The association of soy food consumption with the risk of subtype of breast cancers defined by hormone receptor and HER2 status

Abstract

Soy-food intake has previously been associated with reduced breast cancer risk. Epidemiological evidence for subgroups of breast cancer, particularly by menopausal and hormone receptor status, is less consistent. To evaluate the role of hormone receptor and menopausal status on the association between soy-food intake and breast cancer risk, we measured usual soy-food intake in adolescence and adulthood via food frequency questionnaire in 70,578 Chinese women, aged 40-70 years, recruited to the Shanghai Women’s Health Study (1996-2000). After a median follow-up of 13.2 years (range:0.01-15.0), 1,034 incident breast cancer cases were identified. Using Cox models, we found that adult soy intake was inversely associated with breast cancer risk (hazard ratio-HR) for fifth versus first quintile soy protein intake=0.78; 95% confidence interval (CI):0.63-0.97). The association was predominantly seen in premenopausal women (HR=0.46; 95% CI:0.29-0.74). Analyses further stratified by hormone receptor status showed that adult soy intake was associated with significantly decreased risk of ER+/PR+ breast cancer in postmenopausal women (HR=0.72; 95% CI:0.53-0.96) and decreased risk of ER−/PR− breast cancer in premenopausal women (HR=0.46; 95% CI:0.22-0.97). The soy association did not vary by HER2 status. Furthermore, we found that high soy intake during adulthood and adolescence was associated with reduced premenopausal breast cancer risk (HR=0.53; 95% CI:0.32-0.88; comparing third versus first tertile) while high adulthood soy intake was associated with postmenopausal breast cancer only when adolescent intake was low (HR=0.63; 95% CI:0.43-0.91). Our study suggests that hormonal status, menopausal status, and time window of exposure are important factors influencing the soy-breast cancer association.

INTRODUCTION

Breast cancer is the leading cancer in women worldwide ¹. Low rates of breast cancer among women in Asian countries and its rapid increase following emigration to western cultures has led to lifestyle factors, particularly dietary factors, being postulated as an explanation for this pattern ^2-4. Epidemiological studies have shown that soy food consumption may decrease the risk of breast cancer and these data were summarized in several recent meta-analyses/systematic reviews ^5-8. However, the conclusions from these studies are mixed with one suggesting that the association may be limited to premenopausal women ⁵, one reporting a stronger effect in postmenopausal women ⁶, and two concluding no modifying effect of menopausal status ^{7, 8}. The majority of previous studies investigating the association between soy food intake and breast cancer risk have been case-control studies which are subject to recall and selection biases. The few reports from cohort studies have also been inconsistent^{9, 10,11} and were limited by small numbers of breast cancer cases ^{9, 11} or did not include all sources of commonly consumed soy foods ¹¹.

Isoflavones, which have a chemical structure similar to 17β-estradiol, are the most abundant phytoestrogen in soy food and have been shown to compete with endogenous estrogen to bind estrogen receptors ¹². Isoflavones have weak estrogenic potency and endogenous estrogen level may affect their action; isoflavones exert estrogenic-like effects in estrogen-deprived environments and anti-estrogenic effects when endogenous estrogen levels are high ¹³. Experimental studies have suggested that soy may be protective against hormone-related cancers¹⁴.

Breast cancer is a heterogeneous disease and biological differences in subtypes depending on the expression of receptors, such as estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor-2 (HER2) have been well recognized. Several risk factors for subtypes of breast cancer also differ; compared to receptor-negative tumors, ER+/PR+ tumors are more strongly associated with some reproductive factors, such as parity, timing of births, and age at menopause ^{15, 16}. Due to the ability of isoflavones to bind estrogen receptors, the association between soy intake and breast cancer risk may vary by hormone receptor status of the tumor, though few studies have assessed this association ^{8, 14}. Of the 8 case-control studies we found in the literature that considered hormone receptor status as a potential effect modifier, 3 reported that the protective association between soy and breast cancer was consistent across all subtypes of ER/PR tumors ^17-19. Two studies found a greater reduction in risk in those with ER+ tumors compared to ER− tumors ²⁰ or ER+/PR+ tumors compared to other ER/PR status ²¹, while another study found reduced risk for ER+/PR+ and ER−/PR− tumors but not mixed subtypes ²². One study only observed a decreased risk of ER+/PR+ breast cancer with high adolescent intake ²³. To our knowledge, only one study evaluated the association by HER2 status, and it reported null results for all types, i.e. ER+, HER2−, and ER+/PR+/HER2-tumors ²⁴. We only identified one prospective study that investigated soy intake and breast cancer risk by ER and/or PR status of tumors and reported no effect modification among postmenopausal women²⁵. This study, however, was limited by small sample size and did not report the results for premenopausal women by hormone receptor status²⁵.

