QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Chan, S. G. Articles by Chong, P. Y. Y. Search for Related Content PubMed PubMed Citation Articles by Chan, S. G. Articles by Chong, P. Y. Y.

---

Nutritional Epidemiology

Validation of a Food Frequency Questionnaire for Assessing Dietary Soy Isoflavone Intake among Midlife Chinese Women in Hong Kong^1,2

³ Department of Community and Family Medicine, Chinese University of Hong Kong, Hong Kong SAR; ⁴ Department of Public Health Sciences, University of Toronto, M5G 2L7 Toronto, Canada; and ⁵ Department of Mathematics and Statistics, University of Guelph, N1G 2W1 Guelph, Canada

^* To whom correspondence should be addressed. E-mail: sieugaen2003{at}yahoo.com.

	ABSTRACT

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

 
Substantial evidence suggests that soy isoflavones may protect against certain chronic diseases. This study aims to assess the reproducibility and validity of a 47-item semiquantitative soy FFQ (SFFQ) designed to measure the usual intake of soy isoflavones among 145 Chinese mid-life women in Hong Kong. Reproducibility of the SFFQ was assessed by the stability of dietary intake obtained at baseline (SFFQ₀) and at 13-mo follow-up (SFFQ₁). Relative validity was measured by comparing data collected from SFFQ₁ with those derived from 23-d, 24-h dietary recalls (DR) collected during the same 1-y validation period. Isoflavone intake was calculated using analytical values from the Chinese University of Hong Kong Soy Isoflavone Database. A total of 3,217 DR were successfully obtained. The median (interquartile range) absolute intake estimated by the SFFQ₁ was 0.91 (–2.7 to 6.8) mg aglucon equivalents/d higher than the 6.3 (3.7–10.7) mg aglucon equivalents/d measured by the DR (P < 0.0057; Wilcoxon's Signed Rank test). Bland-Altman analysis further demonstrated the presence of significant proportional bias between methods among Cantonese women with above-median intake (Spearman correlation coefficient; r = 0.44; P = 0.0005). Nonetheless, the intraclass and Pearson correlation coefficients, respectively, were 0.84 and 0.72 for non-Cantonese and 0.65 and 0.50 for Cantonese, showing moderate to good levels of reproducibility and validity of the SFFQ (difference between 2 intraclass correlation coefficients, P = 0.09; difference between 2 Pearson r, P = 0.16). The unadjusted and BMI-adjusted correlations were of similar magnitude. The SFFQ is a reasonably valid instrument for assessing dietary soy isoflavone exposure in Hong Kong Chinese mid-life women.

	Introduction

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

Soy is a traditional food among Asians and the variations of soy intake would allow studies to assess the relations between soy isoflavone exposures and the potential health effects in these populations. The FFQ for estimating the usual intake of an individual over a specified period is currently the most widely used dietary assessment method for investigating diet and disease relationship.

Several analytic epidemiological studies recently carried out among Chinese populations have used FFQ for the assessment of dietary soy exposure (5–8). However, these FFQ contained a limited number of soy products or were originally developed for the estimation of other nutrients, such as calcium intake (6). Moreover, the estimation of soy isoflavone content was based on published databases from the West, as a local database was not available.

Thus, the development and validation of a FFQ for quantifying soy isoflavone exposure would form the basis for further research into the potential health effects of soy isoflavone intake. This article reports on the development of a semiquantitative soy FFQ (SFFQ)⁶ for the evaluation of soy isoflavone intake over a 12-mo period among Chinese midlife women in Hong Kong and assesses its validity and reproducibility using 23-d, 24-h dietary recalls (DR) over the same 12 mo as the reference method.

	Materials and Methods

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

 
Study population

Subjects were participants of the Hong Kong Perimenopausal Osteoporosis Study (HKPOST), an ancillary study to the 1996 cross-sectional survey of "Women's Health at Midlife." Descriptions of the study design, sampling methods, and selection criteria for research participants have previously been reported (9). Briefly, 1903 women aged 45–55 y were recruited between June 1994 and March 1996 through random dialing of the directory-listed residential telephone numbers. Eligible participants from this survey and from the University Family Medicine Clinic were invited to join the HKPOST. Women aged 45–55 y who were free from medical conditions and were not taking medications known to affect bone metabolism were considered eligible. Among the 949 eligible women identified from these sources, 438 women joined the HKPOST with a response rate of 46%. Their characteristics were largely similar to those of the nonparticipants. Assessments of bone mineral density were made at baseline and at 9-, 18-, 30-, 47-, 74-, and 86-mo follow-up.

