点击显示 收起
1 From the Fred Hutchinson Cancer Research Center, Cancer Prevention Research Program (CLF, REP, NKH, MLN, HES, and JWL) and the Departments of Epidemiology (CLF, REP, and JWL), Pharmaceutics (TFK), and Medicinal Chemistry (WNH), University of Washington, Seattle.
2 Supported by the Fannie E Rippel Foundation, the Fred Hutchinson Cancer Research Center, and the National Institutes of Health (T32 CA09661, R01 CA53996, R01 CA70913, and R03 CA80648). 3 Address reprint requests to JW Lampe, Fred Hutchinson Cancer Research Center, 1100 Fairview Avenue North, MP-900, PO Box 19024, Seattle, WA 98109-1024. E-mail: jlampe{at}fhcrc.org.
ABSTRACT
Background: Soy foods may have various health benefits, but little is known about the patterns and correlates of soy consumption among postmenopausal women in the United States.
Objective: We assessed the reliability and validity of a soy food-frequency questionnaire (FFQ) and examined demographic, lifestyle, and dietary correlates of plasma isoflavone concentrations in postmenopausal women.
Design: In this cross-sectional study, soy isoflavone intake and plasma isoflavone concentration were analyzed in 96 postmenopausal women aged 50–79 y; the data were obtained at 2 visits that were 1 wk apart. Intake was determined with a 20-item soy FFQ and a comprehensive FFQ that included questions about tofu and soymilk. Fasting plasma daidzein and genistein concentrations were determined with liquid chromatography–mass spectrometry.
Results: Intraclass correlations between week 1 and week 2 values were >0.98 for both the soy and comprehensive FFQs. Median reported isoflavone intake was <2 mg/d. Pearson’s product-moment correlation coefficients relating isoflavone intakes with plasma isoflavone concentrations ranged from 0.35 to 0.43. Plasma isoflavone concentrations were positively associated with age, fiber consumption, servings of fruit and vegetables, and dietary supplement use and were inversely associated with caffeine consumption; no associations with body mass index, education, dietary beliefs, activity level, alcohol intake, or fat intake were observed.
Conclusions: Within a population with low soy consumption, the soy FFQ and comprehensive FFQ showed good reliability and moderate validity. Associations of plasma isoflavone concentrations with other dietary behaviors suggest that these compounds may serve as biomarkers of health behaviors in populations with low soy consumption.
Key Words: Soy • isoflavones • postmenopausal women • reliability • validity • food-frequency questionnaire
INTRODUCTION
Physiologic changes associated with menopause are thought to increase the risks of various chronic diseases, including cardiovascular diseases and osteoporosis (1, 2). One potential mechanism for these increased risks is the sharp decline in endogenous estrogen production during menopause. Historically, hormone replacement therapy (HRT) has been used to alleviate these risks. However, it is not yet clear whether HRT is beneficial in preventing cardiovascular disease, and accumulating evidence suggests that HRT may be associated with an increased risk of breast cancer (3–6). Soy foods have been suggested as an alternative to HRT (7, 8) because they contain isoflavones, which are phytoestrogens that exhibit lower binding affinity and lower estrogenicity than does estradiol in experimental and animal models (9, 10). A variety of physiologic effects of isoflavones have been observed in animal models and in vitro; these include antiestrogenic, weak estrogenic, antiproliferative, antioxidant, and antiangiogenic effects (11), some of which may confer health benefits.
Epidemiologic evidence suggests that diets high in soy are associated with enhanced maintenance of bone mineral density and improvement of cardiovascular disease risk factors, such as increased HDL-cholesterol concentration and decreased fasting insulin concentration (12–15). However, there is conflicting evidence from observational studies that evaluated improvement of cardiovascular disease risk factors in relation to isoflavone intake (14, 16, 17). Results from short-term intervention studies of soy protein that have used different doses suggest that doses higher than those consumed in US populations are needed to elicit physiologic effects (15, 18, 19).