We have previously reported the association of adolescent and adult intake of soy with breast cancer risk in the Shanghai Women’s Health Study (SWHS).¹⁰ In the present study, we updated the association between soy food intake and breast cancer risk with a longer follow-up time and provided new information on the association by ER, PR, and HER2 status of the breast cancer.

MATERIALS AND METHODS

Study Population

This study utilized the resources generated from the population-based Shanghai Women’s Health Study (SWHS) for which the methodology has been previously described ^{26, 27}. Briefly, 74,942 women aged 40 to 70 years were recruited to the study from seven urban communities in Shanghai, China between 1996 and 2000, with a participation rate of nearly 93%. Information on demographic and lifestyle factors, as well as participant characteristics including reproductive factors, menstrual history, and medical history, was collected using structured questionnaires through in-person interviews by trained interviewers. Using a standard protocol, anthropomorphic measures, including height, weight, and waist and hip circumferences, were taken at the baseline interview. Self-reported weight information was collected at the 3^rd and 4^th follow-up interviews. Clinical information and tumor characteristics, including hormone receptor status (ER, PR, and HER2), were extracted from medical charts. Written, informed consent was obtained from all participants and the institutional review boards at all participating institutions approved the study.

Ascertainment of Breast Cancer Cases

Active surveys were conducted every 2-3 years in the SWHS to collect information on occurrence of cancer and other chronic diseases with the following response rates: 1^st in-person follow-up, 99.8%; second, 98.7%; third, 96.7%, and fourth, 92%. Annual record linkage to the population-based Shanghai Cancer Registry was used to identify cancer cases. Women with a first cancer diagnosis of breast cancer (ICD-9 code 174) were defined as cases for this study ²⁸. Cancer diagnosis was verified through in-person visits and review of medical charts obtained from the diagnostic hospital.

Dietary Assessment

Habitual dietary intake was measured using a validated food frequency questionnaire (FFQ) at baseline and 2-3 years later at the 1^st follow-up interview; the latter was completed by 92% of cohort members who were alive at the time of the 1^st follow-up. The FFQ used in this study was a comprehensive dietary assessment and was designed to measure the consumption of soy foods commonly consumed in Shanghai, including soymilk, tofu, fresh soy beans, and other soy foods ²⁷. Estimated energy and nutrient intakes, including soy and isoflavones, were calculated by summing products of food intake amount multiplying the nutrient content of the specific food item based on the Chinese Food Composition Tables 2002 ²⁹.

Statistical Analysis

Women were excluded for the present study if they reported a previous cancer diagnosis at baseline (n=1,598), extreme baseline total energy intakes (<500 or ≥3500 kcal/day, n=125), or prior or current hormone replacement therapy (HRT) use (n=2,585) resulting in a total sample size of 70,578 women.

Adult habitual dietary intake of soy food was assessed from the baseline and 1^st follow-up surveys. Analyses performed used two methods for estimating adult soy intake: (1) the soy protein and soy isoflavone intake based on the baseline survey data only and (2) an average soy protein and isoflavone intake calculated by averaging the soy intake levels from the baseline and 1^st follow-up surveys. The latter method was not used for women who developed cancer, diabetes, myocardial infarction, or stroke between the baseline and follow-up surveys because the diagnosis of these conditions may have modified some women’s dietary habits. For these women, the baseline soy measure was used for all analyses of adult intake. Adolescent soy intake was assessed at the baseline interview; women were asked to report their intake of commonly consumed soy foods between the ages of 13 and 15. Baseline variables were evaluated for their association with breast cancer and soy protein intake using generalized linear models for continuous variables and chi-square tests for categorical variables.