In July 2001, 247 members who had completed the HKPOST 47-mo follow-up study in 1999 were invited by letter followed by telephone calls to participate in the soy questionnaire validation study. Up to 9 attempts at different times and days of the week were made to reach a potential participant. Of the 247 HKPOST members, the first 145 women who agreed to participate were enrolled in the validation study between September 2001 and February 2002. The validation study was approved by the Health Sciences I Research Ethics Board of the University of Toronto and the Ethical Committee of the Chinese University of Hong Kong. Written informed consent of the validation study was obtained from subjects for their participation in the study.

Dietary assessment

    Twenty-four-h DR (reference method). Description of the 24-h DR collection method has been reported in an earlier article (9). Two nonconsecutive, unannounced 24-h DR, randomly selected within a 1-mo interval over a 12-mo period, were obtained from all participants from October 2001 to November 2002. Both weekdays and weekend days were included as recall days to take into account dietary intake pattern throughout the whole week. Sixteen weekdays and 7 weekend days were included for each participant. The first DR was administered 2 wk after the baseline soy food frequency questionnaire (SFFQ₀) by trained interviewers using the USDA multiple pass technique (10). Details about each reported food were gathered using standardized food probes listed in the Food Instruction Booklet for the Continuing Survey of Food Intakes by Individuals 1998 (11).

    Development of SFFQ (test method). Selection of food items The SFFQ was designed to assess the habitual dietary intake of soy isoflavones in midlife women. As the soy isoflavone database for local foods has not yet been developed, the food items and range of portion sizes selected for the SFFQ were first developed using dietary soy protein intake data from the China Health and Nutrition Survey (CHNS) conducted in 1993 (12). Based on 3-d, 24-h DR data collected from 2643 female CHNS participants aged 40–60 y, food items that contributed to 95% of total population intake of soy protein were identified by contribution analysis (13). Estimates of the expected contribution of individual soy products were obtained primarily from the 1991 Food Composition Table for China (14,15). The initial SFFQ contained 13 food items. To ensure comprehensiveness of food items to be included, data from earlier local dietary surveys were also examined (5,6,8,16). In addition, local "wet markets" and supermarkets were visited and regional food composition tables (17–19) and cookbooks were reviewed to identify soy products commonly consumed by the local Chinese or dialect subgroups in Hong Kong. Thirty-eight additional items were identified and subsequently added to the initial food list of 13, yielding a SFFQ consisting of 51 soy food products.

Development of portion size Respondents chose from a selection of 3 portion sizes, small, medium, or large, to describe the intake quantity for each of the food items contained in the SFFQ. The quantity of food for the 3 portion sizes was derived using the 25th, 50^th, and 75th percentiles for the specific food from the 1993 CHNS survey data as cutoff values (20). The derived portion sizes for foods included in the SFFQ were reviewed and modified by one coauthor (S. C. Ho) experienced in conducting nutritional epidemiologic studies in the local population. Color photographs of the 51 foods, each depicted in the respective portion sizes, were used to assist participants to better estimate the quantity of foods consumed. Participants were asked about their usual intake of the soy food during the 1-y period preceding the date of interview. Seasonal variation of soy food intake was also taken into account by asking the participants to report intake frequency during both in-season and off-season periods for each item. Frequency of consumption was recorded as never, or number of times per day, week, month, or year, and portion size equivalent to either small, medium, or large.

    Pretesting the SFFQ. The draft SFFQ was pretested on 20 women recruited from women's health talks or by word of mouth. Using a retrospective, think-aloud technique developed in cognitive science (21), the first 10 volunteer participants were instructed to think back about how they comprehended and interpreted the terms "soy products" and "seasonal foods." Participants were also asked to explain the recall strategies used in their estimation of consumption frequency and the amount of 6 soy products (2 solid, liquid, and amorphous) eaten over a 6-mo period prior to the interview. We adjusted the wording of some questions to improve the ease of understanding. The revised SFFQ was pretested again on another 10 women of similar age range. No further changes were made and the revised SFFQ was tested on another sample of 39 women aged 40–60 y to evaluate their ability to answer the questions and the overall flow of the questionnaire. Participants for this final stage of the pretesting were identified from a sample of 2000 addresses from the list of housing quarters maintained by the Hong Kong Census and Statistics Department. Soy products not consumed by participants were excluded (n = 4) from the final SFFQ. The final SFFQ contained 47 soy products.