Little is known about soy food consumption patterns among postmenopausal women in the US population. The available data suggest that soy consumption in this population is low (20) and that it may be important to consider intake of isoflavones from nontraditional sources, including foods that do not contain soy, such as coffee and lentils (20, 21). However, isoflavone-containing foods other than soy foods appear to make a minimal contribution to total isoflavone exposure (20, 21). Although isoflavone intakes by postmenopausal women in the United States appear to be low, associations between intakes and disease risk factors have been reported in observational studies (14, 17). This raises the question of whether soy may be a marker for other dietary and lifestyle behaviors that influence disease risk. The objective of this study was to assess the reliability and validity of a soy food-frequency questionnaire (FFQ) in postmenopausal women and to identify dietary and nondietary factors associated with plasma isoflavone concentrations.
SUBJECTS AND METHODS
Participants
The participants were postmenopausal women aged 50–79 y who were recruited from King County, Washington, as part of a cross-sectional study. The study assessed self-reported measures of dietary intake and biomarkers of nutrient intake. Women were recruited through newspaper advertisements, fliers, and direct mailings addressed by using the Washington State Department of Licensure list. Women were chosen to be representative of women in the Women’s Health Initiative (WHI). Exclusion criteria for the study included: a history of Crohn disease, ulcerative colitis, inflammatory bowel disease, diabetes or hypoglycemia, kidney disease or proteinuria, chronic lung disease, liver disease, or systemic steroid use; loss or gain of >4.5 kg in the 2-mo period before participation; and consumption of 2 servings of alcohol/d (24 g).
Study procedures
Participants visited the Fred Hutchinson Cancer Research Center on 2 occasions that were 1 wk apart. Before the first visit, each participant received a packet of study questionnaires by mail. The packet contained a 20-question soy-specific FFQ, a comprehensive FFQ from the WHI (22), and a Beliefs and Lifestyle Questionnaire, which included the Physical Activity for Elderly questionnaire (23). Participants completed these questionnaires before the first visit, at which time they gave the questionnaires to study personnel and provided a fasting blood sample. Also at the first visit, participants received the second soy FFQ and comprehensive FFQ. They completed these FFQs and returned them to study personnel at the second visit, when another fasting blood sample was obtained. Height and weight were measured at both of the visits. The Institutional Review Board at the Fred Hutchinson Cancer Research Center approved all procedures, and written informed consent was obtained from all participants.
Food-frequency questionnaires and isoflavone content of foods
The soy FFQ takes 5 min to complete and contains questions about the consumption of 20 soy foods and supplements, including traditional soy foods (such as tofu, tempeh, and miso soup), manufactured soy foods (such as vegetable soy burgers and soy cheese), and supplements (such as isoflavone pills and liquid nutrition drinks). The foods were selected after researching soy foods available in a random sample of large and small grocery stores and natural and specialty food stores in Seattle. Participants were asked how frequently each food or beverage was consumed during the past 3 mo, with responses ranging from "never or less than once per month" to "2+ per day" for foods and "6+ per day" for beverages. The size of a medium serving was specified, and participants indicated whether their usual serving was small, medium, or large. A small serving was one-half and a large serving was one and one-half the size of a medium serving. The frequency of intake was multiplied by the usual portion size (0.5 for small, 1.0 for medium, and 1.5 for large) to obtain servings/wk for each of the line items on the soy FFQ. The milligrams of genistein and daidzein per medium serving were multiplied by the servings per week for each line item.
Sources of published data were used to determine the genistein and daidzein content of the soy foods listed on the questionnaire (24–26). For foods with no published isoflavone data available, the isoflavone content was estimated by using known amounts in similar foods. For example, the isoflavone content of soymilk was used as the amount of isoflavones in low-fat soymilk. Recipes were used to determine the isoflavone contents of soy foods without isoflavone information, such as the miso in miso soup. For manufactured foods with the soy content indicated on the labels, this information was used to estimate the amount of isoflavones in these foods. After the development of this soy FFQ, the US Department of Agriculture database of the isoflavone contents of foods was published (27). We used the database to update the isoflavone contents of some foods (ie, those foods for which there was information in the database).
