| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Division of Preventive and Behavioral Medicine (Y.M., D.C., B.C.O., P.A.M.), University of Massachusetts Medical School, Worcester
Division of Cardiovascular Medicine (I.S.O.), University of Massachusetts Medical School, Worcester
Department of Health and Clinical Sciences, University of Massachusetts, Lowell (R.N.), Massachusetts
Address correspondence to: Yunsheng Ma, M.D., Ph.D., Division of Preventive and Behavioral Medicine, Department of Medicine, University of Massachusetts Medical School, 55 Lake Avenue North, Worcester, MA 01655. E-mail: Yunsheng.Ma{at}umassmed.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
Objectives: To compare the effects of a water-washed soy protein concentrate with a milk-protein based control on blood lipid levels in hyperlipidemic men and women.
Methods: A randomized, double-blind, controlled clinical trial including 159 subjects. After a 3-week run-in period during which all subjects consumed a milk protein-based supplement, participants were randomized into one of two groups: a control group (continued milk protein) and an intervention group (soy protein) for a five-week period. Fasting venous blood draws for lipid measurement were obtained at baseline, towards the end of the run-in period and at the end of the intervention. Blood isoflavone concentrations were measured at the end of the study.
Results: Blood lipid levels were not significantly different between groups at any point in time; and there were no significant associations between blood isoflavones and lipid levels. Significant decreases in total cholesterol (19 mg/dL), and LDL-cholesterol (11 mg/dL), were observed during the run-in period, with no further decreases in lipids during the intervention period in either group.
Conclusions: These results do not support the hypothesis that water-washed soy protein has an effect on blood lipids. Several hypotheses are discussed, highlighting the selective nature of the effect of soy consumption in the population. The cholesterol-lowering effect during the run-in period may be explained by the "regression to the mean effect" and by other factors related to study participation, mainly nutrient displacement induced by the protein supplement.
Key words: soy protein, isoflavones, LDL cholesterol, diet
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
Since that time, several studies have re-examined the effect of soy protein and reported smaller effects of 2–7% or no effect at all [5,7,8]. Studies from Gardner and colleagues [9] and Crouse and colleagues [10] demonstrated that the potential benefits of ingesting soy protein may be attributable to its isoflavone content. This observation is supported by the fact that no effect on lipid levels was observed when alcohol-washed soy protein supplement was used, since the isoflavone content is depleted through ethanol-based extraction, whereas the hypocholesterolemic effect was observed when water-washed soy protein was used, which preserves the isoflavone content. However, there are conflicting reports in the literature regarding the importance of soy proteins on serum lipid levels [5,8,11]. We conducted a randomized, double-blind clinical trial to investigate the effects of a water-washed soy protein concentrate as compared to a milk-protein based control on serum lipid levels in moderately hyperlipidemic men and women. We also measured blood isoflavone concentrations at the end of the study to correlate with blood lipid changes during the soy treatment.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
Of the 843 individuals who were contacted, 182 were found to be eligible and consented to participate, 22 dropped out (six subjects dropped out before they were randomized, while eight subjects dropped out from each of the intervention groups), leaving 160 eligible subjects who completed the study. Of the 160 eligible subjects who completed the study, we excluded one subject (from the intervention group) because his triacylglycerol concentrations were greater than 400 mg/dL for all five measures, thus interfering with the LDL-C calculation, as described below. Therefore, 159 were included in the final analysis. All subjects completed the five lipid measures.
Design
After recruitment, subjects were provided with a milk-protein supplement for a three week run-in period, and were advised that their overall usual dietary patterns should not change, and they should try to modify their protein and carbohydrate intake accordingly. At the end of the 3rd week, subjects were randomized into one of two groups: a control group (milk-protein) and an intervention group (soy protein). The intervention period lasted five weeks. All subjects were blinded to their group assignments, as were the research assistants who met with the study participants.
