ABSTRACT Alterations in plasma cholesterol concentrations, especially increases in low-density lipoprotein (LDL), are
well-known risk factors in the development of atherosclerosis. Numerous studies have examined the lipid-lowering effects of
functional soy-containing foods, but few have specifically examined soymilk, with equivocal findings reported. In September
2008, a single-blind, randomized, controlled trial was conducted on 32 postmenopausal women at Baylor University, Waco,
TX, USA. After a 2-week run-in period, subjects were randomly assigned to consume three servings of vanilla soy (n ¼ 16) or
reduced-fat dairy (n ¼ 16) milk per day for 4 weeks. Plasma lipid profiles were obtained pre- and post-supplementation.
Plasma high-density lipoprotein, LDL, and triglycerides were not significantly different between groups post-intervention
(P ¼ .45) or from baseline (P ¼ .83). Separate analysis of plasma total cholesterol levels yielded similar results (P ¼ .19 and
P ¼ .92, respectively). Furthermore, subanalyses controlling for dyslipidemia (n ¼ 23) and lipid-lowering medication usage
(n ¼ 28) did not significantly alter results. Despite good dietary compliance, our study failed to show a significant hy-
pocholesterolemic effect of soymilk consumption in this postmenopausal female population. Potential reasons for this non-
significant finding are discussed, and future research directions are presented.
650
SOYMILK AND LIPID LOWERING 651
Unfortunately, many of the clinical studies addressing the protocol. Exclusionary criteria included active use of hor-
lipid-lowering potential of soy often fail to incorporate mone replacement therapy, presence of coronary artery
commercially available soyfoods, such as soymilk. Because diseases and=or other significant uncontrolled cardiovascu-
it is plausible that synergistic effects may exist among iso- lar, renal, hepatic, gastrointestinal, mental, and endocrine
lated soy constituents, there is considerable need for studies disorders including diabetes mellitus, a body mass index of
examining whole soyfood consumption. To date, only a few <19 kg=m2 or >35 kg=m2, active smoking status (within the
studies have specifically examined the lipid-lowering ben- past 3 years), an average intake of more than two alcoholic
efits of soymilk consumption,11–15 and the results of these drinks per day, and=or the consumption of any dietary
studies have been equivocal. Due to the lack of consensus in supplements that could affect antioxidant status (excluding
the literature, as well as the current reevaluation of the multivitamins) 3 months prior to beginning the study. All
support for the Food and Drug Administration health claim, eligible subjects were required to obtain medical clearance
the purpose of this study was to examine the effects of from their personal physician and sign Baylor University
soymilk consumption on plasma lipid levels in a postmen- (Waco, TX, USA)-approved informed consent documents.
opausal female population.
Study design
MATERIALS AND METHODS This study was conducted as a 4-week single-blind, ran-
Primary and secondary outcomes domized, controlled trial. Prior to randomization, the pro-
tocol included a 2-week run-in phase in which participants
This analysis was a part of a trial originally designed to met with a registered dietitian and were educated and in-
examine the ability of soymilk to attenuate exercise induced structed to limit the amount of dairy products consumed
inflammation and oxidative stress. The primary outcome of (avoid dairy milk completely and limit dairy servings to two
this substudy was to examine the lipid-mediating properties per day) and to avoid isoflavone-containing soy products.
of soymilk. We hypothesized that participants randomized This was done in attempt to minimize baseline dietary dif-
to consume three servings of soymilk per day for 4 weeks ferences between groups with regard to dietary treatments.
would have statistically greater reductions in traditional At this time, participants also provided a 24-hour baseline
markers of cholesterol than the control (dairy) group. Spe- dietary recall to the registered dietitian so that baseline
cifically, we examined plasma levels of TG, TC, LDL, and protein intake could be assessed using the Food Processor
HDL in the entire group of participants post-intervention. Dietary Assessment Software program (ESHA Research
Secondary outcomes included the change in the aforemen- Inc., Salem, OR, USA).