    Administration of the SFFQ. The 47-item SFFQ was administered in person by trained interviewers at baseline (SFFQ₀). A set of visual aids including a page-numbered food photo album and a set of common household measures was given to participants at the completion of the SFFQ₀ to take home for reference for the 24-h DR. Another (SFFQ₁) was administered in person at 13 mo, which was 1 mo after the completion of the last DR.

    Validation procedures. Reproducibility of SFFQ was assessed by the stability of dietary intake obtained at SFFQ₀ and at SFFQ₁. The relative validity was assessed by comparing soy isoflavone data collected from SFFQ₁ against the estimated amount from the 23-d, 24-h DR during the same 1-y validation period.

Statistical analysis

    Data processing. We checked the FFQ and DR for completeness prior to food coding and data entry for analysis. Participants with 12 DR (n = 4) were excluded, leaving a total of 141 participants with 3217 DR available for analysis.

    Isoflavone content of foods. The isoflavone intake for each food was calculated by multiplying the reported frequency of consumption of specific food by the isoflavone content of the specified portion. Values for isoflavone content in foods (mg aglucon equivalents/100 g) were obtained from the CU Soy Isoflavone Database (S. Chan, P. Murphy, S. Ho, N. Kreiger, G. Darlington, K. So, P. Chong, unpublished data). Intake values were then summed across all foods to obtain the total intake of isoflavones for each individual.

    Statistical methods. Median and interquartile range for isoflavone intakes were computed from both FFQ and from the means of the 23-d, 24-h DR. All other summary statistics are expressed as means ± SD or number and percentage, unless otherwise stated. Because the distributions of isoflavone intakes were right-skewed, the Box-Cox power transformation, y = (x – 1)/, was carried out to improve distribution normality prior to formal analysis. The values for SFFQ₀, SFFQ₁, and DR were 0.3, 0.3, and 0.4, respectively.

Proportions were compared with the chi-square test of contingency. Differences between groups were performed using the Mann-Whitney U test. Within-subject comparison was examined using the Wilcoxon's Signed Rank test.

Agreement between the 2 dietary intake methods was further evaluated using the graphical method described by Bland and Altman (22). With this method, the arithmetic difference in isoflavone intakes between the 2 methods for each individual was plotted against the mean values of the 2 methods. To examine whether the differences between methods varied systematically with the means obtained from the 2 methods, Spearman rank correlation procedures were performed. Limits of agreement were established defining the range or interval within which most of the differences between the methods are expected to lie. Because the Bland-Altman plots indicated the differences between methods were intake dependent, and log transformation, as suggested by Bland and Altman (23), could not correct the magnitude of differences, 95% limits of agreement were therefore derived based on the regression method for nonuniform differences.

The ability of the 2 dietary intake methods to rank subjects into the same quartile was also examined. Isoflavone values obtained from the SFFQ₁ and through the 24-h DR were divided into quartiles using cut-off values derived separately for each dietary assessment method. The proportion of subjects correctly classified into the same quartile, and misclassified into extreme quartiles, was examined.

To evaluate the reproducibility of the SFFQ administered on 2 separate occasions 13 mo apart, an intraclass correlation coefficient (ICC_1,1) was computed using a 1-way random effects ANOVA for a single measurement (24,25). The ICC, defined as a ratio of between-person variance to total variance, ranges from 0 to 1. An ICC 0.76 indicates excellent test-retest reliability and an ICC <0.4 represents poor test-retest reliability (26).

