Key Questions

Five general questions are addressed in this report:

1.

In the clinical trial literature, what formulations of soy were used? At what dose? For what purpose(s) (e.g., trial endpoints)?

2.

Does current clinical trial evidence indicate that whole soy products and individual constituents of soy have an effect on:

a.

Cardiovascular events, risk factors, and measures.

b.

Menopausal symptoms.

c.

Endocrine function.

d.

Cancer and tumor-related biomarkers.

e.

Osteoporosis and osteoporosis risk factors.

f.

Reproductive health.

g.

Kidney function.

h.

Other outcomes, based on results of Key Question 1?

3.

What is the scientific evidence of a dose-response effect of different forms of soy and individual constituents of soy for the conditions specified in Key Question 1?

4.

What are the frequency and type(s) of adverse events associated with consumption of soy that are reported in the scientific literature (both trials and epidemiology)?

5.

What is the scientific evidence of a dose-response effect of whole soy products and individual constituents of soy on their safety?

Approach to Analyzing the Literature

Literature Search Strategy

We conducted a comprehensive literature search to address the key questions.* Primary literature searches for English language publications on soy studies were conducted in EMBASE® on March 25, 2004, in MEDLINE® on April 20, 2004, and in CAB Abstracts on June 24, 2004. Search terms included subject headings and textwords with filters to limit the publications to English language and primary studies of the adult and adolescent human populations. Subject headings and text words were selected so that the same set could be applied to each of the different databases. A supplemental search was performed in MEDLINE® on April 30, 2004, to retrieve articles using the textword “miso”. A search update was performed in MEDLINE® In-Process & Other Non-Indexed Citations and MEDLINE® on September 30, 2004, and CAB Abstracts on October 4, 2004. A search of the TOXLINE® database was conducted in March 31, 2005 to identify additional reports for adverse events in humans. Additional sources of published and unpublished data were sought by contacting members of the TEP, and from reference lists of selected review articles and meta-analyses.

Methodological Quality Grade

We used a 3-category grading system (A, B, C) to denote the methodological quality of each study. This system defines a generic grading system that is applicable to varying study designs including randomized controlled trials, cohort, and case-control studies:

• A—Least bias; results are valid; a study that mostly adheres to the commonly held concepts of high quality.
• B—Susceptible to some bias, but not sufficient to invalidate the results, study that does not meet all the criteria in category A, above.
• C—Significant bias that may invalidate the results; a study with serious errors in design, analysis, or reporting.

Applicability Grade

In this report, the focus is on the U.S. population and on specific subgroups within that population (i.e., post-menopausal women, peri-menopausal women, pre-menopausal women, men, and people with relevant medical histories, such as breast cancer). Even though a study may focus on a specific target population, limited study size, eligibility criteria, and the patient recruitment process may result in a narrow population sample that is of limited applicability, even to the target population. To address this issue, we categorized studies within a target population into 1 of 3 levels of applicability that are defined as follows: sample is representative of the target population; sample is representative of a relevant sub-group of the target population, but not the entire population; sample is representative of a narrow subgroup of subjects only, and is of limited applicability to other subgroups.

Meta-analysis

Meta-analysis was performed for several cardiovascular outcomes. We used the random effects model for continuous outcomes to combine studies. We also performed several random effects model meta-regression analyses to explore possible reasons for discrepancies across studies and to address Key Questions related to dose-response.

Soy Products

Soy supplements were used in about three-quarters of all the trials analyzed in this report; soy foods were used in the remaining trials. In this report, soy milk was categorized as a soy supplement. Among the soy supplement trials, 57 percent used soy protein with isoflavones, 36 percent used isoflavones alone, and 6 percent soy protein without isoflavones. In about one-half of the soy foods trials, textured soy protein was used. Soy flour was used in about one-quarter of the soy foods trials. There are 146 separate treatment arms of soy supplementations and 68 separate treatment arms of soy foods or diets. Across studies, the total isoflavones ranged from 0 mg to 185 mg per day and the total protein intake from soy ranged from 0 g to 154 g per day. Of note, the median soy product dose across studies (36 g soy protein per day) was equivalent to over a pound of tofu daily or about 3 soy protein shakes daily.

