Reliability and validity of an assessment of usual phytoestrogen ... - UAB

2 downloads 66 Views 161KB Size Report
P. L. Horn-Ross (&) Ж V. S. Lee Ж C. N. Collins Ж .... These included St. John's wort, ..... Horn-Ross PL, Lee M, John EM, Koo J (2000) Sources of phy-.
Cancer Causes and Control (2006) 17:85–93 DOI 10.1007/s10552-005-0391-6

ORIGINAL PAPER

Reliability and validity of an assessment of usual phytoestrogen consumption (United States) Pamela L. Horn-Ross Æ Stephen Barnes Æ Valerie S. Lee Æ Christine N. Collins Æ Peggy Reynolds Æ Marion M. Lee Æ Susan L. Stewart Æ Alison J. Canchola Æ Landon Wilson Æ Kenneth Jones

Received: 31 March 2005 / Accepted: 5 August 2005  Springer-Verlag 2006

Abstract Objective To evaluate the reliability and validity of a food-frequency questionnaire (FFQ) and database designed to quantify phytoestrogen consumption. Methods This study included 195 members of the California Teachers Study (CTS) cohort who, over a 10-month period, completed four 24-h dietary recalls, a pre- and poststudy FFQ, and provided two 24-h urine specimens. Participants (n = 106) in a parallel study (and 18 women who dropped out of the long-term study) completed a single recall and FFQ, and provided one 24-h urine specimen. All work was performed at the Northern California Cancer Center with the exception of the laboratory analyses which were performed at the University of Alabama, Birmingham. P. L. Horn-Ross (&) Æ V. S. Lee Æ C. N. Collins Æ A. J. Canchola Northern California Cancer Center, 2201 Walnut Ave., Suite 300, Fremont, CA, 94538, USA e-mail: [email protected] Tel.: +1-510-608-5014 Fax: +1-510-608-5085

Urinary phytoestrogens were determined using liquid chromatography–mass spectrometry. Reliability and validity were evaluated using Shrout–Fleiss intraclass correlations and energy-adjusted deattenuated Pearson correlations, respectively. Results Correlations reflecting the reproducibility of the FFQ phytoestrogen assessment ranged from 0.67 to 0.81. Validity correlations (FFQ compared to dietary recalls) ranged from 0.67 to 0.79 for the major phytoestrogenic compounds (i.e., daidzein, genistein, and secoisolariciresinol) and 0.43 to 0.54 for the less common compounds. Compared to urinary levels, validity correlations ranged from 0.41 to 0.55 for the isoflavones and 0.16 to 0.21 for total lignans. Conclusion Our isoflavone assessment is reproducible, valid, and an excellent tool for evaluating the relationship with disease risk in non-Asian populations. Further research is needed before these tools can accurately be used to assess lignan consumption. Keywords Phytoestrogens Æ Reliability Æ Validity Æ Dietary assessment Æ Urinary excretion

S. Barnes Æ L. Wilson Æ K. Jones Departments of Pharmacology and Toxicology and Biochemistry and Molecular Genetics, University of Alabama, Birmingham, AL, USA P. Reynolds California Department of Health Services, Environmental Health Investigations Branch, Oakland, CA, USA M. M. Lee Department of Epidemiology and Biostatistics, School of Medicine, University of California, San Francisco, CA, USA S. L. Stewart Division of General Internal Medicine, Department of Medicine, University of California, San Francisco, CA, USA

Introduction In recent years, substantial interest has developed in the effects of phytoestrogens on the risk of cancer and other chronic diseases. Two classes of phytoestrogens are of primary interest: the isoflavones, found primarily in soybased foods, and the lignans, found in various whole grains, seeds, legumes, fruits, and vegetables [1, 2]. Most

