Phytoestrogen intake from foods, during adolescence and adulthood, and risk of breast cancer by estrogen and progesterone receptor tumor subgroup among Ontario women
Phytoestrogen intake may reduce breast cancer risk and limited evidence suggests this association may hold for hormone receptor-positive tumors only. The study aims were to assess whether the association between phytoestrogen intake during adolescence and adulthood and breast cancer risk varies by estrogen and progesterone receptor (ERPR) tumor subgroup. Cases were identified from the Ontario Cancer Registry (2002–2003), and ERPR status was ascertained from pathology reports for 81% of cases (n = 2,438). Controls were identified through random digit dialing of Ontario households (n = 3,370). Published phytoestrogen food values were applied to food frequency questionnaire responses to assess isoflavone, lignan and total phytoestrogen intake, during adolescence and adulthood. Polytomous multivariate logistic regression was used to estimate adjusted odds ratios (ORs) for association between phytoestrogen intake and breast cancer risk by hormone receptor ERPR tumor subgroups. Among premenopausal women, few associations were observed for adolescent or adult phytoestrogen intake across all tumor subgroups. Among postmenopausal women, adolescent phytoestrogen intake (isoflavone, lignan and total) was associated with reduced risk across all hormone receptor subgroups; however, statistical significance was most consistent within the ER+PR+ subgroup. For example, ER+PR+ postmenopausal breast cancer risk was associated with adolescent phytoestrogen intake (highest vs. lowest: OR = 0.79; 95% confidence interval: 0.65–0.96). Among all women and postmenopausal women, ORs for high adult lignan intake were all below 1.0 within each tumor subgroup, suggesting reduced breast cancer risk, although none reached statistical significance. In conclusion, adolescent phytoestrogen intake was associated with reduced postmenopausal breast cancer, particularly for ER+PR+ tumor subgroup.
Phytoestrogens are estrogen-like plant compounds that bind to estrogen receptors (ER) and may protect against breast cancer by altering estrogen metabolism and inhibiting cancer cell growth.1–3 The most common classes of phytoestrogens are isoflavones (e.g. genistein and daidzein) and lignans (e.g. secoisolariciresinol). Isoflavones are found at high levels in soy foods (e.g. soy beans, soy milk and tofu), whereas lignans are generally found at lower levels in many commonly consumed plant sources (e.g. nuts, seeds, cereals, breads, vegetables and coffee) and flaxseed is a particularly rich source.4 Three meta-analyses have all reported inverse associations between isoflavone/soy intake and breast cancer risk; however, these were inconsistently found among both pre- and postmenopausal women,5 only among postmenopausal women6 or only among high-soy consuming Asian populations.6, 7 Another meta-analysis reported that lignan intake is significantly associated with reduced breast cancer risk among postmenopausal women only.8 Early life phytoestrogen exposure may be important with regard to breast cancer risk.9, 10 The few studies of adolescent phytoestrogen intake have reported inverse associations with breast cancer risk among premenopausal11 or postmenopausal women only12 or both.13–15
Breast cancer risk factor profiles differ by estrogen and progesterone receptor (ERPR) tumor status (e.g. Refs.16–19); several reproductive factors are most consistently associated with hormone receptor-positive tumors, possibly mediated through estrogen exposure.17, 18 Given that an important proposed mechanism of action for phytoestrogens is via ER, the association between phytoestrogens and breast cancer risk may be primarily among hormone receptor-positive tumors. Studies by hormone receptor subgroups, however, have been inconclusive.20–34 Only some studies20, 24, 27–30, 34 reported any evidence of heterogeneity, and a few actually suggested inverse associations for ER− breast cancer only20, 34; differences by hormone receptor status have been observed most consistently for isoflavones27–30 but also for lignans24, 34 and coumestrol.20 A recent meta-analysis found that the inverse association did not differ between lignan intake and ER+ and ER− breast cancer risk among postmenopausal women.8 Few of the previous studies were in North American populations21, 33, 34; most were in Asia25, 27, 29–32 and Europe,20, 22–24, 26, 28 where the consumption of phytoestrogen containing foods and the content of the food databases may differ from that in North America. Despite the potential importance of early life exposures,9, 10, 35 previous studies have not examined phytoestrogen intake during adolescence and hormone receptor defined breast cancer risk.