tioned cholesterol values over the supplementation period After the run-in phase, participants were then scheduled
and two subanalyses of the dataset controlling for select to report to the Exercise & Sport Nutrition Laboratory at
confounders. Because of concerns that lipid-lowering Baylor University for baseline testing. This testing session
medications may affect study results, we performed a sub- included personal and medical questionnaires, height and
analysis omitting study participants who reported current weight measurements, blood sampling, dietary intake as-
use of any prescription lipid-lowering or lipid-controlling sessment, and aerobic fitness (maximal oxygen uptake
medication. Additionally, subanalysis of the individuals [VO2 max]) testing by performing a cardiopulmonary graded
with dyslipidemia, as defined by the National Cholesterol exercise test on a treadmill ergometer (Quinton, Inc., Bo-
Education Program Adult Treatment Panel III criteria16 in at thell, WA, USA). After the testing session, participants were
least one of the lipid markers measured was also conducted matched based on baseline protein intake and cardiopul-
to determine if baseline cholesterol status impacted treat- monary fitness level, as both have been shown to affect
ment efficacy. circulating blood lipids,7,17,18 and were randomly assigned
to consume three servings of either soymilk or dairy milk
Participants per day for 28 days. The milk was provided to the partici-
pants in black bags labeled ‘‘A’’ or ‘‘B’’ in such a way that
Thirty-two apparently healthy, recreationally active,
investigators were blinded to treatment order. Participants
postmenopausal women between the ages of 40 and 60 years
were instructed to return to the Exercise & Sport Nutrition
were used as participants in the study. A power calculation
Laboratory 2 weeks into the supplementation period, to pick
revealed that 16 participants per group were necessary to
up additional milk, and again at 4 weeks, for follow-up
detect a significant difference between groups in markers of
testing including the same battery of tests performed at base-
lipid lowering (LDL effect size of 20 mg=dL11) given a
line with the exception of the aerobic fitness test.
Type I error rate of 0.05, SD of 19 mg=dL, and a power of
0.80. Participants were recruited from the Central Texas
Milk products
area and were considered eligible to participate in the study
if they were physically active but not trained (not engaged in Both milk products were provided by WhiteWave Foods,
an exercise program involving either resistance or endur- Inc. (Broomfield, CO, USA). The soy product was the SilkÒ
ance training more than three times per week for 1 year) and brand very vanilla flavored soymilk, and the dairy product
postmenopausal (naturally or surgically amenorrheic >1 was the HorizonÒ brand reduced-fat organic dairy milk. The
year prior to start of study) and could adhere to the study commercial beverages were matched as closely as possible
652 BEAVERS ET AL.
for fluid volume, total caloric, and macronutrient intake. assess the adequacy of our randomization scheme. Study
Nutrient profiles of the milk beverages obtained from the compliance, defined as the percentage of prescribed bever-
nutrition facts panel are shown in Table 1. Because the milk age consumed, was assessed across groups using a two-
products were not matched based upon dietary fat or cho- sample Wilcoxon rank-sum test as these data were
lesterol intake and because slight alterations in macro- and demonstrably non-normal.
micronutrient intake were likely over the 4-week supple- We analyzed the reported dietary intake variables to
mentation period (despite instructions to maintain a similar evaluate the potential confounding effects of certain nutri-
diet to baseline), equations were used to predict changes in tional variables known to interact with plasma lipid levels.
TC19 and LDL.20 The prediction equations used dietary We then compared means across groups at baseline using
intake of saturated and polyunsaturated fat and cholesterol univariate t tests, and, using repeated-measures analyses of
to predict changes in plasma cholesterol values. variance on each variable, we assessed the effect of the
time=intervention on the dietary intake.
Dietary intake Group-specific means were compared for the primary
study outcome for the full analysis and subgroup analyses
To assess dietary intake, each participant was asked to
using a multivariate analysis of variance. The outcome of
keep 4-day dietary records at baseline and during the second
within-individual change from baseline was assessed using a
half of the 4-week supplementation period. The dietary re-
repeated-measures analysis of variance; specifically of in-
cords were then analyzed using the Food Processor Dietary
terest is whether a grouptime interaction was statistically
Assessment Software program. In an attempt to determine
significant. We then repeated these analyses using the uni-
compliance with the milk supplementation protocol, each
variate outcome of plasma TC and the change in plasma TC
participant was instructed to keep a compliance log of daily
from baseline. All numerical summaries and statistical
milk ingestion to be submitted to investigators at the 4-week
comparisons were conducted using SAS version 9.1.3 (SAS
testing session. Authors of similar studies21,22 have sugges-
Institute, Cary, NC, USA), and the Type I error rate (a) was
ted the standard of compliance is between 80% and 100%;
held constant at 0.05 for all tests.
