeks occurred after each of the first and the second milk phases.
Blood samples were obtained on 15 days. Immediately prior to starting each of the three milk consumption phases, fasting blood samples were collected on two separate days within a 3-day window. A single non-fasting blood sample was collected during the middle of each phase for analysis of plasma isoflavonoid concentration and to identify equol producers. In the final days of each 4-week milk phase, fasting blood samples were again obtained on two separate days within a 3-day window. All blood samples were collected in EDTA tubes.
An oral glucose tolerance test (OGTT) was performed at four time points: at the end of the run-in phase (i.e., just prior to the start of the first milk phase), and at the end of each of the three 4-week milk phases. A standardized bolus of a 50 g glucose solution was consumed after obtaining a fasting blood sample, followed by additional blood sampling at one and two hours post glucose load.
Milk Products
The WB milk was the vanilla flavored Silk® brand produced by White Wave Foods. The SPI milk was the vanilla flavored 8th Continent® brand produced by a joint venture between DuPont and General Mills and made with Solae® brand soy protein isolate. The DM was an organic, 1% fat brand produced by Horizon Organic. All milk products were purchased from a local merchant on a weekly basis and distributed to participants by the study coordinator. Energy content and macronutrient composition of the milk products were obtained from label information and by direct communication with the manufacturers. Milk products from different lots with different expiration dates were collected, aliquoted in duplicate pairs, and frozen at three different time points during the study for analysis of 7s and 11s protein subunits and isoflavone content.
The endothermic protein composition of all three types of study milk was compared to a standard soy isolate with known peaks for 7s and 11s globulins using Differential Scanning Composition Densitometry. The patterns of protein bands were then identified using polyacrylamide gel electrophoresis (PAGE) and quantified using scanning densitometric techniques. The amount of each protein subunit was measured as a percentage of the total amount of protein (Camden and Chorleywood Food Research Association Group, Gloucestershire, United Kingdom). Isoflavone content of the soymilks was determined for daidzein, genistein, glycitein and their glucosides by liquid chromatography/photo diode array detection [25]. Plasma isoflavonoids, including equol, were determined by liquid chromatography/mass spectrometry [26] as recently revised [27,28].
Adherence and Diet Assessment
Participants kept daily milk consumption logs and submitted them weekly during each of the three 4-week milk consumption phases.
Three day food records were completed at five time points during the study: baseline (prior to starting the run-in phase), end of the 4-week run-in phase, and once in the middle of each of the three milk phases. Nutrient composition was analyzed using Food Processor (version 8.4, ESHA, Salem, OR).
Plasma Lipid, Insulin and Glucose Assessments
Laboratory staff members conducting analyses of plasma samples were blinded to treatment assignment. Plasma total cholesterol and triacylglycerol (with subtraction of a free glycerol blank) were measured by enzymatic procedures using established methods of the Stanford Clinical Chemistry Laboratory (Beckman Synchron LX20 instrument) [29,30]. HDL-cholesterol (HDL-C) was measured by liquid selective detergent followed by enzymatic determination of cholesterol [31]. LDL-C was calculated according to Friedewald [32]. Lipid assays were monitored by the Lipid Standardization Program of the Centers for Disease Control and Prevention and were consistently within specified limits (monthly coefficients of variation were all 3.1%).
The study milks were unmatched for total fat, types of fat, and cholesterol, so established equations were used to project the impact of these differences on LDL-C concentrations. The Hegsted equation (
LDL-C (mg/dL) = 1.74 x saturated fat (%) – 0.766 x polyunsaturated fat (%) + 0.0439 x mg cholesterol/1000 Kcal)[33] and the Mensink and Katan equation (LDL-C (mg/dL) = [1.28 x (% carbohydrate – % saturated fat)] – 0.24 x [% carbohydrate – % monounsaturated fat] – 0.55 x [% carbohydrate – % polyunsaturated fat]), [34] were used for these projections.
Plasma insulin was measured by radioimmunoassay [35], and an area under the curve was determined using the trapezoidal method. Plasma glucose was measured by a modification of the glucose oxidase/peroxidase method described by Trinder [36].
Statistical Analysis
All analyses were conducted using SAS 9.1.3 with Service Pack 3 (SAS Institute Inc., Cary, NC). Descriptive statistics and graphs (PROC UNIVARIATE and PROC MEANS) were used to summarize the characteristics of the study population. When variable distributions violated basic testing assumptions, appropriate transformations of the data were used (e.g., log triacylglycerols). Matched pair t-tests and one way analysis of variance with repeated measures were used to test for differences in nutrient intakes in different phases of the study. For the primary outcome (LDL-C) and secondary outcomes (other lipid variables, insulin and glucose) we conducted one way analysis of variance (ANOVA) with repeated measures using PROC GLM. The model included milk type, sequence assignment, and milk x sequence interaction; subject was used as a repeated measure. When the ANOVA for the main effect of milk type was statistically significant, the contrasts comparing pairs of types of milks were examined. Post hoc analyses were conducted for subsets of study participants who were below vs. above the median LDL-C concentration, and for equol producers vs. non-producers. These potential moderators of effect were added to the model, separately. All statistical tests were two-tailed using alpha = 0.05.