123

86

studies of disease risk have been conducted in Asian populations, where the intake of traditional soy-based foods is high. However, the effects of phytoestrogens on disease risk in ‘western’ populations is also of interest. In ‘western’ populations, the consumption of traditional soy foods is substantially lower and a portion of total isoflavone consumption is derived from soy protein and soy flour added to a variety of foods [3]. Nonetheless, isoflavones are still consumed in greater amounts (on average 1–3 mg/d) than are lignans ( < 1 mg/d); however, the relative bioactivity per mg of these two types of phytoestrogens in various mammalian tissues is not well established. In order to evaluate phytoestrogen consumption in epidemiologic studies, several phytoestrogen databases have been developed [1, 4–6]. Most of these databases are based on compilations of phytoestrogen values found in the literature [4–6] and thus, are based on a limited number of foods and determined using various laboratory techniques. We developed a database including a much wider range of foods, using common laboratory methodology to measure the phytoestrogen content of each food, and including foods with soy additives, which may be an important source of isoflavones in non-Asian populations [1]. The objective of this paper is to describe the reliability/reproducibility and validity of this latter database and our assessment of usual phytoestrogen intake using a FFQ.

Materials and methods This validation study was conducted in the context of the California Teachers Study (CTS) [7]. Established in 1995– 1996, the CTS is a prospective cohort study of 133,479 female public school teachers and administrators. The validation substudy included a random sample of 386 CTS participants, who were age 85 or younger at baseline and resided in the substudy area (western Alameda, Santa Clara, San Mateo, Santa Cruz, Monterey, or northern San Benito counties, California). Forty-six (12%) women were not contacted because they had died, moved out of the substudy area, or could not be located. The 340 women who were contacted were asked to participate in a 10month study that included four phases. Phase I (conducted in early 2000) involved (i) participating in an in-person interview that included a 24-h dietary recall and ancillary questions on such topics as menopausal status and the use of herbal and botanical supplements, antibiotics, oral contraceptives, and hormone replacement therapy, (ii) allowing the interviewer to take anthropometric measurements, (iii) collecting a 24-h urine sample following the interview and completing a 1-page checklist of foods consumed during this period, and (iv) completing a food-frequency questionnaire (FFQ) covering usual dietary intake during

123

Cancer Causes and Control (2006) 17:85–93

1999. Phases II–IV (occurring at 3-month intervals) involved a 24-h dietary recall administered via telephone on a random day within the assigned month. Participants were also asked to collect a second 24-h urine specimen during one of the final three phases (which was chosen at random for each participant). At the end of the study, participants were asked to complete a second FFQ covering usual dietary intake during 2000. Of the 340 women invited to participate, 195 (57%) completed the four dietary recalls, 23 (7%) began the study but dropped out prior to its completion, 108 (32%) declined to participate, and 14 (4%) were not interviewed for other reasons (most notably, illness). Of the 195 women participating in all four dietary recalls, 185 (95%) completed the pre-study FFQ, 182 (93%) completed the post-study FFQ, 180 (92%) provided a 24-h urine specimen following the first recall, and 175 (90%) provided a second 24-h urine specimen following either the 2nd, 3rd, or 4th dietary recall. We conducted a second parallel study in which participants completed only phase I of the study (as described above). Participants in this study were selected from among CTS members based on the same eligibility criteria as those in the validation substudy. Of 149 women identified for this study, 144 (97%) were contacted. Of these 144 women, 110 (76%) participated, 30 (21%) declined to participate, and four (3%) were not interviewed for other reasons. Of the 110 participating, 107 (97%) completed the FFQ and 106 (96%) provided a 24-h urine specimen. For selected analyses involving the urinary biomarkers, we included these 106 women as well as 18 of the 23 women who subsequently dropped out of the long-term study but completed phase I and provided a 24-h urine specimen. Since these 124 women completed only a single dietary recall and FFQ they were not included in other parts of the analysis that relied on the presence of more extensive data. This research was approved by the Institutional Review Boards (IRB) of the Northern California Cancer Center and the California State Health Department. The CTS is also overseen by the IRBs at the University of Southern California and the University of California, Irvine.

