Fruit Intake Associated with Urinary Estrogen Metabolites in Healthy Premenopausal Women Abstract Urinary Concentrations of 2:16-hydroxyestrone (2:16-ohe 1 ) Approximate Concentrations of 2-ohe 1 and 16α-ohe 1 in Breast Tissue. as Estrogens Are Purported to Be Involved in Breast Cancer Deve- Lopment

to estrogen metabolites may thus be influenced by diet.

Research has sought to determine whether modifiable factors influence formation of certain estrogen metabolites.Prior studies indicate that estrogen metabolism may be altered in response to intake of certain dietary constituents, with support coming from laboratory studies, as well as human cross-sectional and intervention studies [6,[15][16][17][18][19][20][21][22][23][24][25].Of the cross-sectional studies investigating associations between a wide range of dietary factors and the 2-and 16α-OHE 1 metabolites, a few have reported associations with fruit, vegetable, and coffee consumption, while one reported an association with high fat/low fiber diets [15][16][17][18].Our study sought to investigate associations between urinary 2-and 16α-OHE 1 metabolites and dietary factors ascertained through 3-day food records in 191 premenopausal healthy women.

Study Design
Participants in the Equol, Breast, and Bone (EBB) study were recruited from the Group Health Cooperative (GHC), a large mixed-model health care system in western Washington State.The methods for this study have been described elsewhere [26].Briefly, women were eligible if they were premenopausal, aged 40 to 45 years, and had received a screening mammogram at GHC prior to recruitment.Women were ineligible if they were currently using or had used hormone therapy or oral contraceptives for more than one month in the past year; had a personal history of breast cancer, or had shown signs of perimenopause.
After obtaining informed consent, EBB participants completed a health and demographics questionnaire.At the clinic visit, weight, height, waist and hip circumference measurements, percent body fat, and fasting blood and spot urine samples (during days 5 -9 of their menstrual cycle) were collected.In addition, all participants were asked to complete a 3-day food record (3-DFR) within two weeks of this clinic visit.Participants were given a serving size booklet which contained pictures of commonly consumed foods in different portion sizes, as well as a ruler, a thickness guide, a serving spoon size guide, and tips on how to estimate servings.Dietary intake data from the 3-DFR were analyzed using the Nutrition Data System for Research software by the Nutrition Assessment Shared Resource at the Fred Hutchinson Cancer Research Center using previously described methods [26].Estimates of daily intake of nutrients, grains, meats and shellfish, egg, dairy, tea and coffee, as well as botanically-defined groupings of fruits and vegetables were obtained.
Measurements of urinary creatinine concentrations were based on a kinetic modification of the Jaffe reaction with the use of the Roche Reagent for Creatinine (Roche Diagnostic Systems, Nutley, NJ).

Statistical Analysis
The ratio of 2:16-OHE 1 was computed from the concentrations of 2-OHE 1 and 16α-OHE 1 .When analyzed separately, the 2-and 16-OHE 1 values were corrected for creatinine by dividing the estrogen metabolite concentration by the creatinine concentration.Lack of normality was assessed for each continuous variable; serum E 2 , 2-OHE 1 , and 16α-OHE 1 were skewed, and thus were log-transformed.Quartiles were created for the dietary factors, with the exception of the botanical groupings.For botanical groupings, analysis was limited to those groups that were consumed by at least 30 (≥15%) participants.For those botanical groups, a variable was created that categorized participants as: not consuming the botanical grouping; those consuming less than the median value; and those consuming more than the median value.
For our primary analysis, we tested associations between the 2:16-OHE 1 ratio and dietary factors using generalized linear models (GLM) adjusted for potential confounding factors.A confounding factor was included in the GLM if it was statistically significantly associated with the 2:16-OHE 1 ratio (as determined using one-way ANOVA) and with dietary factors; all models included confounding factors (including serum E 2 , ethnicity and smoking history), in addition to total energy.Correlations were also examined between continuous variables using one-way ANOVA to ensure that variables with high correlations were not included in the same model.
To follow up on associations observed for fruit and vegetable consumption, we examined whether associations were observed between botanical groupings and 2:16-OHE 1 .In addition, to follow up on associations we observed between dietary factors and 2:16-OHE 1 we used GLM to estimate the associations between dietary factors and the 2-OHE 1 and 16α-OHE 1 metabolites separately; these analyses were adjusted for creatinine, and total energy, in addition to confounding factors (ethnicity and smoking history).Benjamini-Hochberg correction was used to adjust for multiple testing in which adjustments were made according to the number of statistical tests for each hypothesis [29].This analysis was conducted using Stata v. 11.