Pearson's product-moment correlation coefficient was computed to quantify the extent of linear association between power-transformed isoflavone intakes derived from the SFFQ₁ and the mean values derived from the 23-d, 24-h DR. Partial correlation coefficient adjusted for BMI was also calculated by multiple linear regression. To correct for the attenuation of correlation coefficient attributable to within-person variation in DR, de-attenuated correlation coefficient was calculated using the formula r_c = r_o {1+[(S²_w / S²_B/n)]^1/2}(27), where r_c is the true correlation, r_o is the observed correlation between power-transformed isoflavone intakes estimated by the SFFQ₁ and 24-h DR, S²_w/S²_b is the ratio of the within- and between-person variances, and n is the number of 24-h DR per subject. Subgroup correlations were also computed, stratifying participants by demographic and anthropometric characteristics. To test the equality of correlations, each subgroup correlation coefficient was first transformed into a Z score using the Fisher's r-to-Z transformation (25). The derived Z scores of 2 and several correlation coefficients were then compared, respectively, using the Z (28) and chi-square formulas (29).

The Statistical Analysis System for Windows (version 6.12, SAS Institute) was used to perform all statistical analyses. All tests were 2-sided and P < 0.05 was considered significant.

	Results

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

 
The baseline demographic characteristics of the participants in this study were similar to those from the original cohort (Table 1). About one-third had only some primary education or below. The majority were married and approximately one-half were housewives.

The Bland-Altman plots for isoflavone intakes for the entire sample (Fig. 1A), Cantonese (Fig. 1B), and non-Cantonese (Fig. 1C) showed the individual between-method differences tended to increase with increasing intake of soy isoflavones, with magnitude of arithmetic differences greatest at high intake and smallest at low intake levels. In addition, the differences between methods were strongly and positively associated with the means obtained from the 2 methods. The Spearman rank correlation coefficients for the aforementioned groups were r = 0.40 (P < 0.0001), r = 0.45 (P = 0.0005), and r = 0.003 (P = 0.9066), respectively. The data suggest the presence of proportional bias (between methods) in the Cantonese dialect group and such bias may have arisen from 1 or both assessment methods. The 95% limits of agreement derived from the regression method indicated 4% of individual differences were spread above or below the lower 95% limit of agreement. Nonetheless, individual data revealed that the limits of agreement between the 2 methods were fairly wide.

View larger version (9K): [in this window] [in a new window]   FIGURE 1  Difference in isoflavone intake estimated with the SFFQ and 24-h DR plotted against the mean of isoflavone intake measured with the 2 methods for the entire study population (n = 141) (A), Cantonese (n = 120) (B), and non-Cantonese (n = 21) (C). The solid lines represent the 95% limits of agreement.

In this study, the overall Pearson correlation coefficient between the SFFQ₁ and 24-h DR for isoflavone intakes of all subjects was 0.53 (P < 0.0001) (Table 2). Cross-classification analysis further revealed that the observed levels of agreement and disagreement between the 2 methods were higher and lower than what might be expected by chance (25 and 12.5%, respectively). Thirty-five percent of the participants were classified into similar quartiles and only 4.3% were grossly misclassified into the extreme quartiles. Because 60% of the 24-h DR had zero consumption of soy isoflavones and Box-Cox transformation could not normalize the distribution of isoflavone values, the ratio of within- to between-person variance in isoflavone intake derived from the 24-h DR could not be calculated. Failure of the data to follow a normal distribution further precluded the calculation of the de-attenuated correlation coefficient, which when corrected for attenuation due to day-to-day fluctuations in individuals would provide a more accurate estimate of the correlation between the SFFQ₁ and multiple DR.

	Discussion

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

 
This study examined the relative validity and reproducibility of a 47-item SFFQ in estimating the habitual isoflavone intake in mid-life Chinese women. A correlation of 0.53 of isoflavone intakes derived from the SFFQ and the 23-d, 24-h DR was found and the results compared favorably with that of other studies using self-reported dietary assessment as the reference methods (30–32). The ICC of the 2 SFFQ measured at 1-y intervals (value = 0.68) is also similar to studies that examined reproducibility of food and nutrient intakes within periods of 1–10 y (33) but considerably lower than those reported in some recent soy validation studies based on shorter durations. Frankenfeld et al. (34) reported an ICC of 0.98 in a 1-wk study, whereas Kirk et al. (35) reported a 2-wk ICC of 0.89 and Rice et al. (36) reported an ICC of 0.72 for a study over 1–4 mo. The relatively short time intervals between the repeated FFQ measurements and thus the lesser variability of dietary intake may explain the fairly high reported correlation values in these studies.