thus, a threshold of <80% consumption of the prescribed
supplement was defined as ‘‘non-compliant’’ a priori. Fi-
RESULTS
nally, a reported side effects questionnaire was administered
at baseline and at 2 and 4 weeks to determine if any sig- Study population
nificant negative side effects occurred during the course of
the study because of the dietary supplementation. Recruitment for this study began September 2008, and
participant testing ended December 2008. A total of 33 in-
Plasma lipid assessment dividuals who met entrance criteria were recruited to begin
the study. Baseline blood work for one participant was un-
Fasting venous blood samples were obtained before and able to be collected because of an early non-intervention-
after the supplementation period from the antecubital vein related dropout, leaving 32 participants retained for
into a 10-mL collection tube using a standard VacutainerÔ analysis. Baseline demographics for the 32 individuals who
apparatus (BD, Franklin Lakes, NJ, USA). Blood samples completed the study are found in Table 2. No statistically
were allowed to stand at room temperature for 10 minutes significant differences were observed in any of these vari-
and then centrifuged. For each sample, the plasma was re- ables between treatment groups. Additionally, no significant
moved and frozen at 808C for later analysis. Plasma HDL, weight changes were observed during the course of the in-
LDL, TG, and TC levels were then assessed by a trained tervention (P ¼ .25).
technician using the Dimension RxL Clinical Chemistry
System (DADE Behring Inc., Deerfield, IL, USA).
Table 2. Baseline Demographic Statistics Including Plasma
Statistical analysis Lipid Values by Treatment Group
Baseline demographic, health, and dietary characteristics Vanilla soymilk Reduced-fat
were summarized as sample mean and SD values. Means group dairy milk
between groups were compared using independent t tests to (n ¼ 16) group (n ¼ 16)
Age (years) 53.88 3.65 54.87 3.05
BMI (kg=m2) 25.36 4.05 26.30 3.82
Table 1. Nutrient Profile of One Serving of the Milk Cardiopulmonary 25.60 4.86 25.57 4.70
Supplements Provided to Study Participants fitness (VO2 max)
Baseline fasting lipid
Vanilla soymilk Reduced-fat
values (mg=dL)
(8.25 fl oz) dairy milk (8 fl oz)
TC 191.50 28.79 208.13 23.56
Calories (kcal) 130 120 TG 103.50 79.35 113.25 82.63
Fat (g) 4 4.5 LDL 106.56 24.22 115.62 20.21
Protein (g) 6 8 HDL 63.06 12.01 66.81 17.21
Carbohydrates (g) 19 12
Data are mean SD values.
SOYMILK AND LIPID LOWERING 653
Table 3. Main Food Components Consumed During the Study by Treatment Group
According to 4-Day Food Record Analysis
Vanilla soymilk group (n ¼ 16) Reduced-fat dairy milk group (n ¼ 16)
300.00
250.00
200.00
mg/dL
50.00
0.00
HDL LDL TG TC
Plasma Lipid Levels
Although multiple studies have reinforced the need for group at baseline, but there were significant changes in
combined isoflavone and soy protein consumption to obtain select dietary variables at 4 weeks. The increases observed
more favorable cardioprotective benefits, perhaps we did not across both groups for kilocalories, protein, fat, and satu-
supply participants with enough of either or both to see a rated fat across time were likely due to the intervention
pharmacologic effect. At present, the Food and Drug Ad- itself as participants in these types of studies often do not
ministraiton health claim states that consumption of 25 g reduce dietary intake to account for supplemental intake.
