点击显示 收起
1 From the Division of Epidemiology and Biostatistics, the School of Public Health, the University of Illinois at Chicago; the Division of Nutritional Sciences, the University of Illinois at Urbana-Champaign; the Immunoassay Core Facility Laboratory, the Robert H Lurie Comprehensive Cancer Center, Northwestern University, Chicago IL; the Department of Pharmacology and Toxicology, the University of Alabama at Birmingham; the Murray Rayburn Laboratory of Biochemical Endocrinology, Strang Cancer Research Laboratory, New York; and Protein Technologies International, St Louis.
2 Supported by the Illinois Soybean Program and the National Cancer Institute (RO3CA64459-01A1). The mass spectrometer used in this study was purchased with funds from an NIH Instrument Grant (S10RR06487) and the University of Alabama at Birmingham (UAB). Operation of the Mass Spectrometry Shared Facility at UAB is supported in part by an NCI Core Research Support Grant (P30 CA13148) to the UAB Comprehensive Cancer Center. Soy protein products were provided by Protein Technologies International, St Louis. 3 Address reprint requests to VW Persky, Room 508, 2121 Taylor Street, Chicago, IL 60612. E-mail: vwpersky{at}uic.edu.
ABSTRACT
Background: The long-term clinical effects of soy protein containing various concentrations of isoflavones on endogenous hormones are unknown.
Objective: We examined the effects of ingestion of soy protein containing various concentrations of isoflavones on hormone values in postmenopausal women.
Design: Seventy-three hypercholesterolemic, free-living, postmenopausal women participated in a 6-mo double-blind trial in which 40 g protein as part of a National Cholesterol Education Program Step I diet was provided as casein from nonfat dry milk (control), isolated soy protein (ISP) containing 56 mg isoflavones (ISP56), or ISP containing 90 mg isoflavones (ISP90). Endogenous hormone concentrations were measured at baseline and at 3 and 6 mo.
Results: The concentration of thyroxine and the free thyroxine index were higher in the ISP56 group, and the concentration of thyroid-stimulating hormone was higher in the ISP90 group than in the control group at 3 and 6 mo (P < 0.05). Triiodothyronine was significantly higher in the ISP90 group only at 6 mo. Thyroxine, free thyroxine index, and thyroid-stimulating hormone at 6 mo were inversely associated with measures of baseline estrogenicity. No significant differences were found for endogenous estrogens, cortisol, dehydroepiandrosterone sulfate, insulin, glucagon, or follicle-stimulating hormone after baseline hormone values were controlled for.
Conclusions: This study does not provide evidence that long-term ingestion of soy protein alters steroid hormone values, but it suggests that soy protein may have small effects on thyroid hormone values that are unlikely to be clinically important. The thyroid effects are, however, consistent with previous findings in animals and highlight the need for future research investigating possible mechanisms of action.
Key Words: Soy protein • hormones • postmenopausal women • isoflavones • National Cholesterol Education Program Step I diet
INTRODUCTION
Previous studies suggested that high soybean intakes are associated with lower concentrations of serum cholesterol (1) and may be related to decreased rates of coronary artery disease (2), cancer (3), and osteoporosis (2). One of the postulated mechanisms of these effects is that soy products, particularly the isoflavones in soy, act through changes in endogenous hormonal balance. Isoflavones were shown to have both estrogenic and antiestrogenic effects in animals (4–9). Although Divi et al (10) noted that genistein inhibits thyroid peroxidase and decreases thyroxine synthesis, most animal studies found that soy protein increases thyroxine (T4) concentrations and has inconsistent effects on triiodothyronine (T3) and thyroid-stimulating hormone (TSH) (11–18).