|
|
RESULTS |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
|
Study Milk Composition
Composition of the study milks is presented in Table 2. The volume of daily milk consumed was standardized to yield 25 g protein/d. The daily volume of the WB milk consumed was almost double that for the DM, with the SPI volume being intermediate between these two. The energy content of the daily WB milk was 100 Kcal higher than the DM, while the SPI milk had an energy content relatively similar to the DM. Grams of total fat were higher in both soy milks than the DM, but the amount of saturated fat was highest in the DM. The proportions of 7S and 11S subunits were similar between the two soy milks. The isoflavone content of the WB soy milk was three-fold higher than the SPI milk, and was also considerably more variable from batch to batch.
|
|
|
Applying Equations of Hegsted and of Mensink and Katan
According to the Hegsted equation, the differences in amounts and types of dietary fat and cholesterol among the study milks should have had a projected effect of lowering LDL-C concentrations by 
4.4 mg/dL for WB vs. DM and by 4.2 mg/dL for SPI vs. DM. According to the Mensink and Katan equation these differences should have had a projected effect of lowering LDL-C by 2.9 mg/dL for WB vs. DM and by 2.7 mg/dL for SPI vs. DM milk. In other words, approximately 1/3 - 1/2 of the total effect on LDL-C concentrations was likely due to dietary fat and cholesterol differences between the milks.
Equol vs. Nonequol Producers
Examination of plasma isoflavone concentrations while on the soy milks indicated that 9 of the 28 participants produced levels of equol >50 nM (mean ± sd of 508 ± 350 nM, range 185–1,213 nM). Fourteen participants had undetectable equol levels and five had levels of 5–37 nM which we grouped with the non-equol producers. The differences in LDL-C concentrations for the nine equol producers during the soy milk phases relative to the dairy milk (WB: –10 ± 17 mg/dL; SPI: –7 ± 17 mg/dL) were virtually identical to the full study sample and not significantly different than for the non equol producers (p = 0.8 for equol x milk type interaction).
Insulin and Glucose Outcomes
There were no detectable differences in insulin AUC or glucose at time 0, 1 or 2 hours for any of the three milk phases compared among themselves or compared to the end of run-in samples taken immediately prior to randomization and the start of the intervention; all values were clinically normal (Table 4). There were also no differences in these outcomes by milk type when examined for equol producers vs. non-producers (data not presented).
|
Adverse Events
The only adverse event occurred for one participant that dropped out of the study due to recurrence of a cancer, which was considered to be unrelated to the short-term consumption of study milks.
|
|
DISCUSSION |
|---|
Conclusions from the 1995 meta-analysis on soy protein and lipids [37] and the FDA's 1999 approval of a health claim for soy protein intakes of 25 g/day [4] have been challenged by recent reviews suggesting the magnitude of effect was previously overstated [5,6,38]. The degree of heterogeneity among different trial designs may be at the heart of this controversy. Baseline lipid concentration has emerged as a potentially important moderating factor such that effects of a clinically relevant magnitude may be detectable only among those with total cholesterol or LDL-C concentrations above a cut point (e.g., LDL-C > 160 mg/dL) [3,8,39]. Notably, pharmacotherapy for the treatment of plasma lipids above that level has become wide-spread (e.g. statins). In the current study, screening eligibility required an LDL-C of 160–220 mg/dL and no current or recent use of lipid lowering medications. To address ethical concerns, we required potential participants to meet the additional eligibility criterion of a low overall Framingham risk score. We acknowledge that the resulting study population was a highly select group of individuals. We also note that in this study the modest effect on LDL-C was similar for those above and below the median LDL-C concentration.
Another topic of debate is whether the isoflavones typically found in natural soy protein products have a lipid lowering role [11,12,40]. In the current study the WB soy milk contained approximately three-fold the isoflavone content of the SPI soy milk, with no difference in effect for LDL-C concentrations. Therefore, these results do not support an isoflavone effect at exposure levels above the lower dose SPI level of 39 mg/d.
Dose is another factor that has varied widely from study to study. The soy protein doses used in the trials included in the 1995 meta-analysis ranged from 17–124 g/day with a mean of 47 g/day [3]. Several recent trials have used doses of 50 g/day of soy protein, double the amount used in the FDA health claim, with very modest effects [11,41,42]. Given that an 8-oz serving of soy milk or a 4-oz serving of tofu contain approximately 7–9g protein, reaching even the 25 g/day level of the FDA health claim using commercially available soy products would be an ambitious dietary modification for most people. In the current study, approximately three quarters of participants reported that their daily regimen of either type of soy milk was more than would be acceptable to them on a daily basis long-term. It would likely be easier to reach a daily dose of 25 g protein/day from varied sources rather than a single source of soy (e.g., tofu, edamame, soy milk, and tempeh). However, the design of a trial to test this approach would require a thorough assessment of foods and nutrients displaced by these different types of soy products.