Dietary intake, including isoflavone intake, was assessed with the comprehensive FFQ, which takes 30 min to complete and asks questions about 122 foods or food groups (22). Estimates of isoflavone intake from the comprehensive FFQ were made on the basis of responses about tofu and soymilk. Adjustment questions at the beginning of the FFQ ask about the type of milk usually consumed as a beverage, with cereal, or in coffee or tea; soymilk is included as a response option for each of these questions. Line items in the body of the FFQ ask about frequency and serving size of milk consumed as a beverage, with cereal, or in coffee or tea. In addition, the comprehensive FFQ has one line item asking about frequency and serving size for "tofu and textured vegetable products." Estimated intakes of genistein and daidzein were calculated on the basis of the answers to these questions in the same manner as for the soy FFQ. The reference genistein and daidzein values for tofu and soymilk used for the comprehensive FFQ were the same as those used for the soy FFQ. As with the soy FFQ, the reference period for the comprehensive FFQ was the past 3 mo.
Beliefs and lifestyle questionnaire
This questionnaire asks about the participant’s beliefs regarding the nature and strength of the relation between diet and disease. Questions about race and ethnicity, education, and marital status are included in this questionnaire. The Physical Activity for the Elderly Questionnaire (PASE) was imbedded in the Beliefs and Lifestyle Questionnaire, allowing for the calculation of activity scores for each participant. The PASE assesses the frequency and duration of low-, moderate-, and high-intensity activities during the previous week and has been used in studies of physical activity levels in older adults (28). The scale gives a total score that can range from 0 to >400; higher scores represent higher activity levels.
Plasma genistein and daidzein analysis
Plasma samples (1.0 mL) were frozen and stored at -70°C until they were extracted and analyzed for genistein and daidzein by using liquid chromatography–mass spectrometry as described previously (29). The intraday CV of the assay was <6% for both compounds and the interday CV over 4 mo of analysis was 6–7% for daidzein and 4–5% for genistein. Given the sensitivity of the assay, plasma concentrations <1 ng/mL were considered below the limit of detection and were assigned the value of 0.5 ng/mL (ie, the midpoint between 0 and 1 ng/mL).
Statistical analyses
The amounts of genistein and daidzein obtained from each of the line items on the soy FFQ were summed to determine the total genistein and total daidzein intakes for each participant. Next, the genistein and daidzein intakes were summed across participants to create summary measures of total genistein and daidzein intakes for the entire sample. Genistein and daidzein intakes were also summed across participants for each line item to obtain total genistein and daidzein intakes for each line item. The percentage contribution by each food was calculated by dividing the total genistein or daidzein from each line item by the total genistein or daidzein from all of the line items combined; this value was multiplied by 100.
The distributions of dietary and plasma values for genistein and daidzein were skewed toward higher values; therefore, we performed a natural logarithmic transformation of these values before analysis. Geometric means were calculated as the anti-logarithm of the mean of the log-transformed genistein and daidzein values. To eliminate the possible contribution of learned responses when participants answered questions on the second soy and comprehensive FFQs, only data from the first soy and comprehensive FFQs were used in analyzing isoflavone intakes. For plasma isoflavone concentrations, we used the average of the values from the 2 visits in the analyses.
Intraclass correlation coefficients were calculated to evaluate reliability between week 1 and week 2 FFQ and plasma values. Pearson’s product-moment correlation coefficients were calculated between 1) dietary intake calculated from the soy FFQ and from the comprehensive FFQ, 2) dietary intake calculated from the soy FFQ and plasma isoflavone concentration, and 3) dietary intake calculated from the comprehensive FFQ and plasma isoflavone concentration. Evaluation of bias between the soy FFQ and comprehensive FFQ was conducted with a Bland-Altman analysis (30). In this analysis, correlation of the difference in intake between FFQs and the mean of intake from the FFQs was utilized to assess proportional bias; a t test of the mean differences was used to assess fixed bias. Linear regression models were used to assess the association of linear variables with plasma isoflavone concentrations. Analysis of variance was used to model the association of categorical variables with plasma isoflavone concentrations. Linear regression models with a grouped linear variable were used to evaluate trend across categories. Significance was defined as P < 0.05. Analyses were conducted with Statistical Analysis Software, release 6.12 (SAS Institute Inc, Cary, NC), with the exception of the Bland-Altman analysis (30), which was performed with STATA, version 7 (Stata Corp, College Station, TX).