Protein Supplements
Dietary supplements containing a mixture of protein, carbohydrate and calcium in powder form, were provided by Central Soya Co., Inc. in sealed packets in two flavors: chocolate and vanilla. Subjects could choose their preferred flavor, and were instructed to consume two packets daily in a beverage form after mixing them with water: one in the morning with breakfast and another one with dinner. The nutrient contents of two supplements are included in Table 1. The soy protein concentrate contained 31.5 g of protein and 120 mg isoflavones in the aglycone form. Calcium was not matched between two supplements.
|
Blood Lipid and Isoflavone Assessment
Fasting (>12 hours) venous blood samples were obtained after sitting for 15 minutes. Blood serum was harvested by low-speed centrifugation at 4°C, aliquoted into individual tubes, and quickly frozen to –70°C. On a regular basis, serum samples were packed in dry ice and shipped for analysis via overnight service to the Centers for Disease Control-standardized laboratory at the University of Massachusetts in Lowell, MA. Total cholesterol (TC), HDL cholesterol (HDL) and triacylglycerol (TG) levels were analyzed using the Abbott VP Autoanalyzer and Sigma reagents. HDL was measured in the supernatant after magnesium-phosphotungstate precipitation of apoB-containing lipoproteins. All assays met the standardization criteria of the CDC-NHLBI Lipid Standardization Program [14]. LDL cholesterol (LDL) was calculated according to the Friedewald formula: LDL = TC-HDL-TG/5 if triacylglycerol concentrations were less than 400 mg/dL [15]. All interassay coefficients of variation for total cholesterol, HDL, and TG were 
2.8%. Serum isoflavones and its metabolites, including dihydrodaidzein, daidzein, glycitein, equol, genistein and ortho-desmethylangolensin (o-DMA), were extracted and analyzed according to Coward and colleagues [16] with electrochemical detector array detection. Using 0.5 mL of blood sample, the lowest detectable level for each isoflavone was 5 ng/ml. Interassay coefficients of variation for isoflavone were <4.9%.
Statistical Analysis
Characteristics between subjects who completed the study and subjects who dropped out were compared using a two-sided t-test for continuous variables and Fisher’s exact test for categorical variables. Similarly, comparisons were made of subjects’ characteristics between the milk-protein group and the soy-protein group at baseline.
Distributions of blood lipids (including TC, LDL, HDL, and triacylglycerols) were examined. Because the histogram of triacylglycerols was skewed, we used natural logarithm transformation of the triacylglycerols data in order to use normality assumptions. Mean lipid values, including TC, LDL, HDL, and triacylglycerols, by visit and study group, were determined using SAS PROC MIXED [17]. The dependent variable was the lipid measure at each point in time. The independent variables were: time of lipid measurement, treatment group, and an interaction term between time and treatment group as fixed effects, with subject treated as random effect. Body weight, dietary intake, and physical activity by visit and study group were estimated in a similar manner.
For subjects in the intervention group, association between blood isoflavones and blood lipid change was examined. Descriptive statistics were computed using uncensored isoflavones (values indicating <5.0 were assumed missing and not included) and change in lipids during treatment period. Lipid difference was computed between two time points (week 8–week 3). The histograms of isoflavone levels indicated skewed distributions. Therefore natural logarithm transformation of the isoflavones data were used in order to meet appropriate normality assumptions for the regression analyses. The distributions of natural log values of isoflavones were relatively symmetric after the transformation. Each of the blood isoflavones was modeled as a function of change in blood lipid. Isoflavones were used as the dependent variable as a matter of ‘convenience’ since there are methods available for regression analysis of censored dependent variables. Less is known about censored values as independent variables. Tobit analysis [18–20] was carried out for each isoflavone where the lower limit of detection was 5 ng/ml (ppb). Tobit analysis is appropriate where the dependent variable has a fixed censoring point for all subjects. We also conducted linear regression analyses using only uncensored isoflavone values.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
Participants were predominantly white, married, educated and employed. The average age was 56 years old (SD = 8.46), and 44% of the subjects were male. Characteristics of the subjects in the soy-protein group and in the milk-protein group are compared in Table 2. The two groups resembled each other in most baseline characteristics, except that subjects in the milk-protein group were more likely to be employed full time as compared with subjects in the soy group (p = 0.002). We noted that distributions of BMI and gender between the two randomized groups were at p = 0.08, although it did not reach statistical significance.
|
Data regarding body weight and dietary intake over time by study group are presented in Table 3. There was a slight, non-significant increase in body weight in both groups throughout the study. Body weights were not significantly different between the milk-protein and soy-protein groups at any point in time. There were significant increases in intake of calories, percent of total calories from protein and calcium in both groups throughout the study, whereas the percentage of total calories from fat and from saturated fat decreased significantly among participants in both groups. Most of the decrease was observed during the run-in period. Percentage of total calories from carbohydrate, as well as intake of dietary cholesterol and fiber remained stable. Physical activity levels, measured in minutes of leisure time physical activity per week, dropped slightly for both groups during the eight week study period. However, no statistically significant difference was found between the two groups in physical activity at any point in time (data not shown).