The association between phytoestrogen intake from foods during adolescence14 and adulthood36 and overall breast cancer risk was previously reported for women in the current study, but only recently have data become available to evaluate hormone receptor status. Intake of lignans, isoflavones and total phytoestrogens during adolescence was found to be associated with reduced breast cancer risk.14 Adult lignan intake was also found to be associated with reduced breast cancer risk, while isoflavone intake was not.36 The objectives of this study were to evaluate the association between phytoestrogen intake from foods during both adolescence and adulthood and breast cancer risk by ERPR tumor subgroups among women in The Ontario Women's Diet and Health Study (OWDHS), a population-based case–control study.
Material and Methods
Data sources and case/control definition
Methods for the OWDHS have been described previously.36 Briefly, cases were women aged 25–74 years with a first pathologically confirmed breast cancer diagnosis identified from the Ontario Cancer Registry in 2002 and 2003. Controls were identified from Ontario households using a random digit dialing method and were frequency age-matched 1:1 to cases. In 2003 and 2004, cases and controls completed a mailed risk factor questionnaire, a modified Block Food Frequency Questionnaire (FFQ) and a short adolescent FFQ. The risk factor questionnaire included 79 questions and collected information on a wide range of lifestyle, reproductive and medical history factors.
In 2009, we ascertained ERPR tumor data from pathology reports for use in this current study. Since 2000, pathology laboratories have determined ERPR status based on immunohistochemical staining and documented results in pathology reports. Although pathology reports are forwarded to the Ontario Cancer Registry, ERPR status was not routinely recorded in the registry database. The study protocols were approved by the University of Toronto Research Ethics Board.
Overall 3,101 of 4,109 (75%) cases and 3,471 of 4,352 (80%) identified controls participated in the Ontario Women's Diet and Health Study by completing the risk factor questionnaire. Joint ER and progesterone receptor (PR) status was available for 2,519 (81%) cases (13 had ER only, 1 had PR only and both ER and PR was unknown for the remaining cases). Adult diet questionnaire data were available on 2,491 of these cases with known ERPR status and 3,427 controls. There were 110 women (53 cases with ERPR data and 57 controls) excluded from all analyses (both adult and adolescent) due to reporting extreme energy intake during adulthood (<600 or >4,500 kcal). Of 2,438 cases with dietary data and known ERPR status, 41 ER–PR+ cases were excluded from all analyses due to small numbers. After all exclusions, data on adult intake were available from 2,397 cases with joint receptor status and 3,370 controls, and data on adolescent intake were available from 2,353 cases with joint receptor status and 3,337 controls.
Derivation of adult and adolescent phytoestrogen intake
The validated adult Block FFQ37 was expanded to 178 items and included 67 additional foods identified as important phytoestrogen sources.36 Women were asked to report both the frequency (how often 2 years ago) and portion size (how much each time) of each consumed item. As part of the OWDHS, a Canadian phytoestrogen database of 119 unique foods was developed.4 Using this database, daily phytoestrogen consumption estimates were derived by applying phytoestrogen values per serving (microgram) to FFQ frequency and portion size responses for each food and by summing over all reported foods.36 Phytoestrogen variables included total isoflavone (genistein, daidzein, glycitein and formononetin), total lignan (secoisolariciresinol, matairesinol, pinoresinol and lariciresinol) and total phytoestrogen (isoflavone, lignan and coumestan) intake. As the primary coumestan (coumestrol) is present in foods at low concentrations,4 and was minimally consumed in our sample, coumestan was not evaluated on its own (although was included in the total phytoestrogen variable).36
The adolescent FFQ included 55 items identified as important phytoestrogen sources.14 Women were asked, “How often did you eat the following foods when you were 10 to 15 years old?” Portion sizes were assigned to each item and four response options were provided (never, monthly, weekly and daily). We previously reported adolescent phytoestrogen intake, using published analytical data to derive ranked intake estimates.14 In this present study, we applied values from our laboratory-based phytoestrogen database4 to foods in the adolescent FFQ using the same methods used for the adult questionnaire described above. A total of 70 individual foods were included in the 55 item adolescent FFQ; when FFQ items contained more than one food (e.g., “blueberries, strawberries or cranberries”), the mean phytoestrogen value for all included foods was used. Values from the phytoestrogen database were used to derive the daily intake for total isoflavones, total lignans and total phytoestrogens (isoflavones, lignans and coumestans).