RESULTS
Demographic characteristics of the study population are presented in Table 1.The median age was 42.4 years; median body mass index (BMI) was 25.7 kg/m 2 .59.4% of the women had a history of breast feeding and 64.6% were never smokers.We observed fruit and vegetable consumption to be positively associated with the 2:16-OHE 1 ratio after adjustment for total energy, ethnicity, BMI, parity, smoking history, and serum E 2 (Table 2).In models adjusting for total energy only, the following dietary factors were associated with 2:16-OHE 1 : total sugars (p = 0.01), caffeine (p = 0.03), and refined grains (p = 0.03; data not shown); however in the fully adjusted model, these factors were no longer associated with the 2:16-OHE 1 ratio (Table 2).Furthermore, after adjustment for multiple tests only the association between 2:16-OHE 1 and fruit consumption remained significant.
We observed two botanical groupings containing fruit to be positively associated with the 2:16-OHE 1 ratio in the fully adjusted model (Table 3).Specifically, Rutaceae (e.g., citrus fruits and juices) and Musaceae (comprised primarily of bananas) were positively associated with 2:16-OHE 1 ; however, after adjustment for multiple testing, the association for Rutaceae was no longer significant.Intake of fruit-containing botanical groupings ranged from 21.1% of women consuming for Anacardiaceae (e.g., mangoes) to 85.6% for Rosaceae (e.g., apples, stone fruits, and some berries).59.8% and 45.9% of women reported consuming Rutaceae and Musaceae, respectively.None of the botanical groupings for vegetables, including Cruciferae (which were consumed by 69.1% of women, although only 8.2% reported eating >1 serving/day), were associated with the 2:16-OHE 1 ratio (p-value = 0.94; data not shown).
We next assessed whether fruit or vegetable intake was associated with either 2-OHE 1 or 16α-OHE 1 separately using the same categorization for fruit and vegetables from Table 2 and for botanical groupings from Table 3.When assessed individually, neither fruit nor vegetable intake, including botanical groupings, were associated with 16α-OHE 1 (Table 4).2-OHE 1 was associated with overall fruit consumption and intake of Rutaceae and Musaceae botanical groupings, after adjustment for confounders.While overall fruit consumption and Musacea groupings were positively associated with 2-OHE 1 , the pattern of association between Ruataceae and 2-OHE 1 was less clear.The association between 2-OHE 1 and Rutaceae became stronger when confounders were included in the model, indicating an influence of ethnicity and smoking history on the relationships between consumption of foods in the Rutaceae grouping and 2-OHE 1 .