The results of human studies have been inconsistent. In men, Ham et al (19) found increases in T4 and in the free thyroxine index (FTI) in hypercholesterolemic men after ingestion of soy-protein isolate for 4 wk, but not after ingestion of soy flour. In premenopausal women, one observational study noted positive correlations of sex hormone binding globulin (SHBG) with excretion of lignans and phytoestrogens (20), another study found no association of SHBG with intake of soy products (21) but an inverse association with serum estradiol (21), and a third study found a positive association of SHBG with dehydroepiandrosterone sulfate (DHEAS) (22). Trials in premenopausal women have also been inconsistent. One mounth of soy ingestion suppressed midcycle surges of follicle-stimulating hormone (FSH) and luteinizing hormone, increased follicular-phase estradiol, decreased cholesterol concentrations (23, 24), and increased menstrual cycle length and decreased serum estradiol, progesterone, and DHEAS (25). Another 3-mo trial in premenopausal women found that ingestion of isolated soy protein (ISP) providing 129 mg isoflavones/d resulted in lower free T3, estrone, and DHEAS values; lower urinary estrogen values; and a higher ratio of urinary 2-hydroxyestrone (2OHE1) to 16-hydroxyestrone (16OHE1) during the follicular phase of the women's cycles than did ingestion of 10 mg isoflavones/d (26, 27).
In postmenopausal women, a similar 3-mo trial with a high-isoflavone diet resulted in a small but significant decrease in estrone sulfate values, a trend toward a decrease in estradiol and estrone values, and a small but significant increase in SHBG values (28). There were no significant differences in thyroid hormones or in measures of estrogenicity in vaginal epithelium or endometrial biopsies (28). In contrast, also in postmenopausal women, Baird et al (29) found no changes in FSH, luteinizing hormone, SHBG, or estradiol, but did note a small estrogenic effect on vaginal cytology after soy ingestion; Petrakis et al (30) found no consistent changes in plasma prolactin, SHBG, estradiol, or progesterone after 5 mo of soy ingestion.
Many of these studies, especially in premenopausal women, were short term (22–25). Yet, chronic exposure to soy alters the kinetics of isoflavone metabolism (31). Several months of exposure may be required to delineate long-term endocrine effects. The present study was undertaken to determine the effects of ingestion of soy protein with different concentrations of isoflavones for 6 mo on endogenous hormones in postmenopausal women.
METHODS
Subjects
Postmenopausal women were recruited from the vicinity of Urbana-Champaign, IL, through flyers, letters, and physician referrals. The participants were studied during 3 separate 24-wk cohort periods: cohort 1 during the summer and fall of 1994 (n = 24), cohort 2 during the spring and summer of 1995 (n = 25), and cohort 3 during the summer and fall of 1995 (n = 25). Within each of these cohorts, the women were randomly assigned to each of the 3 diet groups described below. Inclusion criteria were the completion of menopause, 1 y since the last menstrual period, and a total plasma cholesterol concentration of 6.2–7.8 mmol/L. Women were excluded if they were receiving hormone replacement therapy, taking other medication known to lower cholesterol, had a history of diabetes mellitus or thyroid disease, had a chronic illness that might affect lipid measurements or limit their ability to participate in the study, or had an allergy to soybean protein.
Of a total of 134 women screened, 81 women aged 49–83 y gave informed consent and 74 women completed the study. Seven women withdrew from the study during weeks 1–4 for various reasons: 1 moved to another state, 3 withdrew because of medical problems unrelated to soy-product consumption, and 3 withdrew because of their inability to comply with the study protocol. One woman who was taking levothyroxine was inadvertently included, and her results were, therefore, excluded from all analyses. Thus, statistical analyses were conducted on data from a total of 73 women. The study was approved by the Institutional Review Boards of the University of Illinois at both Chicago and Urbana-Champaign.