RESULTS
The mean (±SD) age of the 96 participants was 61 ± 8 y. Most of the participants (94%) were white, 63% of participants had earned at least a Bachelor’s degree, and 57% of participants were married or living as married. Isoflavone intakes assessed by using the soy FFQ and comprehensive FFQ, plasma concentrations, and intraclass correlation coefficients for each measure are shown in Table 1. The range of intakes was wide, but approximately one-half of the participants consumed the equivalent (or less) of one-half serving of tofu or soy milk/wk. On average, plasma concentrations below the detectable limits were measured for genistein in 19 participants (20%) and for daidzein in 31 participants (32%). The reliability of the FFQ estimates and plasma concentrations assessed over a 1-wk period was high, with intraclass correlations for all measures >0.90.
View this table:
TABLE 1 . Dietary intakes and plasma concentrations of the soy isoflavones genistein and daidzein in 96 postmenopausal women1
The contributions of the different line items of the soy FFQ to dietary isoflavone intake and to plasma isoflavone concentration are shown in Table 2. Items that contributed more to the total dietary intake were also likely to be significant predictors of plasma concentration. On the basis of responses to the first administration of the soy FFQ, 26% of women did not consume any of the line items. All line items of the soy FFQ were consumed by 1 participant. On the basis of responses to the first administration of the comprehensive FFQ, 63% of women did not consume soymilk or tofu.
View this table:
TABLE 2 . Intakes of foods (line items) on the soy food-frequency questionnaire (FFQ) and their ability to predict plasma isoflavone concentrations1
Thirty-five women (36%) were classified as soy nonconsumers by the comprehensive FFQ but were soy consumers according to the soy FFQ. No women were classified as nonconsumers by the soy FFQ but classified as soy consumers by the comprehensive FFQ. The geometric mean plasma genistein concentration was 7.9 nmol/L for participants who were classified as consumers by both FFQs (n = 35) and was 4.2 nmol/L for those classified as consumers by the soy FFQ but as nonconsumers by the comprehensive FFQ (n = 35). Women classified as nonconsumers by both FFQs had a mean plasma genistein concentration of 3.6 nmol/L. The mean plasma genistein values of these 3 groups of participants showed a significant increasing trend (P = 0.005).
Daidzein and genistein intakes calculated from the soy and comprehensive FFQs were significantly correlated with plasma daidzein and genistein concentrations, respectively (P < 0.001) (Figure 1). Correlations with plasma concentrations were slightly greater for genistein intake than for daidzein intake for both FFQs.
FIGURE 1. . Log-transformed dietary intakes of daidzein and genistein calculated from the first administration of the soy food-frequency questionnaire (FFQ) and of the comprehensive FFQ were significantly correlated with log-transformed plasma daidzein and genistein concentrations, respectively (P < 0.001), in 96 postmenopausal women aged 50–79 y. For daidzein, r = 0.37 for the soy FFQ and 0.35 for the comprehensive FFQ; for genistein, r = 0.43 for the soy FFQ and 0.38 for the comprehensive FFQ. The linear regression equations are represented by open circles and solid lines for the soy FFQ and by x’s and dashed lines for the comprehensive FFQ; r values are Pearson’s product-moment correlation coefficients. Each plasma concentration was the average of 2 values obtained at 2 visits that were 1 wk apart.
The correlation between differences and means of the FFQs was significant for daidzein and genistein (r = 0.51 and 0.49, respectively; P < 0.001 for both), which was indicative of proportional bias between the 2 FFQs. However, results of the t test for the mean difference of log-transformed values also suggested fixed bias; relative to the comprehensive FFQ, the soy FFQ presented consistently greater values for daidzein (geometric mean difference: 4.1; 95% CI: 3.3, 5.2) and genistein (geometric mean difference: 4.5; 95% CI: 3.6, 5.8). The estimates of genistein and daidzein intake were significantly correlated between the 2 FFQs (r = 0.67 for both, P < 0.001).