DISCUSSION
We observed fruit consumption to be positively associated with the 2:16-OHE 1 ratio among premenopausal women, and this was mostly due to its association with 2-OHE 1 concentrations.We also observed that botanical food groupings containing citrus and bananas were associated with 2-OHE 1 concentrations.To our knowledge, this is the first study to examine the association between botanical groupings and 2:16-OHE 1 in premenopausal women.
Numerous phytochemicals, including quercetin and other flavonoids (including naringenin found in citrus fruits), have also been shown to influence the activity of phase II enzymes, including UGT, GST, QR, and SULT enzymes [4][5][6].These relationships are complex, as flavonoids are demonstrated to induce some phase II enzymes, including multiple UGT enzymes and QR, while inhibiting others, such as SULT, but overall the influence of flavonoids on phase IIenzymes is consistent with chemoprevention.
Further support for the impact of dietary intake on estrogen metabolites come from human feeding and intervention studies.Feeding studies and interventions have also observed increased consumption of isoflavones (genistein and daidzein), flaxseed, I3C supplementation, and lowfat diets to be associated with increased 2-OHE 1 concentrations and the 2:16-OHE 1 ratio in women [30][31][32][33], findings that are consistent with the in vitro studies discussed above.With flaxseed and berries being a good source of lignans [34], and our finding that fruit was associated with 2:16-OHE 1 , we investigated the potential for berries (botanical groupings Ericaceae and Rosaceae) to be associated with 2:16-OHE 1 .Although we did not see evidence of an association between Ericaceae and Rosaceae and 2:16-OHE 1 it is possible that the levels of berry consumption were too low to detect associations and that lignans may actually contribute in part to fruit consumption being associated with 2:16-OHE 1 .Our study observed fruit intake to be associated with the 2:16-OHE 1 ratio.While it is possible that these findings are spurious, the association between 2:16-OHE 1 and fruit intake and the Musaceaegrouping remained significant after the correction for multiple tests.Furthermore, in the context of the aforementioned support from prior animal and human studies [15,16,[20][21][22], this finding may have validity.The observed association between vegetables and 2:16-OHE 1 should be interpreted with caution as these associations were not significant after correction for multiple testing, and as a result may be more likely due to chance.
Limitations of our study include the use of participant recall for dietary intake.While dietary assessment tools in general may misclassify dietary intake, a 3-day food record (which our study used) arguably has higher validity than a food frequency questionnaire, particularly for major food groups [35][36][37].Another limitation centers on the issue of generalizability.The women in our study representa more health-conscious segment of the general population, with for example a mean BMI of 25.7 in our study and only 4% of participants being current smokers.Thus their dietary intake may not be representative of women from the US population as a whole.However, these characteristics may enhance the internal validity because confounding factors, such as cigarette smoking, are minimized in this rather homogeneous population.An additional limitation involves the use of a single spot urine collection for assessment of the 2-and 16α-OHE 1 concentrations.First, previous studies have indicated that there was no difference between a spot urine and a 24-hour urine for assessment of 2:16-OHE 1 [38], and that urinary 2:16-OHE 1 is correlated with plasma 2:16-OHE 1 (r 2 = 0.83 among non-OC users whose urine samples are collected mid-cycle, which matches our study's protocol) [39].However, despite the high reproducibility of the assays for the 2-and 16α-OHE 1 as indicated by the low CVs, Williams et al. estimated that 5 collections would be ideal in order to capture the variability in urinary hydroxy estrogens [40].Lastly, as with all cross-sectional studies, we were unable to assess the temporality of the diet and 2:16-OHE 1 association.However, a strength of cross-sectional studies are their ability to investigate a variety of food groups associated with 2:16-OHE 1 , and in the case of diet and 2:16-OHE 1 multiple diet interventions have previously demonstrated that diet precedes changes in the 2:16-OHE 1 [30][31][32][33].Furthermore, investigation of associations with botanical groupings in particular allowed for an examination of specific sources of phytochemicals within the diet.
Our study adds to the large body of literature that indicatesdietary intake is associated with 2:16-OHE 1 [6,[15][16][17][18][19][20][21][22][23][24][25] in premenopausal women.With urinary 2:16-OHE 1 representing the 2:16-OHE 1 ratio in the breast [14], this line of research may have implications for modifiable factors related to breast cancer.However, while we observed that fruit consumption, including the botanical grouping Musaceae, was associated with increasing 2-OHE 1 , these results would need to be replicated in larger, more generalizable studies of premenopausal women before definitive conclusions can be drawn.Such studies would ideally be designed to report on botanical groupings in relation to urinary estrogen metabolites, including the 2-and 16α-OHE 1 in order to shed light on the particular aspects of the diet that are associated with estrogen metabolism among premenopausal women.

Table 1 .
Demographic characteristics of the EBB study population.

Table 2 .
Associations of nutrient and food groups with the urinary 2:16 hydroxy-estrone ratio.