Diet
Two weeks before the study, participants completed a 2-d dietary intake record and were interviewed by a registered dietitian to calculate their daily energy requirement for a basal low-fat, low-cholesterol National Cholesterol Education Program Step I diet. Each participant received a booklet published by the American Heart Association containing a long list of foods, along with a calculated fat gram prescription. All participants followed the basal diet for 14 d. After this, baseline blood samples were drawn on 2 separate days, and participants were randomly assigned to 1 of 3 dietary treatment groups. Most of the participants were already consuming diets similar to the National Cholesterol Education Program Step I diet before enrolling in the study. All 3 groups continued to consume their basal diets to which 40 g test protein/d was incorporated as ISP (Supro 675; Protein Technologies International, St Louis) containing moderate (56 mg) concentrations of isoflavones (ISP56 group; n = 24), as ISP containing higher (90 mg) concentrations of isoflavones (ISP90 group; n = 23), or as casein (0 mg total isoflavones/g protein; New Zealand Milk Products, Wellington, New Zealand) from nonfat dry milk (control group; n = 26) (Table 1). Units of isoflavones presented in this paper reflect only the aglycone weight of the molecule. The ISPs were fortified to provide 800–900 mg Ca/d in the form of calcium bisphosphate. This amount was consistent with the amount of calcium provided through the dairy component for the control group. Participants were advised not to ingest other soy products during the study and were blinded to the source of test protein during the trial period. Investigators involved in the production of test protein–containing products were aware of dietary treatment codes; however, laboratory personnel were not.
View this table:
TABLE 1 . Isoflavone content of the 3 test proteins in the 3 diet groups1
The Department of Food Science and Human Nutrition at the University of Illinois developed recipes that incorporated the casein and ISPs into baked products. Ready-to-mix beverages and soups containing ISP or casein were also used. Participants were informed of the amount of protein, fat, and energy provided by each of the test protein items. All baked products were prepared by a professional baker in the University of Illinois Quantity Foods Laboratory and were individually packaged and distributed to participants during a breakfast held at the study site 3 d/wk. Cereals, fruit, tea, coffee, juice, and low-energy carbonated beverages were also offered during the breakfast. Throughout the study, all participants were instructed by a registered dietitian to maintain their body weights and physical activity levels and about dietary requirements and acceptable food and beverage intakes. Additionally, every 4 wk, participants were asked to complete a dietary intake record on 3 d (including 2 weekdays and 1 weekend day) chosen at random. Actual daily nutrient intakes were analyzed by using a nutritional analysis software program (NUTRITIONIST IV, version 3.0; N-Squared Computing, Salem Park, OR). A summary of the isoflavone content of the 3 diets is presented in Table 1.
Blood and urine sample collection
Blood samples for hormone measurements were collected at baseline and 3 and 6 mo after initiation of the diets. Blood was collected at 0600–0900 after a 12-h fast from either the antecubital or dorsal hand vein. Serum and plasma were separated by centrifugation at 1190 x g for 15 min at 4°C and stored at -70°C before shipment to the laboratories of 2 of the authors (RC and SB) for analysis of hormones and isoflavones, respectively. For estrogen metabolites, 24-h urine specimens (each bottle containing 100 mg ascorbic acid) were collected only from the second 2 cohorts at baseline, 3 mo, and 6 mo, and aliquots were stored at -70°C before shipment to Bradlow's laboratory for analysis.
Weight measurements
Body weight was measured weekly by using a scale equipped with a beam balance (Detecto Physician's scale; Detecto Scale Company, Webb City, MO); the scale was calibrated weekly.
Hormones
Measurements
The Delphia system (Walla, Gaithersburg, MD) was used for the assay of serum SHBG. This system is a solid-phase, two-site, time-resolved fluoroimmunometric assay that uses a sandwich technique. The intraassay and interassay CVs were 5.0% and 5.3%, respectively. DHEAS was measured in unextracted serum by using radioimmunoassay (RIA) as described previously (32). The intraassay and interassay CVs were 4.6% and 10.9%, respectively. Plasma cortisol was measured with a direct assay described by Seth and Brown (33) as used in other studies (34). The intraassay and interassay CVs were 7.7% and 9.6%, respectively. FSH was measured with the use of a double-antibody method described previously (35, 36). The intraassay CV was 2.6%; the interassay CVs were 3.2% and 4.2% for the means of 38.7 and 4.2 U/L, respectively. Serum estradiol was measured with the use of a solid-phase competitive immunoassay (Delphia) that uses a europium-labeled estradiol and time-resolved fluorescence measurement for quantification. Intra- and interassay CVs were 5.8% and 8.2%, respectively. Estrone sulfate was measured with the use of a double-antibody RIA with 125I estrone sulfate from Diagnostic Systems Laboratory, Webster, TX. The intraassay CV from duplicate determinations was 6.0%; the interassay CV was 6.4%. SHBG-bound estradiol was measured as described by Bonfrer et al (37, 38). The intraassay CV was 3.3%. Insulin was measured with the use of a solid-phase RIA (39) and glucagon with a double-antibody RIA (40), both from Diagnostics Product Corp, Los Angeles. The total T4 concentration and the percentage of unbound T3 were measured with the use of an enzyme immunoassay (41, 42) on a Hitachi 704 Auto-analyzer (BM/H 704 analyzer; Boehringer Mannheim, Indianapolis). The FTI was calculated from total T4 and T3 uptakes. CVs for external quality controls were 4.06% and 1.24% for T4 and T3, respectively. The total T3 concentration was assayed by solid-phase RIA (42) and TSH by double-antibody RIA (43), both from Diagnostics Product Corp. Interassay CVs for insulin, glucagon, and thyroid hormones were all <1%. Intraassay CVs were set at <10%, and any duplicate result that differed from the original result by >10% was repeated.