|
Fig. 1 presents lipid values over time. There were significant differences among the five time points for TC, LDL cholesterol, and HDL within each group. We found that TC, LDL, and HDL dropped significantly during the first 2 weeks of the run-in period. For example, the decrease in total cholesterol levels was 19.2 mg/dL (SD = 30.0), and 11 mg/dL (SD = 26.4) for LDL cholesterol. However, during the intervention period these values seemed to stabilize, both within and between groups. There was no significant change in triacylglycerol levels over time. No differences in lipid values were statistically significant between the two groups at any time points.
|
or < median) value for each lipid distribution. The pattern observed in the subjects in the upper half of the total cholesterol (264 mg/dL) and LDL cholesterol (179 mg/dL) distribution at baseline was similar to that observed for the whole group. However, the magnitude of change was larger. For example, the TC levels decreased by 25.2 mg/dL (SD = 33.2) and LDL cholesterol by 19.8 mg/dL (SD = 25.9) for subjects in the upper half of the LDL cholesterol distribution during the first 2 weeks of the run-in period. These subjects also had greater changes in dietary fat and saturated fat intake during the run-in period. For example, percentage of total calories from fat and from saturated fat decreased by 5.1% and 1.5%, respectively, for subjects in the upper half of the LDL cholesterol distribution, while the decreases were 4.3% and 1.3%, respectively, for subjects in the lower half the distribution. Of 81 subjects in the intervention, 80 subjects had available blood isoflavone data. Table 4 presents blood isoflavones and lipid changes. There were a large number of undetectable values for dihydrodaidzein, glycitein, and equol. Equol producers, as defined by blood equol levels above 20 ng/ml, were 26.3%, similar to data published in the literature [21]. Isoflavone values exhibited a large range, for example, genistein levels ranged from 7.90 to 845.5 ng/ml.
|
|
|
DISCUSSION |
|---|
The possible biological mechanisms of the effect of soy on blood lipid level may be associated with several of its components, including isoflavones, trypsin inhibitors, phytic acid, saponins, fiber, and small peptide fractions [2,22,23]. The hypocholesterolemic effects of these components have been under investigation since the release of the FDA recommendation related to a beneficial effect of soy on the lipid profile, with isoflavones receiving considerable attention. Isoflavones are found in many foods, such as soybeans, tofu, tempeh, and soy flours, and in concentrates prepared by a water extraction process. Isoflavones have a common diphenolic structure that resembles the potent synthetic estrogen diethylstilbesterol and hexestrol [24]. Just as estrogens lower LDL cholesterol levels and increase HDL levels, it has been proposed that isoflavones may have a similar effect [25]. Studies by Crouse and colleagues [10] and Gardner and colleagues [9] suggest that the cholesterol-lowering effect of soy protein is predominantly related to isoflavones. Several other investigations have not supported this finding [26–31], and more recently a study by Kreijkamp-Kaspers and colleagues found that use of soy protein supplement containing isoflavones did not lower blood lipids in healthy postmenopausal women [8]. Three recent meta-analyses have discussed this issue [32–34] and two concluded that isoflavones do not appear to have a lipid lowering effect [32,33]. It should be noted, however, that isoflavone content is depleted in alcohol-washed soy protein, which was used by the study of Sirtori and colleagues [29].
We were unable to demonstrate a decrease in blood lipids after consumption of soy protein with isoflavones, when compared to a milk protein supplementation. Several factors could account for this lack of association, including: dose of isoflavones, timing between dose and blood draw, intestinal absorption of isoflavones, length of treatment, method of processing of soy protein, background diet, population issues, and baseline cholesterol levels; all of these factors contributed to the existence of a selective pattern of response to phytoestrogens.
Daily isoflavone intake in our study from soy was 120 mg, a dose which is higher than what is used in most studies, for example Gardner’s study, which showed an inverse relationship between isoflavone intake and blood lipids, used 80 mg of isoflavones per day [9].