Soy intake was analyzed using quintiles when sample size allowed, where the lowest quintile served as the reference group, in order to assess the pattern of association with a greater range of intake. Using Cox proportional hazards regression models, the hazard ratios (HR) and 95% confidence intervals (CI) for the association between soy intake and breast cancer risk were estimated. We used age as the time scale for the analyses, where age at enrollment was defined as the entry time and age at diagnosis or censoring was defined as the exit time. Censoring was defined as date of death, last date of follow-up, or the latest date of record linkage. The covariates included in the model were age at enrollment, body mass index (BMI), age at first live birth, physical activity (yes/no), education, family history of breast cancer, season of recruitment, total energy intake, and menopausal status. Menopausal status, which was updated at follow-up interviews, was treated as a time-varying covariate in the analyses. Menopausal status was further evaluated as an effect modifier by stratifying results by menopausal status. Categorical soy intake variables were treated as ordinal variables to evaluate the linear trend.

Further analyses stratified by hormonal receptor status were performed. For these analyses, soy intake variables were categorized into tertiles of intake for analysis in each stratified group. The joint effect of average adult and adolescent soy protein intake was evaluated using tertiles of intake. These results were additionally stratified by menopausal status. Results from analyses on soy protein and soy isoflavones were very similar, therefore only the former is included in the tables (see supplemental tables 2 and 3 for soy isoflavone analyses). All statistical analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC).

RESULTS

Over a median follow-up time of 13.2 years, 1,034 incident breast cancer cases were identified. The average (standard deviation) age at diagnosis was 59.6 (9.3) years. Consistent with the literature on the established risk factors for breast cancer ³⁰ and our previous study,¹⁰ breast cancer cases were more likely to have a higher education, higher income, younger age at menarche, older age at menopause, longer total years of menstruation, older age at first live birth, fewer number of live births, and higher BMI compared to non-cases (Supplemental Table 1). Women in the highest quintile of soy protein (median intake 16.4 grams per day) were more likely to have a higher education, to exercise regularly, to be older at menopause onset, to be younger at age of first live birth, to have a family history of breast cancer, to have a higher BMI, to have a higher waist-to-hip ratio, and to consume more overall calories as well as more meats and vegetables. These factors were included in the analysis as covariates. The rate of cigarette smoking was very low in this cohort and was not related to breast cancer risk. Additionally, soy food intake was not related to weight change during the follow-up period.

Table 1 shows the association between averaged adult and adolescent soy protein intake and breast cancer risk overall and by menopausal status at diagnosis. These updated results are similar to those previously published.¹⁰ Overall, soy protein was associated with a modest decrease in risk of breast cancer. Compared to the lowest quintile of soy intake, the highest quintile of averaged soy protein intake during adulthood was associated with reduced breast cancer risk (HR=0.78; 95% confidence interval (CI): 0.63, 0.97; P_trend=0.007); the association was predominantly seen in premenopausal women (HR=0.46; 95% CI: 0.29, 0.74; P_trend=0.004). No significant association was observed among postmenopausal women. Additionally adjusting for adolescent soy intake did not materially change the observed associations (data not shown). Analyses based on baseline soy intake information showed very similar results (data not shown). Adolescent soy intake was not significantly associated with breast cancer risk among premenopausal women; although the HR for the fifth quintile compared to the first quintile was in the direction of reduced risk (HR=0.69; 95% CI: 0.45, 1.04; P_trend=0.08). No association between adolescent soy intake and postmenopausal breast cancer risk was observed.

Table 1.