Quality control
External quality-control standards for steroid hormones were obtained from the American College of Pathologists. For internal quality control, a single batch of each of the quality-control materials and antisera for each of the hormones analyzed were prepared and reserved for all assays during the 2 y of the study to minimize interassay variability. Blood specimens were analyzed by technicians unaware of the participants' diet group, and 2.3% of the samples were submitted to the laboratory as blind duplicates as an additional quality control for steroid hormones. CVs from the 5 split serum samples were 10.6% for cortisol, 36.5% for DHEAS, 4.5% for SHBG, 2.4% for SHBG-bound estradiol, 11.1% for FSH, 24.0% for estradiol, and 29.2% for estrone sulfate. Serum was stored at -70°C in 3–4-mL aliquots for measurement of additional hormones if subsequent research indicates that these hormones may be affected by soy or isoflavone intake.
Enzyme-linked immunoassay for 2-hydroxyestrone and 16-hydroxyestrone in urine
The hormones 2OHE1 and 16OHE1 were measured in urine with the use of a competitive solid-phase enzyme immunoassay (Immuna Care Corporation, Bethlehem, PA) (44). Both assays indicated 100% recovery of metabolites with serial dilution and addition of exogenous estrogens into urine samples. The intraassay CV was 6% and the interassay CV was 10%. Values for the individual metabolites were divided by the total urine volume.
In the present study, 11.6% of duplicate urine samples were submitted blind as an additional quality control. The CVs from the split urine samples were 16.0% for 2OHE1 and 36.4% for 16OHE1. One-half of the limit of detection for the assay (0.325 nmol/d) was imputed for 13 samples that had nondetectable amounts of 16OHE1. Concentrations of urinary estrone metabolites were multiplied by the total volume of 24-h urine samples.
Plasma isoflavones and lipids
Plasma total and unconjugated isoflavones were measured with the use of a modified version of an HPLC–mass spectrometric method described previously (45). For a 1-mL blood sample, the overall detection limit ranged from 4 to 10 nmol/L. The median value for the CV was determined for 2 ranges: from the limit of detection for each compound to 40 nmol/L (10.2%–21.5%) and from 40 nmol/L upward (3.1–7.7%). CVs from split samples were 20.1% for equol, 25.0% for daidzein, 23.4% for dihydrodaidzein, 34.4% for O-desmethylangolensin (O-DMA), and 21.8% for genistein. Plasma lipids were analyzed as previously described (46, 47).
Bone mineral density
Bone mineral density of the lumbar spine (L1–L4), proximal femur (including the femoral neck and Ward's triangle), and total body was measured at baseline and 6 mo after the study began as described previously (47). In vivo precision errors for the lumbar spine, proximal femur, and total body bone mineral density were 1.3%, 1.8%, and 1.0%, respectively.