The study found differences in the validity of SFFQ among the 2 dialect subgroups. The regularity of food habits and homogeneity in soy consumption might have contributed to a higher measure of correlation coefficient among the non-Cantonese (9). Nonetheless, this notion of a more consistent diet among the non-Cantonese was not supported by the 24-h DR, because the variation in isoflavone intake (measured as the semiinterquartile range divided by median) was higher in the non-Cantonese than in the Cantonese. Closer scrutiny revealed that the non-Cantonese were better educated than the Cantonese and education has previously been shown to be related to better consistency of recalls (33).

The DR method provides snapshots of dietary intake at different time points during the reference period. The number of days of 24-h DR needed to correctly rank individuals varies with nutrients is a function of the desired correlation between observed and true nutrient intakes and the ratio of within- to between-subject variances (37). Our data revealed substantial within-person or day-to-day variations in isoflavone intakes among the study participants. The within- to between-subject variance ratio (Sw²/Sb²), obtained by the SAS procedure PROC VARCOMP for non-normally distributed isoflavone data, was 9.0. The number of days, therefore, needed to obtain a correlation of 0.8 between the 24-h recall-derived mean intake and the true intake would be 16 d [d = (r²/1 – r²) (Sw²/Sb²) = (0.82/1–0.82) (9.0) = 16] (37). Thus, this study has obtained sufficient number of 24-h DR from each participant to achieve a ranking of acceptable accuracy.

Our data revealed that the SFFQ yielded a higher median isoflavone intake value than that obtained from the 24-h DR. A validation study by Yamamoto et al. (30) also noted a significantly higher mean absolute intake estimated by the FFQ method than that obtained from the 24-h DR. Possible explanations include an overestimation of the portion sizes assigned to food items in the SFFQ leading to a higher reported amount than that actually consumed by the participants. A number of studies (38,39) have also noted that increasing the number of food items in the FFQ might result in an overestimation of the intake, whereas combining several nutritionally similar food items into single items could lead to an underestimation. Our SFFQ, comprising 47 individual food items, is an extensive questionnaire containing soy products commonly consumed by various Chinese dialect groups in Hong Kong. The comprehensiveness of the SFFQ could thus also give rise to a higher reported consumption of soy isoflavones.

Epidemiological studies of diet and disease relationship seek to compare disease risks across categories of intake. Although consistent over- and underestimation of intake by the FFQ method would not alter the rankings of individuals along the distribution of intake, differential misclassification could lead to inaccurate rankings and erroneous exposure estimates in diet-disease relations (40,41). In this study, despite a moderate positive correlation between the dietary assessment methods, the Bland-Altman plot showed that the magnitude of between-method differences was not consistent across the range of intake for individual participants. Further analysis of the data indicated that the degree of agreement was strongly associated with the dialect grouping of the participants and a bias might possibly exist among the Cantonese dialect group with an above-median dietary soy intake.

The SFFQ was designed to assess an individual's usual intake during the past year whereas the 24-h DR capture snapshots of actual intakes at different time points over the reference period. The marked discrepancy between estimates derived from the 2 assessment methods in participants with high soy intake could be due to the insufficient number of DR or the inability of participants with high intake to accurately recall what they had consumed. This finding thus suggests that bias might exist in the risk estimates at certain levels of exposures and caution should be taken in using these results.

Due to limited funding, our study had not incorporated the measurement of biomarkers. Biomarkers of diet do not rely on the subject's reporting and possess measurement errors that are not likely to be correlated with those of the self-reported dietary assessment methods (33,42). Lee et al. (43), in a dietary isoflavone assessment validation study among men in Shanghai, China, compared the validity of a 8-item SFFQ with 4 spot urine samples collected over a 12-mo period and found a Spearman correlation coefficient of 0.48. The ICC reproducibility value for isoflavonoid levels measured in spot urine samples was 0.39. These findings demonstrate that measurements sampled at multiple time points can yield acceptable reproducibility and urinary isoflavones can serve as a time-integrating biomarker reflecting longer-term intakes (44). Thus, future validation studies of self-reported dietary assessment methods should also consider the incorporation of dietary biomarkers to compare and evaluate the validity of the findings.