of soy protein=day is needed to see a reduction in the risk However, while this increase could confound the plasma
of heart disease. In our study (assuming an average of 98% lipid changes from baseline, it would not explain the lack
compliance with the dietary protocol), total soy protein of difference between groups post-intervention. Conversely,
consumed by the soymilk group was roughly 18 g=day. the overall increased amount of saturated fat intake observed
Although more recent studies suggest that as little as 20 g of in the dairy group (particularly post-intervention) as well as
soy protein=day can lower non-HDL cholesterol and have the grouptime interaction for fiber and cholesterol in-
favorable cardiovascular effects,24 the possibility remains take may explain the nonsignificant post-intervention find-
that the amount of soy protein provided in approximately 3 ings with regard to plasma lipids, although this is quite
cups of soymilk per day was not enough to significantly speculative.
improve blood lipid levels. In terms of isoflavone content, Attempts were made to control for dietary fat intake by
recent meta-analyses support the notion that intakes of soy using predictive equations. No changes in TC were pro-
protein coupled with isoflavone above 40 mg=day have jected based upon dietary intake of saturated and polyun-
lipid-lowering effects.6 Although the isoflavone content of saturated fat and cholesterol, but it was predicted that the
the soymilk used was not analyzed, Gardner et al.11 used a soy group would have a decrease in LDL, while the dairy
similar product and showed that 3 cups of soymilk supplies group would observe an increase in LDL. Unfortunately,
approximately 90 mg of isoflavones, although the content these predictive equations do not incorporate other dietary
per sample was noted to be highly variable. Despite this, it is variables associated with alterations in plasma cholesterol,
likely that our study design provided a sufficient amount of such as fiber or select antioxidants. Additionally, error with
isoflavones to potentiate a reduction in cholesterol levels. the use of predictive equations also exists. Similar to our
Lastly, the possibility remains that in spite of reported die- results, a meta-analysis examining the effects of dietary fat
tary compliance, participants may not have adhered to the on serum cholesterol found that the predicted TC and LDL
study protocol, rendering intake insufficient. Unfortunately, levels were significantly lower on a high polyunsaturated
because of funding restraints we were unable to analyze diet relative to a high monounsaturated fat diet; however,
blood or urinary isoflavones and were dependent on par- the observed lipid differences were not significant.25 Fi-
ticipant truthfulness in compliance reporting. We implore nally, it is possible that inaccuracies in using a 4-day dietary
the need for future studies to assess these markers with more record to assess food intake confounded our results. Dietary
objective indicators of compliance as a means of process records are considered accurate for estimating stable dietary
evaluation. components, such as total kilocalories and percentage of
macronutrients, but more variable nutrients such as dietary
Time course of administration cholesterol may require 2–3 weeks of recording to accu-
rately estimate intake.26
Although the amount of time allotted for supplementation
in our study was comparable to11 or only slightly shorter in
duration than12,13 other studies observing favorable out- CONCLUSIONS
comes, our study failed to find a significant result. Results In light of increasing use, it is important to understand the
from the meta-analysis performed by Zhan and Ho6 suggest health effects of soymilk. Although numerous studies have
that the strongest lowering effects of soy protein containing reinforced the need for combined isoflavones and soy pro-
isoflavones on TC, LDL, and TG occurs within the short tein consumption to obtain favorable cardioprotective ben-
initial period of intervention (6 weeks). Therefore, it is efits, our study failed to show an improvement in the lipid
likely that our study was of sufficient duration to see initial profile. Future research should seek to clarify the biological
improvements in these specific markers. HDL improve- mechanism responsible for the lipid-lowering benefit of
ments, however, appear to require a longer intervention soyfoods, as well as the importance of study population
period as Zhan and Ho5 found that increases in HDL cho- characteristics, intervention type, dosage, and time course of
lesterol were only observed in studies of greater than 12 soyfood consumption necessary to achieve a significant im-
weeks of duration. provement in lipid profiles.
Dietary control
ACKNOWLEDGMENT
This study was conducted in free-living individuals.
Although this may have hindered protocol adherence, di- We would like to thank Dr. Mark J. Messina for
etary and compliance analysis suggests that it did not. All thoughtfully reviewing this manuscript and significantly
dietary variables included in this analysis were similar by contributing to its completion.
656 BEAVERS ET AL.
1. Beth H Rice, Erin E Quann, Gregory D Miller. 2013. Meeting and exceeding dairy recommendations: effects of dairy consumption
on nutrient intakes and risk of chronic disease. Nutrition Reviews 71:4, 209-223. [CrossRef]