Statistical analyses
Arithmetic means and 95% CIs were calculated for the isoflavones and hormones at baseline and at 3 and 6 mo. Geometric means are presented for the values of all isoflavones, estradiol, estrone sulfate, SHBG, DHEAS, TSH, insulin, insulin:glucagon, 2OHE1, 16OHE1, and the ratio of 2OHE1 to 16OHE1, which were log-transformed before analyses to meet the normal distribution assumption of the statistical test. Student's t tests were used to compare concentrations of serum isoflavones and endocrine outcomes at baseline between the ISP56 and control groups and between the ISP90 and control groups. For isoflavones, regression models were used to compare each diet group with the control group after adjustment for baseline values.
The effects of dietary supplementation on endocrine outcomes at 3 and 6 mo were further analyzed by using random-effects models (PROC MIXED) to evaluate the ISP56 diet (ISP56 compared with control) and the ISP90 diet (ISP90 compared with control), adjusting for clustering of hormone measures within participants at 3 and 6 mo, baseline hormone concentration, and time of sample (3 mo compared with 6 mo), and including interaction terms for diet group and time of sample. Group x time interaction terms indicated differences in effects between treatment groups at 3 and 6 mo. When a group x time interaction term was significant, the model was recoded to indicate differences in effects between treatment groups at 3 and 6 mo. When group x time interaction terms were not significant, they were removed from the model. Treatment group terms then indicated differences in the averages of 3- and 6-mo effects between treatment groups. The outcomes analyzed included body mass index (wt/ht2); serum T3, T4, FTI, TSH, estradiol, estrone sulfate, cortisol, insulin, glucagon, insulin:glucagon, DHEAS, FSH, SHBG, and SHBG-bound estradiol; and urinary 2OHE1, 16OHE1, and 2OHE1:16OHE1.
Analyses of 2OHE1 and 16OHE1 included (data presented in tables) and did not include (data not presented) data from 5 participants who had metabolic values or ratios of 2OHE1 to 16OHE1 >3 SDs from the mean. The results were not altered by exclusion of these 5 women.
Spearman's correlation coefficients were used to associate changes in isoflavone concentrations (6 mo concentration - baseline concentration) with changes in hormone values (6 mo concentration - baseline concentration). Partial Pearson's correlation coefficients, after baseline thyroid hormone values and treatment group (as defined for the multiple regression analyses) were controlled for, were used to assess the relation of baseline estrogenicity with changes in thyroid hormones at 6 mo.
CVs for split samples of hormones were computed as the technical error per mean. Technical error was estimated as (di2/2n)1/2, where di is the difference in values between the 2 identical samples and n is the number of pairs of split samples. SAS (version 8.0; SAS Institute, Inc, Cary, NC) was used for all analyses. P 0.05 was considered statistically significant.
RESULTS
Daily energy intakes of the diet groups were not significantly different at baseline or during the study (46). There were also no significant differences in intakes of selected nutrients, except for total protein intake, which increased in all 3 groups during the study (46). The dietary intervention resulted in expected changes in serum isoflavone concentrations. After baseline concentrations were adjusted, women in both ISP groups had higher concentrations of isoflavones at 6 mo than did women in the control group (Table 2).
View this table:
TABLE 2 . Plasma isoflavone concentrations in the 3 diet groups at 0 and 6 mo1
There were no significant differences at baseline in mean age or body mass index between the 3 diet groups (data not shown). The average ages were 59.3, 61.9, and 61.0 y and the average body mass indexes were 28.0, 26.9, and 29.6 in the ISP56, ISP90, and control groups, respectively. Similarly, there were no significant differences in body mass index after 3 or 6 mo of dietary intervention. Few women smoked (2–3 in each diet group) or drank any alcohol (4–7 in each diet group).
There were no significant differences in any of the estrogen values at baseline or over time (Table 3). Similarly, there were no significant differences in DHEAS, FSH, insulin, or glucagon between groups (Table 4). SHBG concentrations differed between the ISP90 and control groups averaging over time, but this difference was of borderline significance (P = 0.06). Cortisol concentrations at baseline were higher in the ISP56 group than in control subjects, but differences were no longer present averaging over 3 and 6 mo.