Adult and Adolescent Soy Protein Intake and Breast Cancer Risk By Menopausal Status

	Soy Protein Intake Level (RRa,b (95%CI))
	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5	Ptrend
ADULT INTAKE (median(g/day))	3.5	6.0	8.2	10.9	16.0
Overall	1.00 (reference)	1.01 (0.83, 1.22)	1.00 (0.82, 1.21)	0.87 (0.71, 1.06)	0.78 (0.63, 0.97)	0.007
N(cases)	212	222	222	196	182
Premenopausalc (N(cases)=273)	1.00 (reference)	0.97 (0.69, 1.36)	0.86 (0.60, 1.24)	0.98 (0.68, 1.42)	0.46 (0.29, 0.74)	0.004
N(cases)	68	65	54	59	27
Postmenopausalc (N(cases)=761)	1.00 (reference)	1.03 (0.82, 1.30)	1.06 (0.84, 1.33)	0.83 (0.65, 1.06)	0.90 (0.71, 1.16)	0.15
N(cases)	144	157	168	137	155
ADOLESCENT INTAKE (median(g/day))	1.8	3.9	6.2	9.2	15.6
Overall	1.00 (reference)	1.04 (0.86, 1.26)	1.12 (0.92, 1.35)	1.04 (0.86, 1.26)	0.95 (0.77, 1.16)	0.43
N(cases)	198	213	226	212	185
Premenopausalc	1.00 (reference)	0.77 (0.54, 1.11)	0.96 (0.68, 1.36)	0.74 (0.51, 1.07)	0.69 (0.45, 1.04)	0.08
N(cases)	63	55	69	51	35
Postmenopausalc	1.00 (reference)	1.17 (0.93, 1.47)	1.19 (0.94, 1.49)	1.18 (0.94, 1.49)	1.06 (0.84, 1.34)	1.00
N(cases)	135	158	157	161	150

^aCox proportional hazards models adjusted for age, body mass index, age at first live birth, physical activity, education, family history of breast cancer, season of recruitment, and menopause (time-varying) were used for analyses

^bAdult intakes additionally adjusted for total energy intake and juvenile intakes adjusted for total juvenile rice intake

^cStratified based on menopausal status at breast cancer diagnosis

When stratified by hormone receptor status, HRs for high adult soy intake were below 1.0 for most subtypes, though only one reached statistical significance (Table 2). Compared to women in the lowest tertile of adult average soy protein intake, there was a significantly decreased risk of ER+/PR+ breast cancer among all women (HR=0.75; 95% CI: 0.58, 0.98; P_trend=0.03), and the association was only significant in postmenopausal women (HR=0.72; 95% CI: 0.53, 0.96; P_trend=0.02). A significantly decreased risk of ER−/PR− breast cancer was observed for premenopausal women (HR=0.46; 95% CI: 0.22, 0.97; P_trend=0.04) in the highest tertile of soy intake. Analyses using baseline adult soy intake showed very similar results (data not shown). HRs associated with higher adult soy intake for HER2+ and HER2− breast cancer were both below 1.0 but none of the point estimates were statistically significant.

Table 2.

Adult and Adolescent Soy Protein Intake and Breast Cancer Risk By Menopausal Status and Receptor Status