Statistical data analysis
All phytoestrogen intake variables (isoflavone, lignan and total phytoestrogen) during both adolescence and adulthood were categorized into tertiles based on intake distributions among the controls. Simultaneous odds ratio (OR) and 95% confidence intervals (CI) estimates were obtained for joint receptor status (ER+PR+, ER–PR− and ER+PR–) using multivariate polytomous logistic regression. The Wald test for heterogeneity was used to evaluate statistical significance between receptor subgroups comparing the ORs across all three ERPR receptor subtypes.
All models were adjusted for age. For consistency with our earlier report on breast cancer risk and adult intake,36 all adult phytoestrogen intake variables (isoflavone, lignan and total) were adjusted for age at first live birth, duration of hormone replacement therapy use (among postmenopausal women only), personal history of benign breast disease, history of breast cancer in a first degree relative and dietary fiber intake. For adolescence, no confounders had been identified previously14 and thus models were adjusted for age only.
The likelihood ratio test was used to test the significance of a multiplicative interaction between phytoestrogen intake and menopausal status, and analyses are presented stratified by menopausal status. Tests for trend were conducted using the median phytoestrogen value within each category. All analyses were conducted using SAS v9.2.
Joint ERPR data were available for 2,438 cases with reported daily adult energy intake within the range of 600–4,500 kcal: 1,432 were ER+PR+, 500 were ER−PR−, 465 were ER+PR− and only 41 were ER−PR+. The 41 ER−PR+ cases were excluded due to the small number, resulting in 2,397 cases for all analyses. Table 1 shows the distribution of several sociodemographic variables and known breast cancer risk factors among controls, the three predominant ERPR case subgroups and cases with missing ERPR data. The mean age of controls was 55 years and for ER+PR+, ER−PR−, ER+PR− and unknown ERPR cases, the mean ages were 57, 54, 59 and 57, respectively. The distributions of ethnicity, education and BMI were similar across controls, ERPR subgroups and women with unknown ERPR status. Family history of breast cancer varied between controls and ERPR case groups; as expected, controls were less likely than all case groups to report a first degree relative with breast cancer. Menopausal status also varied by ERPR status with the fewest premenopausal women among the ER+PR− subgroup.
Table 1. Distribution of selected characteristics among controls and breast cancer cases categorized by joint hormone receptor status in the Ontario Women's Diet and Health Study
Table 2 shows the frequency distribution, OR estimates and 95% CI for isoflavone, lignan and total phytoestrogen intake during both adulthood and adolescence among ERPR defined breast cancer subgroups. Adult phytoestrogen intake was not consistently associated with any breast cancer subgroups, and no significant heterogeneity by ERPR was observed. Only one significant OR was observed and this was a positive association between the highest tertile of isoflavone intake and ER−PR− breast cancer risk (OR = 1.38; 95% CI: 1.05–1.81; ptrend = 0.01). While all but one point estimate for lignan intake ranged from 0.80 to 0.97 across the tumor subgroups (suggesting a possible reduction in risk), none of these reached statistical significance. For adolescent intake, nearly all point estimates for isoflavone, lignan and total intake across all three tumor subgroups were below 1.0, although few of these reached statistical significance. Significant inverse associations were observed for isoflavone intake and ER+PR− breast cancer risk (highest vs. lowest tertile: OR = 0.77; 95% CI: 0.60–0.99; ptrend = 0.05) and lignan intake and ER+PR+ breast cancer risk (highest vs. lowest tertile: OR = 0.86; 95% CI: 0.73–1.00; ptrend = 0.04). Statistically significant tests for heterogeneity of ORs between all hormone receptor subgroups were observed for adolescent intake of both lignan (p = 0.005) and total phytoestrogen intake (p = 0.007).