As shown in Table 3, older age was associated with higher plasma genistein and daidzein concentrations. Because this finding has not been reported elsewhere, linear regression with reported genistein intake as the outcome was evaluated, and each increasing year of age was associated with a 0.7 mg/wk increase in genistein intake (P < 0.01). Education (level completed or degree earned) was not associated with plasma daidzein (P = 0.34) or genistein (P = 0.39) concentrations. Participants’ beliefs regarding the relation between diet and disease (categorized as strong, moderate, and weak or none) were also not associated with plasma daidzein (P = 0.31) or genistein (P = 0.46) concentrations. The association of smoking could not be assessed because of the low number (n = 2) of current smokers in this group.
View this table:
TABLE 3 . Univariate associations of plasma daidzein and genistein concentrations with participant characteristics and dietary intakes
Several dietary factors were associated with plasma daidzein and genistein concentrations. Intake of caffeine was negatively associated with plasma concentrations of daidzein (P = 0.05) and genistein (P = 0.06) (Table 3). Positive associations with plasma genistein and daidzein concentrations were found for dietary fiber intake (P 0.01) and total servings of fruit and vegetables (P 0.06). Alcohol and fat consumption were not associated with plasma concentrations of daidzein or genistein. Regular use of any dietary supplement during the previous year was positively associated (P 0.01) with plasma concentrations of daidzein and genistein (Table 4).
View this table:
TABLE 4 . Univariate associations of plasma daidzein and genistein concentrations with dietary supplement use1
DISCUSSION
Data are lacking about the determinants of isoflavone intakes and plasma concentrations; these are important components of understanding the associations between isoflavone intakes and disease risk factors. In this study, we showed that in addition to soy intake, other components of diet and subject age are associated with plasma isoflavone concentrations. We also showed that the 20-item soy FFQ and the comprehensive FFQ had excellent reliability over the 1-wk period of measurements. Validity estimates were within the range typically observed for associations of isoflavone intake with plasma or urinary concentrations. Studies comparing intake, as assessed with FFQs or dietary records, with overnight urine or plasma isoflavone concentrations have reported correlation coefficients ranging from 0.31 to 0.54 (31–35). One study that examined the correlation between plasma concentrations and isoflavone intake estimated from soymilk intake found a Spearman correlation >0.70, but this correlation may have been biased toward greater values because subjects were selected on the basis of prior knowledge of soy intake (36).
The soy FFQ gave higher estimates of isoflavone intake than did the comprehensive FFQ. On the basis of the Bland-Altman analysis, the bias between the soy and comprehensive FFQs is most likely proportional; at greater intakes, the difference between the 2 FFQs is larger. The Bland-Altman analysis does not provide information on the validity of either FFQ compared with plasma concentrations.
The range of isoflavone intakes in our population was wide, from 0 to >1050 mg/wk. This range of values is wider than are ranges previously observed in US populations (<1 to 80 mg/wk) (20, 34), but the mean intake is considerably less than that observed in Asian populations. Reported mean genistein intakes in Japan have ranged from 104.3 mg/wk in older adults to 210.7 mg/wk in adult women of various ages; daidzein intakes have ranged from 66.5 to 114.8 mg/wk (32, 33, 37). Most participants in our study had isoflavone consumption that was less than the equivalent of half a serving of tofu or soymilk per week. The number of women classified as soy nonconsumers depended on the number of soy items on the FFQ, suggesting that assessment of only soymilk and tofu intakes may underestimate total soy consumption in this population.
A potential limitation of our study was that approximately one-fifth to one-third of the women had nondetectable concentrations of isoflavones, and approximately one-quarter were soy nonconsumers. However, this did not limit our ability to detect significant correlations between each FFQ and plasma concentrations. Inclusion of women with a wide range of intakes and plasma concentrations allowed for adequate assessment of the linear relation between isoflavone intake, as assessed with the FFQs, and plasma isoflavone concentration. Furthermore, the women in this sample were chosen to reflect the women in the WHI, and they may reasonably represent a population of healthy postmenopausal women in the United States.