View this table:
TABLE 3 . Estrogen values at 0, 3, and 6 mo1
View this table:
TABLE 4 . Non-estrogen hormone values in the 3 diet groups at 0, 3, and 6 mo1
The effects of ISP ingestion on thyroid hormones are shown in Table 5. T4 concentrations and the FTI were significantly higher in the ISP56 group than in the control group when averaged across 3 and 6 mo. TSH was significantly higher only in the ISP90 group when averaged across time. T3 was also significantly higher only in the ISP90 group, but for T3, there was a significant group x time interaction, indicating that the difference was significant only at 6 mo.
View this table:
TABLE 5 . Thyroid hormone values in the 3 diet groups at 0, 3, and 6 mo1
Changes in T4, FTI, and TSH between baseline and 6 mo were inversely related to measures of estrogenicity at baseline (Table 6). Thus, the change in T4 was significantly and inversely associated with total bone mineral density, serum estradiol, and urinary 2OHE1 concentrations. The change in the FTI was significantly and inversely related to body mass index, total bone mineral density, serum estradiol, serum estrone sulfate, urinary 2OHE1, and urinary 16OHE1. The change in TSH concentration was significantly and inversely related to body mass index, total bone mineral density, and lumbar spine bone mineral density at baseline.
View this table:
TABLE 6 . Pearson's partial correlation coefficients (r) for the effects of baseline estrogenization on changes in thyroid hormones at 6 mo1
Changes in isoflavone concentrations were also significantly related to changes in measures of estrogenicity, as reflected by changes in bone mineral density and HDL-cholesterol concentrations at 6 mo compared with baseline (data not shown). Changes in the concentrations of genistein, daidzein, dihydrodaidzein, O-DMA, total daidzein, and total isoflavones were all positively associated with changes in lumbar spine bone mineral density at 6 mo, and changes in the concentrations of genistein, O-DMA, total daidzein, and total isoflavones were significantly and positively associated with changes in HDL concentrations at 6 mo. Changes in T4 and the FTI at 6 mo, however, were significantly and positively associated only with changes in the concentrations of the daidzein metabolites dihydrodaidzein and O-DMA (data not shown).
DISCUSSION
This study provides no evidence that long-term ingestion of ISP containing moderate or high concentrations of isoflavones alters serum or urinary estrogen, cortisol, DHEAS, insulin, glucagon, or FSH in postmenopausal women. There is a suggestion that SHBG concentrations may be increased in the ISP90 group, but the relation was of only borderline significance. Petrakis et al (30) found no change in plasma estrogen concentrations in 10 postmenopausal women after they ingested 38 g soy protein for 6 mo. Duncan et al (28) noted small but significant decreases in estrone sulfate concentrations, a trend toward lower estradiol and estrone concentrations, and small but significant increases in SHBG concentrations in postmenopausal women who had consumed a high isoflavone diet for 3 mo. Several studies in premenopausal women found decreased estrogen concentrations and changes in estrogen metabolite concentrations (21, 25–27), and one study in premenopausal women found an increase in the concentration of follicular estradiol and a decrease in the midcycle surge in the concentrations of luteinizing hormone and FSH (23, 24); however, these studies used shorter intervention periods than those in the present study. Differences between those findings and the findings in the present study may be due to altered isoflavone metabolism after long-term ingestion of soy protein, as noted previously (31). It is also possible that soy may act as an estrogen antagonist in premenopausal women with high endogenous estrogen concentrations and as an estrogen agonist in postmenopausal women with lower endogenous hormone concentrations, with estrogenicity better reflected in end-organ effects than in serum hormone concentrations. This possibility is supported by the findings of Baird et al (29) that soy ingestion was associated with increased vaginal estrogenicity but was not associated with any alterations in the concentrations of serum estradiol or FSH. This possibility is also consistent with reductions in hot flushes (48), increased vaginal cell maturation (49), and perceived severity of vasomotor symptoms (50) noted in other studies, and with the increases in bone density found in our previous study (47) and in animal studies (51, 52).