Table 2. Odds ratio (OR) estimates and 95% CI for isoflavone, lignan and total phytoestrogen intake during adulthood and adolescence among hormone receptor-defined breast cancer cases and controls
Additional multivariate analyses adjusting the adolescent models for all of the covariates included in the adult models revealed no substantial differences (data not shown). The magnitudes of most multivariate-adjusted ORs were very similar to the age-adjusted models (i.e. no strong evidence of confounding) and no associations became statistically significant.
Analyses stratified by menopausal status are presented in Table 3, although formal tests for interaction were not statistically significant. Among premenopausal women, most of the ORs were hovering near 1.0, however, for the ER+PR+ case subgroup the ORs for the highest categories of adult intake were consistently less than 1.0 (although not statistically significant) (Table 3). Tests for heterogeneity across ERPR subtype were not statistically significant and only two significant associations were observed among premenopausal women. The highest category of adult total phytoestrogen intake was significantly associated with increased ER−PR− breast cancer risk (OR = 1.65; 95% CI: 1.06–2.57; ptrend = 0.04) and the middle category of adolescent total phytoestrogen intake was significantly associated with reduced risk of ER−PR− breast cancer (OR = 0.66; 95% CI: 0.44–0.99; ptrend = 0.70) (Table 3).
Table 3. Odds ratio estimates and 95% CI for isoflavone, lignan and total phytoestrogen intake among hormone receptor-defined breast cancer cases and controls, stratified by menopausal status
Among postmenopausal women, the results were generally null or not significant for all adult phytoestrogen intake categories with the exception of a positive association between adult isoflavone intake and ER−PR− breast cancer (highest vs. lowest tertile: OR = 1.50; 95% CI: 1.05–2.15; ptrend = 0.04) (Table 3). None of the tests for heterogeneity were significant for adult intake. Point estimates for the highest tertile of adult lignan intake ranged from 0.74 to 0.93 across the tumor subgroups, suggesting a possible reduction in risk, although none of these reached statistical significance. Adolescent intake of isoflavones, lignans and total phytoestrogens was associated with reduced risk of all three breast cancer subgroups (ER+PR+, ER−PR− and ER+PR−); although these findings were most noticeably statistically significant for ER+PR+ and significant tests for heterogeneity were observed for lignan and total phytoestrogen intake. Within the ER+PR+ tumor subgroup, the highest tertiles of adolescent isoflavones (OR = 0.81; 95% CI: 0.67−0.98; ptrend = 0.09), lignans (OR = 0.78; 95% CI: 0.64−0.95; ptrend = 0.01) and total phytoestrogens (OR = 0.79; 95% CI: 0.65−0.96; ptrend = 0.04) were all significantly associated with reduced breast cancer risk. Significant inverse associations were also observed within the ER−PR− subgroup and the middle categories of lignan and total phytoestrogen intake but the tests for trend were not statistically significant. ER+PR− breast cancer risk was also significantly reduced with highest tertile of isoflavone intake during adolescence (OR = 0.68; 95% CI: 0.51–0.90; ptrend = 0.01).
Overall, the results of our study suggest that among premenopausal women, there were minimal differences across tumor subgroups and very few consistent or significant associations were observed. Among postmenopausal women, adolescent phytoestrogen intake (isoflavones, lignans and total) was associated with reduced breast cancer risk. This was found most consistently and significantly among the ER+PR+ tumor subgroup; however, some reduced risks were also found for the other tumor subgroups. No statistically significant reductions in risks were apparent for adult phytoestrogen intake within any particular hormone receptor subgroup of postmenopausal breast cancer; however, point estimates for high adult lignan intake (vs. low) were all below 1.0 for each tumor subgroup, suggesting a possible reduction in breast cancer risk, although none of these reached statistical significance. Independent of ERPR status, we previously reported that adult lignan intake was associated with reduced breast cancer risk overall.36
Phytoestrogens are similar in structure to estradiol and exert weak estrogenic activity.5, 38 They have the potential to limit estrogen production by inhibiting aromatase synthesis, and they can bind to ER, competitively blocking the binding of more potent natural estrogens.2, 38, 39 Lifetime estrogen exposure is a known breast cancer risk factor.40, 41 Thus, it has been widely hypothesized that possible associations between phytoestrogen intake and breast cancer risk may be strongest among hormone receptor-positive cases. All studies to date by ERPR tumor subgroup have only examined phytoestrogen intake in adulthood.