Another potential explanation for the lack of effect on blood lipids could be differential absorption of isoflavones. Several authors have highlighted the mechanism through which phytoestrogens found in soybeans are absorbed and become bioactive [21, 35–37]. Equol, an important bioactive plant-derived nonsteroidal estrogen considered a potent antioxidant, is not a phytoestrogen per se, since it is not present in plants, but rather is a byproduct of the bio-transformation of diadzein by the microflora in the large intestine [21]. However, equol is not produced in all people in response to dietary stimulus with soy. Setchell and colleagues have suggested that the population could be classified on the basis of their equol producing capacity, as "equol producers" and "non-equol producers". Several studies have suggested that the proportion of "equol producers" in the general population is variable, from 14–70% [21, 35, 38–41], which could account for the relatively large proportion of studies looking at the effects of soy on blood lipids reporting inconsistent results. There may be some people who eat a diet higher in carbohydrate and fiber, and lower in fat, who are equol producers. These people have a higher concentration of bacteria in the colon [21,35,41]. The intestinal floras are also adversely affected by antibiotic use. In our study the proportion of equol producers was 26.3%, which could justify the lack of response in blood lipids.
Another possible explanation for the lack of an association may be the length of the intervention. The soy protein treatment period in our study was five weeks, while in the study by Gardner and colleagues [9] it was 12 weeks. When we plotted the response in LDL-C change over time from both studies, we found that the decrease in LDL-C in the study by Gardner started at around week 7, suggesting that our intervention may not have lasted long enough to be able to observe the effect. However, in support of our study results, a recent report by Kreijkamp-Kaspers and colleagues [8], with a one-year intervention, also showed no effect on blood lipids. Recent meta-analysis formed by Zhan and colleagues [34] showed a significant effect of soy protein containing isoflavones. Most studies included in this meta-analysis have longer treatment duration.
The method of processing soy protein may also affect the response in blood lipid levels. The hypothesis is that when the alcohol-washed method was used, the isoflavone content is depleted through ethanol-based extraction, whereas the water-washed method preserves the isoflavone content. Of the studies included in the meta-analysis by Anderson et al, [3] on which the FDA based its recommendation, five out of the twenty-two positive studies (95% CI not including zero) [42–47] used textured vegetable protein (TVP) as the substrate. Although in our study we used water-washed soy protein, which retains the isoflavone content, the processing of the soy protein to a powder form may alter the proportion of the components and absorption rate in the source food, and end up not exerting its effect: i.e., the administration of the individual components added together may not equal the effect of the administration of the food as a whole.
The intricate phenomenon of nutrient displacement that occurs in dietary intervention studies has been cited often, yet it still is poorly understood, given the complexity of the interactions. In our study, the protein supplements added approximately 250 kcal per day to subjects’ diet during the study period. However, the total caloric intake increased only by 200 kcal during the run-in period and gradually increased thereafter. In examining the macronutrient content, we found that there were significant increases in the intake of protein (mostly related to the supplement itself) and significant decreases in the intake of fat and saturated fat, whereas carbohydrate intake remained relatively constant, as a percentage of total caloric intake. This observation reinforces the idea of nutrient displacement, as opposed to a simple addition of the nutrients in the supplements to those of the regular diet. However, depending on the background diet, the specific nutrient displacement, and its effect, may be significantly different.
Again, taking a closer look at the meta-analysis by Anderson and colleagues [3], we find that all of the studies with positive findings (5/22) [42–47] were performed in Italy (except one, which was a multi-center study including Italian and Swiss sites) [43]. This leads to an interesting question: Are there certain characteristics of the Italian diet that may make that population more susceptible to the effects of TVP supplementation e.g., higher carbohydrate and fiber content which increases the micro flora in the colon and therefore enhances the capacity to bio-transform daidzein to equol, as opposed to the characteristic high fat diet in the US population?
Finally, Anderson and colleagues [3] also note in their meta-analysis that baseline cholesterol level was a significant predictor of the change in cholesterol in response to soy supplementation. In the six populations (reported in 5 studies) [42–47] where a significant LDL lowering effect, related to increased soy intake was observed; all of these studies were performed on subjects with type II familial hyperlipoproteinemia, with average baseline cholesterol levels over 300 mg/dL. Our subjects had averaged initial cholesterol levels of 270 mg/dL (range 240–350), still quite elevated, but may not have been due to a genetic predisposition to elevated cholesterol, thus incurring a potentially different response to treatment.