Dietary assessment All dietary recalls included reporting all foods, beverages, and vitamin and mineral supplements that were consumed, when they were consumed, how foods and beverages were prepared, and what was added to the foods or beverages (i.e., fats, salt, spices, condiments, etc.) during the 24-h period directly preceding the interview. During the in-person recall, portion size was estimated using visual aids, such as standard dinner plates, small wood cubes

Cancer Causes and Control (2006) 17:85–93

grouped into four portion sizes, glasses, measuring spoons and cups, several meat models, and two-dimensional images (actual size) of different portions of bananas, pizza, bread, and several other foods. After the in-person recall, interviewers oriented the participants to additional twodimensional portion size pictures that were left with the participant and used during the telephone recalls. The FFQ was an early version of the Block95 questionnaire [8, 9] with changes made to enhance measurement of phytoestrogen intake [1, 3, 10]: (i) canned peaches and dried apricots were asked as separate items; (ii) raisins were included as an individual item rather than grouped with ‘other fruits’; (iii) garbanzo beans, ceci beans, or chick peas were included as an individual item rather than subsumed under the general category of beans; and (iv) ‘regular’ (mung) bean sprouts, soybean sprouts or seeds, and soy milk were added. The FFQs were self-administered and checked by study staff for completeness. In addition, during the phase I in-person interview, we obtained information on the history of use of 14 herbal and botanical supplements. These included St. John’s wort, echinacea, and ginkgo biloba, all of which are rich in lignans, and soy supplements (in pill or powder form) that are rich in isoflavones. Participants also completed a checklist indicating if they had eaten any of 15 phytoestrogen-rich foods during the 24-h period during which they collected their urine. We quantified consumption of seven phytoestrogenic compounds (i.e., the isoflavones: daidzein, genistein, formononetin, and biochanin A; the coumestan: coumestrol; and the lignans: matairesinol and secoisolariciresinol), using the database we had developed previously [1]. The phytoestrogen values obtained in developing this database were applied to both the FFQs and the 24-h recall data based on their content per 100 g of food/ beverage. Urine collection and analysis Participants were instructed to collect all urine produced in the 24-h period starting immediately following the interview. Interviewers instructed participants on the use of collection ‘hats’, containers, and cold-packs. Participants were asked to record the time of each urination during the 24-h period and whether the capture was complete, partial, or missed. The samples were collected in plastic jugs and participants were asked to use the cold-packs and, if willing, to refrigerate the samples during the 24-h collection period. At the end of the 24-h, interviewers collected the sample, stored them in 20 C freezers until they were transported to the main laboratory (between 0.5 h and 2 weeks after collection; the average was 1 week), where they were thawed, aliquoted, and frozen at 70 C.