	Adult Soy Protein Intake Level (RRa,b (95%CI))				Adolescent Soy Protein Intake Level (RRa,b (95%CI))
	Tertile 1	Tertile 2	Tertile 3		Tertile 1	Tertile 2	Tertile 3
	median=4.5g/day	median=8.2g/day	median=13.5g/day	P trend	median=2.6g/day	median=6.2g/day	median=12.5g/day	P trend
OVERALL
ER+ (N(cases)=550)	1.00 (ref)	0.96 (0.78, 1.18)	0.86 (0.68, 1.07)	0.16	1.00 (ref)	1.28 (1.04, 1.57)	1.09 (0.87, 1.35)	0.76
ER− (N(cases)=288)	1.00 (ref)	1.25 (0.95, 1.66)	0.87 (0.63, 1.21)	0.32	1.00 (ref)	1.13 (0.86, 1.49)	0.98 (0.73, 1.32)	0.77
ER+/PR+ (N(cases)=409)	1.00 (ref)	0.91 (0.72, 1.15)	0.75 (0.58, 0.98)	0.03	1.00 (ref)	1.32 (1.03, 1.68)	1.16 (0.90, 1.48)	0.46
ER−/PR− (N(cases)=246)	1.00 (ref)	1.20 (0.89, 1.62)	0.83 (0.59, 1.18)	0.25	1.00 (ref)	1.12 (0.82, 1.52)	1.12 (0.82, 1.54)	0.52
ER+/PR− (N(cases)=124)	1.00 (ref)	1.18 (0.75, 1.85)	1.27 (0.78, 2.05)	0.35	1.00 (ref)	1.05 (0.69, 1.59)	0.81 (0.51, 1.28)	0.32
HER2+ (N(cases)=158)	1.00 (ref)	1.04 (0.72, 1.51)	0.79 (0.51, 1.21)	0.26	1.00 (ref)	1.23 (0.85, 1.77)	0.79 (0.52, 1.19)	0.17
HER2− (N(cases)=434)	1.00 (ref)	0.92 (0.73, 1.16)	0.83 (0.65, 1.07)	0.15	1.00 (ref)	1.11 (0.88, 1.40)	1.13 (0.89, 1.43)	0.37
PREMENOPAUSAL
ER+ (N(cases)=135)	1.00 (ref)	1.00 (0.66, 1.50)	0.91 (0.57, 1.44)	0.67	1.00 (ref)	1.10 (0.74, 1.63)	0.86 (0.55, 1.35)	0.46
ER− (N(cases)=88)	1.00 (ref)	0.91 (0.56, 1.46)	0.60 (0.33, 1.11)	0.11	1.00 (ref)	1.31 (0.81, 2.11)	0.77 (0.43, 1.39)	0.30
ER+/PR+ (N(cases)=103)	1.00 (ref)	0.92 (0.58, 1.48)	0.91 (0.54, 1.52)	0.72	1.00 (ref)	1.01 (0.64, 1.61)	1.01 (0.62, 1.66)	0.97
ER−/PR− (N(cases)=68)	1.00 (ref)	0.85 (0.50, 1.44)	0.46 (0.22, 0.97)	0.04	1.00 (ref)	1.30 (0.73, 2.30)	1.09 (0.58, 2.05)	0.89
ER+/PR− (N(cases)=30)	1.00 (ref)	1.32 (0.56, 3.12)	1.04 (0.38, 2.83)	0.97	1.00 (ref)	1.48 (0.68, 3.25)	0.33 (0.09, 1.21)	0.08
HER2+ (N(cases)=49)	1.00 (ref)	1.07 (0.56, 2.05)	0.73 (0.33, 1.61)	0.44	1.00 (ref)	1.43 (0.75, 2.72)	0.64 (0.28, 1.47)	0.22
HER2− (N(cases)=122)	1.00 (ref)	0.90 (0.58, 1.38)	0.93 (0.58, 1.51)	0.78	1.00 (ref)	0.84 (0.55, 1.27)	0.84 (0.54, 1.32)	0.49
POSTMENOPAUSAL
ER+ (N(cases)=415)	1.00 (ref)	0.95 (0.75, 1.21)	0.84 (0.65, 1.09)	0.19	1.00 (ref)	1.35 (1.06, 1.72)	1.16 (0.90, 1.49)	0.48
ER− (N(cases)=200)	1.00 (ref)	1.52 (1.07, 2.15)	1.05 (0.71, 1.57)	0.90	1.00 (ref)	1.05 (0.75, 1.48)	1.06 (0.75, 1.50)	0.76
ER+/PR+ (N(cases)=306)	1.00 (ref)	0.90 (0.69, 1.19)	0.72 (0.53, 0.96)	0.02	1.00 (ref)	1.45 (1.09, 1.92)	1.21 (0.90, 1.62)	0.44
ER−/PR− (N(cases)=178)	1.00 (ref)	1.44 (1.00, 2.08)	1.04 (0.69, 1.58)	0.91	1.00 (ref)	1.05 (0.73, 1.51)	1.13 (0.78, 1.62)	0.52
ER+/PR− (N(cases)=94)	1.00 (ref)	1.14 (0.66, 1.94)	1.35 (0.78, 2.33)	0.28	1.00 (ref)	0.90 (0.55, 1.48)	0.94 (0.57, 1.54)	0.84
HER2+ (N(cases)=109)	1.00 (ref)	1.02 (0.65, 1.61)	0.81 (0.49, 1.36)	0.40	1.00 (ref)	1.12 (0.72, 1.75)	0.83 (0.51, 1.34)	0.37
HER2− (N(cases)=312)	1.00 (ref)	0.94 (0.71, 1.23)	0.80 (0.60, 1.08)	0.14	1.00 (ref)	1.26 (0.95, 1.66)	1.26 (0.95, 1.68)	0.15