Cardiovascular Endpoints

No study evaluated clinical cardiovascular events. A total of 68 randomized studies reported data on total cholesterol, LDL, HDL, and/or triglycerides. The total isoflavones ranged from 0 mg to 185 mg per day, with a median of 80 mg. Among studies with soy protein, the total protein intake from soy ranged from 14 to 113 g per day, with a median of 36 g.

There is a great deal of heterogeneity of effects found on lipoprotein and triglyceride levels. Overall, the majority of studies reported small to moderate effects on the lipids, despite a wide range of net effects for total cholesterol, LDL, and triglycerides. Sixty-one studies reported data on the effect of consumption of soy products on total cholesterol levels. The median net change compared to control found was approximately -5 (interquartile range -10, +1) mg/dL decrease (about -2.5%). A meta-analysis of 52 studies that reported data on the effect of soy consumption on LDL levels yielded a statistically significant net decrease of 5 (95% confidence interval [CI] -8 to -3) mg/dL (about -3%). A meta-analysis of 56 studies that reported data on the effect of soy consumption on HDL levels found a statistically non-significant net change of +0.6 (95% CI -0.5, +1.8) mg/dL. A meta-analysis combining 54 studies that reported data on the effect of soy consumption on triglyceride levels yielded a net change of -8 (95% CI -11, -5) mg/dL (about -6%).

Across studies, there is the possible suggestion that higher doses of soy protein are associated with greater LDL reduction among those with elevated baseline LDL (although not if studies with minimal soy protein doses are excluded), but not with HDL or triglycerides. Dose of isoflavones was not associated with effect for any lipid. Higher baseline LDL or triglycerides may also be associated with net effect for these 2 lipids; the effect of baseline HDL is unclear. For all lipids, in individual studies the effect of dose and baseline was generally inconsistent.

A total of 22 studies reported data on the effect of consumption of soy products on systolic and diastolic BP. Overall, across studies, there was no discernable effect.

Some of the well known emerging risk factors for cardiovascular disease included for analysis in this report are: Lp(a), CRP, homocysteine, endothelial function, systemic arterial compliance, and oxidized LDL. The total number of studies that reported data on the effect of soy consumption are: 18 studies on Lp(a), 3 on CRP, 5 on homocysteine, 10 on endothelial function, 3 on systemic arterial compliance, and 13 on oxidized LDL. Across these studies, there is no discernable effect based on the type of soy products. The majority of studies were of poor quality with a narrow range of applicability. Given the limited evidence and poor quality studies, no conclusions could be drawn on the beneficial or harmful effects of consumption of soy protein on these putative risk factors for cardiovascular disease.

Menopausal Symptoms

A total of 21 trials examine the effects of soy and/or its isoflavones on hot flashes and night sweats in women. These trials generally measured frequency and severity of the symptoms. However, a large number of vasomotor symptom scores or indexes that employed a variety of frequency intervals were used by the investigators. These factors made meta-analyses unsuitable and limited the comparisons of results across studies. Furthermore, many of the studies had high withdrawal or dropout rates that were frequently uneven between soy treatment and control arms, further limiting the validity of these trials.

Overall, the effects of soy protein and/or its isoflavones are inconsistent across studies. Every trial found a decrease in hot flash frequencies or scores in both the treatment groups and the control groups. Thus, the results are difficult to interpret. A third of the studies found no or worsening effects compared to control; two-thirds showed soy protein and/or its isoflavones either non-significantly or significantly decreased hot flash frequencies or scores compared to control in post-menopausal women. The evidence of a benefit was stronger among the randomized trials of isoflavone supplements, which mostly showed positive results—the net reduction in weekly hot flash frequency ranged from 7 to 40 percent. However, these trials are mostly rated as poor quality due to high dropout rates.

There are only 4 studies that evaluated the effect of soy consumption on menopausal symptoms in peri-menopausal women or those receiving breast cancer therapy. Among these studies there is no evidence that soy consumption is better than control to reduce menopausal symptoms.

Endocrine Function

Measures of endocrine function were reported from 50 trials in 47 articles. Five studies with a total of 179 participants reported testosterone levels in healthy males before and after soy consumption. Four of these trials found a statistically non-significant decrease in testosterone levels. The small total number of subjects as well as the low quality of these studies precluded any meaningful conclusion. No statistically significant effect was found on FSH level, which is commonly measured in the initial evaluation of male and female infertility; results were conflicting.