Our study has several limitations. We had not adjusted the isoflavone intake values for total dietary energy intake, because the nutrient composition data for local foods were unavailable. Energy adjustment would presumably improve the correlation coefficient between estimates from 2 dietary assessment methods (33,45), because it removes variations in nutrient intake due to different energy requirements among individuals and also minimizes measurement errors from under- or over-reporting of energy intake in FFQ. When the crude validity coefficient in our study was adjusted for BMI (an indirect measure of total energy intake), the unadjusted and BMI-adjusted correlations were similar (0.5220 and 0.5239, respectively), suggesting that the variation in isoflavone intake is not related to variation in dietary energy intake.

Also, because of the highly skewed distribution with a high frequency of zero intake values in the DR, the de-attenuated correlation coefficient was not calculated. Though a high degree of reproducibility was observed, particularly among the non-Cantonese, the level of agreement between the 2 assessment methods varied with a relatively wide scatter of differences observed among the Cantonese. The Bland-Altman plot indicated that individual differences between methods were dependent on intake levels, with better agreement at the lower levels of intake. Thus, more work is required to establish the presence of differential bias in dietary intake estimates derived from the SFFQ in population subgroups under investigation. Moreover, as the sample size of the non-Cantonese subjects was rather small, additional research is warranted to examine the effectiveness of the SFFQ in larger subgroups of non-Cantonese populations.

Despite these limitations, this SFFQ has been validated against 3217 DR with very low attrition from the study participants. It had acceptable validity and reproducibility and is thus a reasonable instrument for assessing dietary soy iosflavone exposure in Hong Kong Chinese mid-life women.

	FOOTNOTES

 
¹ Supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (project no. CUHK 4047/01M).

² Author disclosures: S. G. Chan, S. C. Ho, N. Kreiger, G. Darlington, E. M. Adlaf, K. F. So, and P. Y. Y. Chong, no conflicts of interest.

⁶ Abbreviations used: CHNS, China Health and Nutrition Survey; CU, the Chinese University of Hong Kong; DR, dietary recall; HKPOST, Hong Kong Perimenopausal Osteoporosis Study; ICC, Intraclass correlation coefficient; SFFQ, soy FFQ; SFFQ₀, baseline soy food FFQ; SFFQ₁, 13-mo follow-up soy FFQ.

Manuscript received 10 April 2007. Initial review completed 29 May 2007. Revision accepted 27 November 2007.

	LITERATURE CITED

TOP ABSTRACT Introduction Materials and Methods Results Discussion LITERATURE CITED

 

1. Setchell KDR, Adlercreutz H. Mammalian lignans and phyto-oestrogens. Recent studies on their formation, metabolism and biological role in health and disease. In: Rowland IA, editor. The role of gut microflora in toxicity and cancer. New York: Academic Press; 1988. p. 315–45.

2. Committee on Toxicity of Chemicals in Food/Consumer Products and the Environment. COT report: phytoestrogens and health. London: Food Standards Agency; 2003.

3. Cassidy A, Albertazzi P, Lise Nielsen I, Hall W, Williamson G, Tetens I, Atkins S, Cross H, Manios Y, et al. Critical review of health effects of soyabean phyto-oestrogens in post-menopausal women. Proc Nutr Soc. 2006;65:76–92.[Medline]

4. Balk E, Chung M, Chew P, Ip S, Raman G, Kupelnick B, Tatsioni A, Sun Y, Wolk B, et al. Effects of soy on health outcomes. Evidence report/technology assessment no. 126. Prepared by Tufts-New England Medical Center Evidence-Based Practice Center under contract no. 290–02–0022. AHRQ publication no. 05–E024–2. Rockville (MD): Agency for Healthcare Research and Quality; August 2005.

5. Ho SC, Woo J, Leung SSF, Sham A, Lam TH, Janus ED. Intake of soy products are related to better plasma lipid profiles in the Hong Kong Chinese population. J Nutr. 2000;130:2590–3.[Abstract/Free Full Text]

6. Ho SC, Chan SG, Yi Q, Wong E, Leung PC. Soy intake and the maintenance of peak bone mass in Hong Kong Chinese women. J Bone Miner Res. 2001;16:1363–9.[Medline]

7. Mei J, Yeung SS, Kung AW. High dietary phytoestrogen intake is associated with higher bone mineral density in postmenopausal but not premenopausal women. J Clin Endocrinol Metab. 2001;86:5217–21.[Abstract/Free Full Text]