Data on the effects of soy on DHEAS are also conflicting. In premenopausal women Lu et al (25) and Duncan et al (27) found decreases in DHEAS concentrations, whereas Persky and Van Horn (22) found increases in DHEAS concentrations. Doses of isoflavones, however, differed among the studies. The average dose of 129 mg/d in Duncan et al's high-isoflavone group (27) was higher than that ingested by the women studied by Persky and Van Horn (22). DHEAS concentrations in the study by Duncan et al were insignificantly higher in the low-isoflavone group than in control subjects after 3 mo (27), suggesting that different doses may account for some of the discrepancies among the studies. In postmenopausal women, Duncan et al (28) noted inconsistent effects on DHEAS of 129 mg soy isoflavones/d in the high-isoflavone group; again, concentrations were higher than in the present study. Alternatively, differences could reflect the fact that the present study compared the effects of 2 concentrations of isoflavones in soy protein with milk protein, whereas the study by Duncan et al (28) was designed to compare the effects of isoflavone-rich protein with the effects of isoflavone-free protein.
The present study provides some evidence that soy protein increases measures of thyroid function. Differences reflect changes in both the experimental and control groups and could occur by chance. The differences are also small and unlikely to be clinically important. This is one of the few studies of the effects of soy on thyroid hormones in postmenopausal women. The results are consistent with many previous studies in animals that showed elevations of T4 and, less consistently, of T3, FTI, and TSH after ingestion of diets containing ISP products (11–16, 53). The results are also consistent with those previously published in a study of men by Ham et al (19). In that study, however, thyroid concentrations of men who ingested soy protein were significantly higher than their concentrations at baseline but were not significantly higher than those of men who consumed control diets, suggesting a general response to dietary change. The results are not consistent with the decrease in free T3 previously noted in premenopausal women (27) or with the lack of change in thyroid hormones noted before in postmenopausal women (28) who ingested higher amounts of isoflavones than consumed in the present study.
The fact that some of the thyroid effects were most apparent in the present study at 6 mo suggests that these may be an adaptation to long-term dietary changes. The dissociation of the effects on T4 and the FTI compared with those on TSH, with changes in T4 and FTI values manifested at moderate concentrations of isoflavones and changes in TSH concentration only at high concentrations of isoflavones, also suggests that there may be 2 mechanisms by which soy protein acts on thyroid function. One effect may result in an increase in the T4 concentration at lower doses of isoflavones, perhaps through displacement of T4 from its binding proteins, similar to effects found previously with flavonoids (54). The other effect, at higher doses of isoflavones, may result in a decrease in the T4 concentration, with a compensatory increase in TSH. The latter would be consistent with the decreased in vitro synthesis of thyroid hormones after ingestion of purified genistein and daidzein noted by Divi et al (10) and with the decreased thyroid hormone concentrations noted in a few reports of infants, most of which were in infants who were iodine deficient and which used relatively high doses of soy in formula (55–58).
The possibility that the effect of soy on thyroid hormones may be related to its estrogenic properties is suggested by the finding that changes in thyroid hormones were greatest in women with the lowest measures of estrogenicity at baseline. Estrogenic effects of soy are also supported by the associations of changes in isoflavone concentrations with changes in bone mineral density and HDL-cholesterol concentrations at 6 mo in the present study. It is possible that at different isoflavone doses soy may also act as an estrogen, as an antiestrogen, or through other mechanisms on thyroid balance. There is evidence that estrogens may increase the sensitivity of the pituitary or thyroid gland to normal feedback mechanisms (59, 60). The dissociation of changes in TSH and peripheral hormones, however, in the present study does not support this hypothesis. Previous studies also suggest that estrogens and thyroid hormones may interact in the regulation of gene expression in selected target tissues, but the effects of this interaction are not yet understood (61).
In summary, the present study does not show significant effects of soy protein on serum or urinary estrogens, SHBG, FSH, cortisol, DHEAS, insulin, or glucagon. The study does show small effects on thyroid hormones that are unlikely to be clinically important. Future animal and human studies are needed to confirm these findings and to address in more detail the mechanisms by which soy could affect thyroid hormones.
ACKNOWLEDGMENTS
We are grateful to Lori Coward for performing the isoflavone assays and to Marion Kirk for performing the mass spectrometry analysis.
REFERENCES