Twelve studies reported estradiol levels at the follicular phase in 434 pre-menopausal women. The overall effect of soy on estradiol levels was not consistent. Most of the studies showed a trend for soy reducing estradiol, although they failed to demonstrate a statistically significant effect. Six randomized trials reported the effect of soy on TSH. No overall effect of soy on TSH and thyroid function is clear.

An additional 11 trials (in 10 publications) evaluated the effect of soy on menstrual cycle length in pre-menopausal women. A wide range of soy interventions was used in these trials making a conclusion of the effects from soy difficult. These trials did not show statistically significant changes in menstrual cycle length after treatments of soy and/or its isoflavones.

Cancer and Tumor-Related Biomarkers

Twenty-four trials evaluated subjects without a history of cancer for effects of soy on tumor-related biomarkers. No study reported the development of cancer as an outcome. Most studies measured the effect of soy on estrogens and estrogen metabolites as well as on estrogenicity indicators. There were also trials that evaluated correlations between soy and possible cellular pathways of cancer prevention. No causal relationship could be established between these markers and cancer because they do not represent known risk factors for cancer disease. Only 4 studies reported on testosterone levels, which is a risk factor for prostate cancer, and are discussed under Endocrine Function.

Bone Endpoints

Overall, 31 studies evaluated various markers of bone health, including bone mineral density (BMD), bone formation biomarkers (bone specific alkaline phosphatase and osteocalcin) and bone resorption biomarkers (urinary hydroxyproline, urinary cross-linked N-telopeptide, urinary pyridinoline, and urinary deoxypyridinoline).

Because there are few long-term randomized trials and a wide variety of soy interventions used across studies, it is difficult to draw an overall conclusion about the effects of soy on bone outcomes. Overall, among the 5 studies of 1-year minimum duration, no consistent effect on BMD was seen with soy consumption. Studies of shorter duration, likewise found no effect of soy. Similar to the results in BMD, studies of bone formation biomarkers generally found no effect of soy consumption when compared to control. While a number of studies reported reductions in 2 markers of bone resorption—urinary pyridinoline and deoxypyridinoline—no effects were found on the other markers of bone resorption and the effects were not consistent across studies. For these markers there is no clear evidence of a dose effect for either soy isoflavones or soy protein.

Only one study found a consistent effect on these markers. The study differed from other studies in that it evaluated a unique formulation of soy genistein and that it excluded subjects with denser femoral neck BMD.

Kidney Function, Neurocognitive Function, and Glucose Metabolism

Only 1 small study in patients with type 2 diabetes assessed the effect of soy on kidney function. No statistically significant changes in glomerular filtration rate was seen after 8 weeks of soy protein diet. Four studies examined the effects of soy on cognitive function of post-menopausal women and college students of both sexes. Overall, no statistically significant or consistent effect was noted on neurocognitive functions such as verbal episodic memory. Six studies evaluated the effect of soy on fasting blood glucose. No statistically significant changes were reported.

Adverse Events

In general, the rates of adverse events reported were greater in the soy treatment arms than in their respective control arms, but adverse events related to soy consumption were generally minor. Overall, soy products including isoflavones were well tolerated in the trials we examined.

The most frequently reported adverse events among a total of 3,518 subjects in 49 studies (including 5 non-randomized and 3 pharmacokinetic studies) that reported adverse events were gastrointestinal in nature. These were reported in 33 of 41 comparison studies of soy diets, soy proteins, isoflavones, and phytoestrogen supplements. Most of the gastrointestinal adverse events were reported in soy diet and soy protein trials, especially the 12 studies that used purified isoflavone interventions in dosages ranging from 40 to 100 mg/day. The amount of soy protein in these trials ranged from 20 to 60 g/day, but there was no clear dose relationship between the amount consumed and subsequent adverse events. Menstrual complaints, reported in 15 studies, were also common. Six of these studies used purified isoflavone interventions in dosages ranging from 40 to 80 mg/day. However, most women in these studies were post-menopausal and the controls frequently included hormone therapy regimens. Other adverse events included musculoskeletal complaints, headache, dizziness, and rashes. In addition, there were somewhat more withdrawals from the soy arms due to taste aversion.