87

Specimens were transported on dry ice via overnight carrier in batches to the University of Alabama, Birmingham for phytoestrogen analyses. Liquid chromatography–electrospray ionization–multiple reaction ion monitoring mass spectrometry was used to measure urinary phytoestrogen levels. After thawing, urine specimens were centrifuged at 3000 g for 10 min to remove particulate material. Two clarified aliquots (1.0 ml) were separately added to 250 ll of 500 mM ammonium acetate buffer, pH 5.0, containing 40 units of b-glucuronidase/sulfate and 50 nmol of apigenin, phenolphthalein glucuronide and 4-methylumbelliferone sulfate and incubated at 37 C for 15 h. Urine positive and negative control samples were processed concurrently with each batch of urine samples that were analyzed. The hydrolyzed phytoestrogens were extracted with 3 · 2 ml of diethyl ether. The ether phases were combined and evaporated to dryness under nitrogen. The dry residue was resuspended in 0.5 ml 80% aqueous methanol. A 20 ll aliquot was subjected to isocratic analysis on a C8 reverse-phase column (100·4.6 mm i.d.) using a 40% acetonitrile–10 mM ammonium acetate mobile phase at 1 ml/min. The eluate was split and 0.1 ml/min passed into the IonSpray interface of a PE-Sciex (Concord, Ontario, Canada) API 3 triple quadrupole mass spectrometer. The instrument was operated in the negative ionization mode with the spraying needle voltage at 4900 V and the orifice potential at 50 V. Multiple reaction ion monitoring (MRM) was used in which parent molecular ions [M H] were selected one at a time in the first quadrupole and then subjected to collision-induced dissociation with Argon gas. Specific daughter ion fragments for each phytoestrogen were monitored with the third quadrupole. The following parent ion/ daughter ion pairs were used: m/z 253/223, daidzein; m/z 255/149, dihydrodaidzein; m/z 257/108, O-desmethylangolensin; m/z 241/108, equol; m/z 269/133, genistein; m/z 283/239 glycitein; m/z 269/149 apigenin (internal standard); m/z 317/93, phenolphthalein; m/z 175/119, 4methylumbelliferone; 297/253, m/z enterodiol; and 301/ 106, m/z enterolactone. The individual ion chromatograms were subject to analysis of peak areas by the program MacQuant provided by the manufacturer. Samples containing five differing amounts of phytoestrogen standards and the hydrolyzed form of the internal standards were also analyzed in order to establish a calibration curve. The appearance of phenolphthalein and 4-methylumbelliferone was used to validate that hydrolysis had taken place. Quantitative corrections were made in all samples and standards by relating the area of the phytoestrogen peaks to the area of the internal standard apigenin. The correlation coefficients for the calibration curves for each analyte were 0.99 or greater. Recovery of the phytoestrogens through the extraction procedure is 82–100% with the best recovery

123

88

for genistein and the lower value for daidzein and dihydrodaidzein. Statistical analysis Chi square and t-tests were used to evaluate differences between the substudy participants and the remainder of the cohort from which they were selected [11]. Reproducibility of the phytoestrogen estimates derived from the pre- and post-study FFQs (often referred to as reliability under the assumption that any reported changes in intake during this period are due to reporting errors rather than true changes in eating habits) was evaluated using Shrout–Fleiss intraclass correlations [12]. Energy-adjusted deattenuated Pearson correlation coefficients were used to compare the validity of the FFQ relative to the average of the 24-h recalls and between the phytoestrogen intake measures and urinary excretion [13, 14]. Urine measures were adjusted for creatinine levels (which had been determined using a standard clinical assay) and volume to standardize for individual differences in urinary output. Since urinary phytoestrogen levels were influenced by both what was consumed during the 24-h recall period as well as the subsequent 24-h urine collection period, for each phytoestrogen-rich food the participant reported eating during the urine collection period, based on the checklist (see above), we added one medium serving of that food to her intake. For subanalyses accounting for soy and botanical supplement use, we added the phytoestrogens that would be received from one tablet or dose of isoflavone supplements, soy protein powder, St. John’s wort, echinacea, and/or ginkgo biloba for women indicating they were regular users of these supplements for a year or more.

Results We compared the 195 substudy participants to the remainder of the CTS cohort (age £85 years) on a variety of factors to evaluate the representativeness of our study sample. Substudy participants were quite similar to the entire cohort in terms of age, menopausal status, body mass index, physical activity, broad dietary composition (i.e., contribution of fat and fruits and vegetables to the diet), and intake of phytoestrogens (Table 1). Substudy participants, however, were more likely to be Latina and less likely to be African American or white; a reflection of the substudy area compared to the state as a whole. In addition, the distribution of energy consumption between the two groups differed slightly. Substantial variation in the intake of the various phytoestrogen compounds was reported by our study sample