Isoflavone intake is an important consideration when evaluating the association between isoflavone intake and disease. Among intervention studies that used soy protein, results suggest that relatively higher doses are required to elicit physiologic changes. In a study of hypercholesterolemic postmenopausal women, consumption of isolated soy protein containing 90 mg isoflavones/d was associated with a significant increase in spinal bone density after 24 wk (15). No significant change was observed for women consuming the control diet or women consuming a diet that provided 56 mg isoflavones/d. In a randomized crossover trial of 3 isoflavone doses (7, 65, and 132 mg/d) as isolated soy protein, dose-dependent reductions in LDL were observed (19). The ratio of LDL to HDL cholesterol was lower in participants consuming the soy intervention diets than in participants consuming the control diet, but the soy diets were not different from each other, showing that isoflavone dose may affect an aspect of cholesterol metabolism. Overall, such results emphasize the importance of using caution when interpreting associations between isoflavone intake and disease or disease risk factors in populations with low soy consumption.
Observational studies have found conflicting results regarding associations between isoflavone intake and disease risk. In a study of postmenopausal Japanese women whose average isoflavone intake was 54.3 mg/d, no differences were observed across intake groups for concentrations of cholesterol, triacylglycerol, or apolipoproteins (16). However, a significant positive association of genistein with HDL cholesterol concentrations and a negative association with insulin were reported in a study of postmenopausal women in California in which the average intake of genistein was 1.3 mg/d and the highest observed intake was 13.9 mg/d (14). Similarly, there were inverse associations of total isoflavone intake with triacylglycerol concentrations and metabolic syndrome score in postmenopausal women in the Framingham Study; the median isoflavone intake was <0.5 mg/d (17).
Healthy lifestyle behaviors may cluster together. For example, Tseng and DeVellis (38) observed that people who consume a vegetable-fruit pattern diet exercised more frequently, used less salt at the table, and were less likely to be smokers than were people who consumed a meat-starch pattern diet. In our study, genistein and daidzein intakes were inversely associated with caffeine intake and positively associated with fruit and vegetable and fiber intakes. Few women were smokers or reported that they believed there was a weak relation or no relation between diet and disease; this limited our ability to detect associations of these characteristics with isoflavone intake. Factors such as smoking or beliefs regarding the relation between diet and disease may be important correlates of isoflavone intake in other populations, such as men or younger women. Age was also positively associated with plasma genistein and daidzein concentrations. This finding was not reported in other studies of soy intake in older women. In the current study, because isoflavone intake was positively associated with age, it is likely that the older women consumed more soy. However, altered renal clearance of isoflavones also may have contributed to higher plasma concentrations in these women. The amount of fluid consumed and renal function were not assessed in this study.
Overall, the soy FFQ and comprehensive FFQ both showed excellent reliability and moderate validity in our population of postmenopausal women. In this population, the soy FFQ captured more soy consumers than did the comprehensive FFQ. Although intakes measured with the soy FFQ were as high as 1050 mg isoflavones/wk, the median intake was <2 mg/wk, suggesting low intake overall in this group. Data from clinical trials suggest that isoflavone doses much higher than 2 mg/wk are necessary to achieve physiologic effects. Therefore, it is important to consider that in populations in which isoflavone intakes and plasma concentrations are low, these indicators of intake may be functioning as markers of other healthy behaviors.
ACKNOWLEDGMENTS
We express our sincerest thanks to Margaret T Grate and Tamsen Johnston for their assistance with the participant visits and materials.
The authors’ contributions to this study were as follows: study design, REP and JWL; data collection, NKH and MLN; laboratory and data analysis, CLF, REP, HES, TFK, WNH, and JWL; and writing of the manuscript, CLF, REP, JWL, and NKH. None of the authors have any financial or personal interest, including advisory board affiliations, in any company or organization that sponsored the research.
REFERENCES