Six cohort studies, represented by seven publications, have reported on adult intake of isoflavones,25 lignans22, 24 or both20, 21, 23, 26 and breast cancer risk by ERPR status and only two concluded that there were differences by ERPR status.20, 24 The remaining four studies reported inverse associations for isoflavone25 or lignan22 intake among postmenopausal women independent of ERPR or no associations at all for isoflavone and lignan intake independent of both menopausal and ER status.21, 26 Of the studies to report heterogeneity by ERPR, one previous French study of lignans only, reported significant reduced ER+PR+ breast cancer risk among postmenopausal,24 but not premenopausal women.23 This is somewhat consistent with our results of an odds ratio less than 1.0 for adult lignan intake and ER+PR+ breast cancer risk but our results were not statistically significant and similar ORs were observed in both premenopausal and postmenopausal women and among ER+PR− postmenopausal women. The other cohort study to report differences by ERPR found that intermediate intake of coumestrol, but not lignans or isoflavones, was inversely associated with ER-PR- tumors among premenopausal and postmenopausal women.20 Given the low intake, we were not able to evaluate coumestrol on its own and it was included in our total phytoestrogen measure only.
Most previous case–control studies of adult phytoestrogen intake and hormone receptor defined breast cancer risk have evaluated soy foods or isoflavones only,27, 29–32 other studies included both lignans and isoflavones28, 33 or lignans only.34 Some,27–30 but not all,31–33 case–control studies of isoflavones or soy intake have reported stronger inverse associations with ER+PR+ or ER+ breast cancer risk. This was, however, inconsistently observed in a study of premenopausal women only,28 among postmenopausal but not premenopausal women,27 or not reported by menopausal status.29, 30 One previous case–control study of lignans only reported an inverse association between lignans and premenopausal breast cancer risk only among ER− cases34; however, the other case−control studies which included lignans concluded there was limited evidence of heterogeneity by hormone receptor status.28, 33 In summary, more cohort and case–control studies reporting differences by hormone receptor status have found this for isoflavones27–30 although there is some evidence of differences for lignans24, 34 and coumestrol.20
There is some evidence that isoflavones have stronger ER binding affinity than lignans (reviewed in Ref.8). Many studies have been conducted among Asian populations25, 27, 29–32 where soy intake is common, and isoflavones have been the primary class of phytoestrogens examined. In North American populations, lignans are the primary class of phytoestrogens consumed,2, 42 and our data demonstrate that lignans are also the major phytoestrogen consumed during adolescence. Isoflavone intake in our study was very low, particularly during adolescence, and differences between adult and adolescent intakes may be partially accounted for by the number of isoflavone foods in each FFQ (119 vs. 55, respectively, of which 15 and 4 were soy foods).
Although the literature describing heterogeneity by ERPR status is inconsistent, the literature on phytoestrogen intake and breast cancer risk suggests a moderate inverse association overall, independent of ERPR (as reviewed in Refs.5, 7 and8). Unexpectedly, we observed positive associations between adult isoflavone and total phytoestrogen intake among ER−PR− cases; however, there was no evidence of a positive association among other ERPR subgroups. It is possible that this may be a spurious finding, although some animal studies43, 44 found isoflavones-stimulated cancer growth, albeit for ER+ tumors. This positive association is not supported by the epidemiological literature (as reviewed in Ref.45).