123

Cancer Causes and Control (2006) 17:85–93

(Table 2). Median values for the primary isoflavones, genistein and daidzein, coumestrol, and the primary lignan, secoisolariciresinol, were higher when quantified from the 24-h recalls than from the FFQ. Intake as measured by FFQ over the 1-year study period was relatively constant, and correlations reflecting the FFQ’s reproducibility for these compounds were high (ranging from r = 0.67 for matairesinol to r = 0.81 for secoisolariciresinol; Table 2). Correlations quantifying the validity of the FFQ assessment, compared to the average of the four 24-h recalls collected over a 10-month period, were high for the primary isoflavones, genistein and daidzein, and the primary lignan, secoisolariciresinol (ranging from 0.67 for daidzein measured by both the pre- and post-study FFQ to 0.79 for secoisolariciresinol measured by the post-study FFQ; Table 2). Correlations were lower for the other compounds, with the lowest correlations (r = 0.43 and 0.44) observed for formononetin (using the post-study FFQ) and biochanin A (using the pre-study FFQ), respectively. Correlations between the recalls and both the pre-study and post-study FFQs were generally similar for all compounds. Validity correlations comparing the phase I urinary phytoestrogen levels to (i) the phase I dietary recall and (ii) the pre-study FFQ were reasonably high for the isoflavones (ranging from r = 0.41 to 0.55 for the dietary recall assessment of genistein and the FFQ assessment of daidzein, respectively) (Table 3). Validity coefficients were slightly higher for the FFQ assessment than for the dietary recall assessment. Using the post-study FFQ and averaging the phytoestrogen levels in the two urine specimens led to similar results (with correlations ranging from 0.50 for genistein to 0.53 for total isoflavones; additional data not shown). Antibiotic use can kill gut bacteria and therefore negatively impact the metabolism of phytoestrogens. Among women who had not used antibiotics in the past year, the correlations with urinary phytoestrogen levels were slightly higher and more similar for the two assessment methods (Table 3). Including total isoflavone exposure from regular use of isoflavone supplements (i.e., pills containing isoflavones or soy protein powder) did not improve the correlation between total isoflavone intake (measured by dietary recall) and urinary excretion (r = 0.52 among women not exposed to antibiotics). Correlations measuring the validity of both the dietary recall and FFQ assessments, relative to urinary excretion, were poor for lignan consumption (r = 0.21 for the dietary recall method and r = 0.16 for the FFQ). Limiting analyses to women who had not used antibiotics in the past year did not favorably impact these correlations (r = 0.17 and r = 0.16, respectively, for the two assessment methods) nor did adding exposure from St. John’s wort, echinacea, or

Cancer Causes and Control (2006) 17:85–93 Table 1 Selected baseline characteristics of participants in the validation substudy and the CTS cohort

Characteristic Age < 45 45–54 55–64 ‡65 Mean (SD) Race/ethnicity White African American Latina Asian/Pacific Islander Other

b

Excluding substudy participants and women age >85 at baseline

c Percentages are based on the total number of women with known data for a given variable d

Includes women with incomplete data

e Average lifetime strenuous and moderate physical exercise

56 (29%)c 68 (35%) 33 (17%) 38 (19%) 51.9 (13.0) 163 2 15 8 6

(84%) (1%) (8%) (4%) (3%)

35,560 (27%) 37,881 (29%) 25,485 (19%) 32,119 (25%) 53.5 (14.2) 113,660 3,536 5,384 4,548 2,750

0.16

0.11

(88%) (3%) (4%) (4%) (2%)

0.05

79 (41%) 97 (50%) 19 (10%)

50,735 (39%) 69,376 (53%) 10,934 (8%)

0.61

Body mass index < 19.1 underweight 19.1–25.8 normal weight 25.9–27.3 marginally overweight 27.4–32.2 overweight 32.3–44.8 obese ‡44.9 morbidly obese Mean (SD)

11 (6%) 97 (63%) 14 (7%) 26 (14%) 18 (10%) 0 (0%) 24.6 (4.7)

6,962 (6%) 79,244 (63%) 8,573 (7%) 20,021 (16%) 10,020 (8%) 733 (1%) 24.9 (5.1)

0.80

Physical activity (hours/week)e