8. Ho SC, Woo J, Lam S, Chen Y, Sham A, Lau J. Soy protein consumption and bone mass in early postmenopausal Chinese women. Osteoporos Int. 2003;14:835–42.[Medline]

9. Chan SG, Ho SC, Kreiger N, Darlington G, So KF, Chong PYY. Dietary sources and determinants of soy isoflavone intake among midlife Chinese women in Hong Kong. J Nutr. 2007;137:2451–5.[Abstract/Free Full Text]

10. Guenther PM, Cleveland LE, Ingwersen LA. Questionnaire development and data collection procedures. In: Tippett KS, Cypel YS, editors. Design and operation: the Continuing Survey of Food Intakes by Individuals and the Diet and Health Knowledge Survey, 1994–96. USDA, Agricultural Research Service, Nationwide Food Surveys report no. 96–1; 1997 [cited 2005 Aug 31]. Available from: http://www.ars.usda.gov/SP2UserFiles/Place/12355000/pdf/DORch_4.pdf.

11. USDA Agricultural Research Service. 1994–6 Continuing Survey of Food Intakes by Individuals (CSFII): food instruction booklet. 1998 [cited 2005 Aug 31]. Available from: http://www.ars.usda.gov/SP2UserFiles/Place/12355000/pdf/FIB.pdf.

12. Zhai F, Guo X, Popkin BM, Ma L, Wang Q, Yu W, Jin S, Ge K. The evaluation of the 24-hour individual recall method in China. Food Nutr Bull. 1996;17:154–61.

13. Block G, Dresser CM, Hartman AM, Carroll MD. Nutrient sources in the American diet: quantitative data from the NHANES II survey. I. Vitamins and minerals. Am J Epidemiol. 1985;122:13–26.[Abstract/Free Full Text]

14. Popkin BM, Lu B, Zhai FY. Understanding the nutrition transition: measuring rapid dietary changes in transitional countries. Public Health Nutr. 2002;5:947–53.[Medline]

15. China CDC. Food Nutrition Components Table. Beijing: People's Medical Publishing House. 1991.

16. Leung SSF, Ho S, Woo J, Lam TH, Janus ED. Hong Kong Adult Dietary Survey 1995. Hong Kong: Chinese University of Hong Kong; 1997.

17. Wang G, Parpia B, Wen Z. The composition of Chinese foods. Washington, DC: ILSI Press; 1997.

18. Holland B, Welch AA, Unwin ID, Buss DH, Paul AA, Southgate DAT. McCance and Widdowson's the composition of foods. 5th rev ed. Cambridge (UK): The Royal Society of Chemistry and Ministry of Agriculture, Fisheries and Food; 1991.

19. Tsang YS, Fung SW. Diet and health guide and composition of food in Southern China. Guang Zhou (China): Zhongshan University Press; 1991.

20. Carolina Population Center at the University of North Carolina at Chapel Hill. China Health and Nutrition Survey. Version current 16 August 2005 [cited 2005 Aug 31]. Available from: https://www.cpc.unc.edu/projects/china/data/datasets/data_downloads.

21. Sudman S, Bradburn NM, Schwarz N. Thinking about answers: the application of cognitive processes to survey methodology. San Francisco: Jossey-Bass Publishers; 1996.

22. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–10.[Medline]

23. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8:135–60.[Abstract/Free Full Text]

24. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420–8.[Medline]

25. McGraw KO, Wong SP. Forming inferences about some intraclass correlation coefficients. Psychol Methods. 1996;1:30–46.

26. Fleiss JL. Statistical methods for rates and proportions. New York: Wiley; 1981.

27. Beaton GH, Milner J, Corey P, McGuire V, Cousins M, Stewart E, de Ramos M, Hewitt D, Grambsch PV, et al. Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation. Am J Clin Nutr. 1979;32:2546–59.[Free Full Text]

28. Cohen J, Cohen P. Applied multiple regression/correlation analysis for the behavioral sciences. Second edition. Hillsdale (NJ): Lawrence Erlbaum Associates; 1983.

29. Steel RD, Torrie JH, Dickey TA. Principles and practice of statistics: a biomedical approach. New York: McGraw Hill; 1997.