Although the three types of milks used in the trial were all standardized on protein content, they differed in their content of other nutrients. As would be expected, the dairy milk was higher in saturated fat and cholesterol than the soy milks, and the soy milks were higher in polyunsaturated fat than the dairy milk. Many soy protein trials, especially those using soy protein isolate powders, have been designed to match all macronutrient levels in order to focus on the type of protein and to avoid other confounding effects [11]. It was our original intent to test the practical effects of simply substituting commercially available soy milk for dairy milk. Therefore, we chose to allow macronutrients other than protein to be present at their natural levels and thereby increase external validity. Rather than matching the diets for these nutrients, we applied both the Hegsted [33] and the Mensink and Katan equations [34] for projecting effects of dietary fat and cholesterol changes on changes in LDL-C concentrations. These calculations suggest that between one third and one half of the LDL-lowering observed during the soy milk phases might have been attributable to the decreased saturated fat and cholesterol and the increased polyunsaturated fat content of the soy milks. Notably, without the contribution of the differences in fat and cholesterol content of the study milks, the magnitude of the LDL-C differences attributable to the source of protein alone would not have yielded statistically significant differences in this study.
Lovati, Duranti, Sirtori and colleagues have conducted a series of studies suggesting that the ’ subfraction of the 7s globulin of soy protein is mechanistically responsible for effects on LDL-C blood concentrations through the upregulation of LDL-receptors [10,43]. We originally selected one soy milk derived from whole beans and another derived from soy protein isolate with the assumption that these two different manufacturing approaches might lead to differences in the levels of protein subfractions. To the contrary, the amounts of 7s and 11s subfractions were determined to be virtually identical in the two soy milks, which would be consistent with the two soy milks having similar effects on LDL-cholesterol concentrations.
A recent potential advance in the investigation of soy intake and lipid effects has been the suggestion that individuals who generate detectable levels of equol from daidzein during enterohepatic circulation may achieve larger benefits than non equol producers [15,16]. Our data do not support this conclusion. Similar to many other studies of predominantly Caucasian, non-vegetarian populations, approximately one third of our participants were determined to be equol producers [13–15]. However, the LDL-C effect among equol producers in the current study was comparable to the non-equol producers. This conclusion should be interpreted with caution since our study sample included only 28 adults and only nine equol producers.
In summary, among a select population of adults with elevated LDL-C concentrations who were otherwise in good health, the overall effect of the soy intervention was modest with no additional benefit for those who were equol producers, or for those whose LDL-C was above the median. Furthermore, in addition to the direct cholesterol-lowering effects of soy protein, cholesterol reduction was likely also attributable to differences in fatty acid profiles between the soy and dairy milks. The effects of the two soy milks were similar suggesting that neither isoflavone content above 39 mg/d nor certain processing methods affect the hypocholesterolemic effects of soy protein. The overall reduction in cholesterol observed in this study is consistent with the results from recent meta-analyses and reviews indicating soy protein reduces LDL-C by only 3 to 5% [5–7]. Nevertheless, although not relevant at the clinical level, even these modest effects are meaningful at the population level as over a period of years each 1% reduction in LDLC has been estimated to reduce CHD risk from 2 to 4% [44,45]. Given the modest cholesterol-lowering effects of soy protein, future research should focus on the potential cardiovascular benefits of soy foods independent effects of serum lipid levels (e.g., reducing LDL particle size [46] improving endothelium functioning [47,48], inhibiting LDL-C oxidation [49], etc.). Finally, rather than focusing on the more processed soy protein products, future trials would be well-advised to use the less processed soy foods such as edamame, tempeh, soy nuts, and others that are low in saturated fat and cholesterol, and good sources of fiber, vitamins, minerals and unsaturated fats, including the omega-3 fatty acid -linolenic acid. These soy foods are not only healthy food choices, but may also act indirectly on health through the displacement of foods of lower nutritional density and quality [6,11,50].
|
|
ACKNOWLEDGMENTS |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
This investigation was supported by an unrestricted gift from the White Wave Foods company, by NCI/NIH grant CA71789, and by Human Health Service grant M01-RR00070, General Clinical Research Centers, National Center for Research Resources, National Institutes of Health.
|
|
FOOTNOTES |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
Received September 30, 2006. Accepted May 2, 2007.
|
|
REFERENCES |
|---|
|
TOP
FOOTNOTES
ABSTRACT
INTRODUCTION
SUBJECTS AND METHODS
RESULTS
ACKNOWLEDGMENTS
REFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. H. Hennekens, W. R. Schneider, E. J. Barice, and P. R. Hebert Modest Dietary Reductions in Blood Cholesterol Have Important Public Health Benefits Journal of Cardiovascular Pharmacology and Therapeutics, June 1, 2009; 14(2): 85 - 88. [Abstract] [PDF]
|
|
|
|
 |
|
 A. A Thorp, P. R. Howe, T. A Mori, A. M Coates, J. D Buckley, J. Hodgson, J. Mansour, and B. J Meyer Soy food consumption does not lower LDL cholesterol in either equol or nonequol producers Am. J. Clinical Nutrition, August 1, 2008; 88(2): 298 - 304. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