^aCox proportional hazards models adjusted for age, body mass index, age at first live birth, physical activity, education, family history of breast cancer, season of recruitment, and menopause (time-varying) were used for analyses

^bAdult intakes additionally adjusted for total energy intake and juvenile intakes adjusted for total juvenile rice intake

^cStratified based on menopausal status at breast cancer diagnosis

Among premenopausal women, only high soy protein intake in both adulthood and adolescence was significantly associated with the risk (HR=0.53; 95% CI: 0.32, 0.88) (Table 3). Among postmenopausal women, high soy protein intake during adulthood was associated with a significantly decreased risk of breast only when adolescent intake was low (HR=0.63; 95% CI: 0.43, 0.91). High adolescent intake alone or high adult and adolescent intake was not significantly associated with the risk among postmenopausal women. None of the tests for multiplicative interaction between adult and adolescent soy intake reached statistical significance. Among breast cancer patients, those who had high soy food intake during both adolescence and adulthood had a later age of cancer diagnosis (mean age=61.3, SD=8.2) than cases who had low level soy food intake during both periods (mean age=57.5, SD=9.1) (p=0.0001). High consumption during adulthood alone (mean age=59.6, SD=8.5) and during adolescence alone (median age=61.7, SD=11.0) were also related to delayed age at cancer diagnosis.

Table 3.

Joint Effect of Adult and Adolescent Soy Protein Intake on Breast Cancer Risk

	Adult Average Soy Protein Intake (RRa (95%CI))
Juvenile Soy Protein Intake	≤6.32	6.33-10.43	≥10.44
Premenopausalb (N(cases)=273)
≤4.24	1.00 (reference)	0.62 (0.38, 1.00)	0.56 (0.31, 1.02)
4.25-8.61	0.86 (0.58, 1.28)	0.86 (0.57, 1.29)	0.78 (0.49, 1.25)
≥8.62	0.56 (0.31, 1.00)	0.80 (0.51, 1.26)	0.53 (0.32, 0.88)
Postmenopausalb (N(cases)=761)
≤4.24	1.00 (reference)	0.96 (0.72, 1.29)	0.63 (0.43, 0.91)
4.25-8.61	1.07 (0.81, 1.43)	1.08 (0.82, 1.43)	0.94 (0.69, 1.26)
≥8.62	0.94 (0.67, 1.32)	1.01 (0.76, 1.35)	0.98 (0.74, 1.29)

^aCox proportional hazards models adjusted for age, body mass index, age at first live birth, physical activity, education, family history of breast cancer, season of recruitment, total adult energy, total juvenile rice intake, and menopause (time-varying) were used for analyses

^bStratified based on menopausal status at breast cancer diagnosis

DISCUSSION

In this population-based cohort study, we found a decreased risk of breast cancer with high soy food intake among all women, particularly in premenopausal women, consistent with several previous reports, ^{5, 31, 32} including our own previous study.¹⁰ We found that the significant association with ER+/PR+ breast cancer was mainly seen among postmenopausal women, while the association with ER−/PR− breast cancer was predominantly confined to premenopausal women. Furthermore, among premenopausal women, high soy intake during both adult and adolescent periods was associated with the biggest reduction in risk; however, among postmenopausal women, high adult soy intake was only associated with a decreased risk of breast cancer when adolescent intake was low. We did not observe the soy and breast cancer risk association to vary by HER2 status. These results suggest that both adult and adolescent soy food intake may be associated with a reduction in risk of breast cancer, and that this association may vary by menopausal status and ER/PR status.