Significant reductions in breast cancer risk were observed for adolescent intake of isoflavones, lignans and total phytoestrogen among all ERPR breast cancer subgroups, though most notably among ER+PR+. This contributes to the growing body of literature that suggests adolescent exposures may be important for cancer prevention.35, 46, 47 Adolescence reflects a critical period of breast development, between onset of puberty and first pregnancy, when breast tissue is undifferentiated and animal studies have shown that phytoestrogen intake in early life may promote breast tissue differentiation.48, 49 Although none of these studies stratified risk by ERPR status, the few studies that have evaluated early life phytoestrogen or soy food exposure have also all reported inverse associations with statistically significant risk estimates ranging from 0.40 to 0.72.11–15 Our results suggest that the inverse associations between adolescent intake of isoflavones, lignans and total phytoestrogens may be relevant to all breast cancer subgroups, though most consistently for ER+PR+ tumors among postmenopausal women. Our finding of significant associations at such low levels of intake is not consistent with the hypothesis among adults of an absolute threshold (for isoflavones) based on observations of moderately reduced risk at intakes of 10 mg/day, maximally reduced risk at ≥20 mg/day, but no association at <1 mg/day, typical among non-Asians.7, 50 However, our finding possibly supports the hypothesis that adolescent exposure may be important even at low levels of intake.9, 50
Despite our relatively large sample size, minimal heterogeneity by ERPR tumor status may have been detected due to limited statistical power, particularly among premenopausal women (premenopausal cases n = 736 and controls n = 1,211; postmenopausal cases n = 1,659 and controls n = 2,154). It is possible that we observed the most significant associations between postmenopausal ER+PR+ because this was the largest sample size. Another limitation is the retrospective nature of the case–control study design, which introduces the potential for measurement error as study participants need to recall earlier dietary intake. This is of particular concern for adolescent diet, because many of the soy foods are consumed relatively infrequently and may not have been routinely available when these women were adolescents (30–40 years ago). It is possible that the intake of these foods (e.g. tofu and soy milk) during adolescence may be somewhat easier to recall than common food items. Lignans are found more widely in commonly consumed plant foods, but the primary contributor to total lignan intake in our adult population was flaxseed,36 which is also consumed infrequently. The adolescent FFQ used in our study has not been validated and given the long recall period it would not be possible to do so using traditional methods such as 24-hr recalls or biomarkers. Although measurement error is a potential concern, recall bias is likely minimal as it is not expected that cases and controls would differentially report their intake of these foods. Individual differences in metabolism and bioavailability cannot be taken into account using a FFQ.
In conclusion, the association between breast cancer risk and phytoestrogen intake, during adulthood seems to be minimal to null and independent of hormone receptor status; with high lignan intake possibly associated with reduced breast cancer risk across all ERPR subgroups for all women and postmenopausal women, although statistical significance was not reached. Phytoestrogen intake during adolescence, however, was significantly associated with a reduction in postmenopausal breast cancer risk, in particular for ER+PR+ breast cancer. In addition to altering estrogen metabolism or competitively binding to ERs, phytoestrogens may have other anticarcinogenic properties including the inhibition of angiogenesis, promotion of cell proliferation and antioxidative and anti-inflammatory actions,2 and these alternate biologic pathways may explain why associations are independent of ERPR status. Our results for adolescent intake by ERPR tumor status are novel, and while there was no marked heterogeneity in risk by subgroup, we consistently observed inverse associations between adolescent intake of isoflavones, lignans and total phytoestrogens and breast cancer risk. Phytoestrogen intake is modifiable through diet and, given that there are so few well-established modifiable risk factors for breast cancer, it could potentially be important for primary prevention. Future studies are needed to replicate our findings and further elucidate the mechanism by which phytoestrogens may reduce breast cancer risk, especially during early life, with the ultimate goal of developing population-based prevention strategies.
The authors thank the study staff, Noori Chowdhury and Leah Palma, for their dedication to this study. This research was funded by the Canadian Cancer Society (CCSRI grant #01985 to M.C.) and the Canadian Breast Cancer Research Alliance with special funding support of the Canadian Breast Cancer Foundation Ontario Chapter (CBCRA grant #13572 to M.C.).