In our study, a significant decrease in cholesterol levels during the run-in period of 19 mg/dL in TC and 11 mg/dL in LDL was found. Several factors could account for this observation: the effect from "regression to the mean", the effect of nutrient displacement induced by the protein supplement, and the intrinsic effects of the increase in dietary protein.
The first possibility is that the decrease in blood lipid levels could be explained by the effect of regression to the mean. There may be an effect of regression to the mean because we recruited participants on the basis of being above a cut-point at the time of their screening eligibility assessment. In order to compare the magnitude of the regression to the mean with other studies, we examined available data from another study with a sample of 774 hypercholesterolemic people from the same geographical area and with similar baseline characteristics. No dietary supplements were provided to this population between the two measures of blood lipids within a one week period, and the first measure also was used to screen hyperlipidemic subjects for recruitment. An average decrease of 6.5 mg/dL in total cholesterol and 6.6 mg/dL in LDL cholesterol was observed (data not published). Therefore, if the comparison is appropriate, it appears that effect of regression to the mean could explain 30 to 50% of the observed decrease between the first and the second measurements.
Studies have shown that a decreased intake of fats and saturated fats has a favorable effects on blood lipids [48–50]. However, the decrease in consumption of fat observed in our study was not absolute, but rather a decrease in the percentage of calories as fat (overall from 
38 to 32% of total calories—Table 2), while the actual amount of fat in grams consumed remained unchanged. Therefore, we would not expect an effect on blood lipid levels.
At the same time, an increase in protein intake may be responsible for the decrease in serum lipids observed in our study, through an intrinsic effect of protein on lipid metabolism as suggested by Gardner and colleagues [9]. The Gardner study reported an unexpected decrease in LDL cholesterol observed in the milk group, and no significant difference in LDL cholesterol between the soy-protein group and the milk group. This observation is consistent with our findings in terms of the effect we noticed in the milk-protein group and the significant decrease in blood lipids during the run-in phase. Gardner and colleagues [9] attribute this observation to "other" factors associated with participation in the study. Although study participation could explain some of the changes observed during our study, the magnitude of the decline in blood lipids is significant and cannot be solely explained by it.
Several other mechanisms have been postulated as responsible for the cholesterol lowering effect of protein itself [51]. Among these are alterations in the glucagon/insulin ratio, (increased glucagon or decreased insulin), which may affect hepatic cholesterol synthesis [51]. Two recent studies [22,52] suggests that if one or more peptides from soy protein can reach the liver after intestinal digestion, they may elicit a cholesterol-lowering effect. The hypothesis that the protein moiety, devoid of isoflavone components, is likely to be responsible for this major biochemical effect of soy protein [22], is supported by the dose-related response (decrease in blood lipids related to increasing amount of soy consumed) reported in the results of the meta-analysis performed by Anderson and colleagues [3]. Other reports suggest that depression of apolipoprotein B-containing lipoproteins via alterations in LDL receptor quantity or activity [26]; increased turnover of VLDL apoprotein B in humans who substitute soy protein for animal protein [53]; and the amino acid composition of foods (increases in arginine) [54] may be responsible for the decrease in blood lipids related to soy consumption.
Strengths of the study includes a larger sample size than many studies previously reported, excellent dietary data (three 7-day recalls at appropriate time points), and high rates of retention and adherence. In addition, the soy protein in our study is greater than what the FDA health claim is based on [1]. Therefore, the lack of significant findings from our study could not be attributed to the dosage.
There are several limitations of our study. First, we acknowledge that the proportion of women in the soy group tended (p = 0.08) to be larger than in the milk group, and it may have affected our results. Second, we were unable to perform gender-specific analysis due to the small sample size. Thirdly, because participants were under "uncontrolled" conditions prior to randomization, the interpretation of any statistical comparisons between baseline and the rest of the study time points should be done with caution. Fourth, we do not have day-to-day measurements to allow us to assess day-to-day variability for the lipid measures. Fifth, we can not isolate the independent effect of isoflavones, as soy protein and isoflavones were both present in the soy group. Finally, we did not match calcium and carbohydrate content in two supplements during the manufacturing of the products. The effect of calcium and carbohydrates on lipids could have influenced our results.
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received January 20, 2005. Accepted May 11, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  L. K Nies, A. A Cymbala, S. L Kasten, D. G Lamprecht, and K. L Olson Complementary and Alternative Therapies for the Management of Dyslipidemia Ann. Pharmacother., November 1, 2006; 40(11): 1984 - 1992. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