30. Yamamoto S, Sobue T, Sasaki S, Kobayashi M, Arai Y, Uehara M, Adlercreutz H, Watanabe S, Takahashi T, et al. Validity and reproducibility of a self-administered food-frequency questionnaire to assess isoflavone intake in a Japanese population in comparison with dietary records and blood and urine isoflavones. J Nutr. 2001;131:2741–7.[Abstract/Free Full Text]

31. Williams AE, Maskarinec G, Hebshi S, Oshiro C, Murphy S, Franke AA. Validation of a soy questionnaire with repeated dietary recalls and urinary isoflavone assessments over one year. Nutr Cancer. 2003;47:118–25.[Medline]

32. Huang MH, Harrison GG, Mohamed MM, Gornbein JA, Henning SM, Go VL, Greendale GA. Assessing the accuracy of a food frequency questionnaire for estimating usual intake of phytoestrogens. Nutr Cancer. 2000;37:145–54.[Medline]

33. Willett W. Nutritional epidemiology. 2nd ed. New York: Oxford University Press; 1998.

34. Frankenfeld CL, Patterson RE, Horner NK, Neuhouser ML, Skor HE, Kalhorn TF, Howald WN, Lampe JW. Validation of a soy food-frequency questionnaire and evaluation of correlates of plasma isoflavone concentrations in postmenopausal women. Am J Clin Nutr. 2003;77:674–80.[Abstract/Free Full Text]

35. Kirk P, Patterson RE, Lampe J. Development of a soy food frequency questionnaire to estimate isoflavone consumption in US adults. J Am Diet Assoc. 1999;99:558–63.[Medline]

36. Rice MM, LaCroix AZ, Lampe JW, van Belle G, Kestin M, Sumitani M, Graves AB, Larson EB. Dietary soy isoflavone intake in older Japanese American women. Public Health Nutr. 2001;4:943–52.[Medline]

37. Nelson M, Black AE, Morris JA, Cole TJ. Between- and within-subject variation in nutrient intake from infancy to old age: estimating the number of days required to rank dietary intakes with desired precision. Am J Clin Nutr. 1989;50:155–67.[Abstract/Free Full Text]

38. Thompson FE, Kipnis V, Subar AF, Krebs-Smith SM, Kahle LL, Midhune D, Potischman N, Schatzkin A. Evaluation of 2 brief instruments and a food-frequency questionnaire to estimate daily number of servings of fruit and vegetables. Am J Clin Nutr. 2000;71:1503–10.[Abstract/Free Full Text]

39. Smith-Warner SA, Elmer PJ, Fosdick L, Tharp TM, Randall B. Reliability and comparability of three dietary assessment methods for estimating fruit and vegetable intakes. Epidemiology. 1997;8:196–201.[Medline]

40. Beaton GH. Interpretation of results from diet history studies. In: Kohlmeier L, editor. The diet history method. London: Smith-Gordon; 1991.

41. MacIntyre UE, Venter CS, Vorster HH, Steyn HS. A combination of statistical methods for the analysis of the relative validation data of the quantitative food frequency questionnaire used in the THUSA study. Transition, Health and Urbanisation in South Africa. Public Health Nutr. 2001;4:45–51.[Medline]

42. Ocke MC, Kaaks RJ. Biochemical markers as additional measurements in dietary validity studies: application of the method of triads with examples from the European Prospective Investigation into Cancer and Nutrition. Am J Clin Nutr. 1997;65 Suppl:S1240–5.[Medline]

43. Lee SA, Wen W, Xiang YB, Barnes S, Liu D, Cai Q, Zheng W, Shu XO. Assessment of dietary isoflavone intake among middle-aged Chinese men. J Nutr. 2007;137:1011–6.[Abstract/Free Full Text]

44. Willett WC. The use of biomarkers in nutritional epidemiology. In: Kok FJ, van't Veer P, editors. Proceedings of the 3rd meeting on nutritional epidemiology "biomarkers of dietary exposure." Rotterdam (The Netherlands): Smith-Gordon; 1991.

45. Flegal KM. Evaluating epidemiologic evidence of the effects of food and nutrient exposures. Am J Clin Nutr. 1999;69 Suppl:S1339–44.

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Chan, S. G. Articles by Chong, P. Y. Y. Search for Related Content PubMed PubMed Citation Articles by Chan, S. G. Articles by Chong, P. Y. Y.