Several systematic reviews and meta-analyses have been conducted to summarize the literature on the association between soy intake and breast cancer risks ^5-8. One meta-analysis, which included 12 case-control studies and 6 cohort or nested case-control studies, concluded that the association was strongest in premenopausal women; however, many of the included studies were not designed to assess the association between breast cancer and soy intake and the studies differed in the level of control for confounding ⁵. Another meta-analysis restricted to prospective studies observed the strongest association in postmenopausal women; however, the studies included were highly heterogeneous which resulted in difficult interpretations of the results ⁶. A more recent meta-analysis additionally considered the region and type of study conducted and found an association between soy intake and breast cancer risk in Asian countries, but not Western countries, and found no modifying effect of menopausal status; however, there was heterogeneity among the included studies and varying dietary intake measures ⁷. In our study we observed that the association between soy intake and reduced breast cancer risk was predominantly seen in premenopausal women. This finding is supported by studies that have shown that genistein, an isoflavone found in soy foods, inhibits growth of breast cancer cells in culture when endogenous estrogen levels are high³³. None of the meta-analyses evaluated the potential modifying effect of hormone receptor status although this has been evaluated in a few previous studies with inconsistent findings. The majority of the previous studies were case-control studies and had several limitations, such as including a small number of breast cancer cases available for analysis ^{18, 20}, low soy intake in the population studied ^{17, 20, 23, 25}, lack of a comprehensive soy intake assessment ^{18, 24}, or recall bias ^17-24. The large sample size in our study allowed for analyses stratified by both ER/PR status and menopausal status. Our findings that soy intake may be associated with a reduction in risk of different breast cancer subtypes in premenopausal and postmenopausal women offers a potential explanation for the inconsistent findings of previous studies.

Soy food is a rich source of isoflavones, which are structurally similar to estrogens; studies have shown that isoflavones compete with endogenous estrogen for the estrogen receptor making it biologically plausible that isoflavones protect against breast cancer development ¹². Other anti-cancer properties of soy food, such as anti-angiogenic effects, anti-proliferative effects, antioxidative DNA topoisomerase I and II inhibition, as well as tyrosine kinases, and proteases effects ^{12, 14}, may also explain the inverse association between isoflavones and breast cancer risk, particularly for estrogen receptor negative breast cancer.

The reduced risk of breast cancer associated with soy intake has primarily been observed in Asian populations, with studies in Western populations generally showing no association,⁶ which is likely due to the lower level of soy food consumption as the soy intake levels in Asian populations are considerably higher than that of Western populations. On the other hand, several studies have also suggested that early or lifetime soy exposure may explain this discrepancy ^{10, 34, 35}. In our study, we found only adolescent and adult soy intake in conjunction were significantly associated with reduced breast cancer risk among premenopausal women, suggesting a relatively long period of exposure may be required for soy food consumption to exert its protective effect. Although our finding that only high adult soy food intake alone was associated with postmenopausal breast cancer risk appears to be contradictory to the long-term exposure hypothesis at first glance, it is likely that soyfood, like most cancer preventive agents, would only be able to prevent some but not all breast cancer in women. Under the long-term exposure hypothesis, these “soy responsive” breast cancers would be prevented when women had sufficient exposure, e.g., high adolescent plus high adulthood exposure for pre-menopausal women, and high adulthood exposure for postmenopausal women. Because “soy responsive” breast cancer would have been already prevented during the pre-menopausal period for women with both high adolescent and adulthood soyfood intake, no additional benefit would be observed among post-menopausal women. This may also explain the null association observed in western populations as soy food consumption has only become popular in recent years. It is noteworthy that we found that soy food consumption during adolescence and adulthood were both related to late age at cancer diagnosis among breast cancer patients. This observation adds to the evidence supporting a causal association between soy food intake and breast cancer risk. More research is warranted to understand the underlying biological mechanisms.

Diets high in soy food may be associated with healthier dietary and lifestyle behaviors. These dietary and other associated lifestyle factors which are associated with soy intake and breast cancer risk, could contribute to the observed inverse association seen in our population. We have carefully adjusted for a wide range of dietary and other lifestyle variables in our study although residual confounding can’t be completely ruled out. Additional adjustment for dietary pattern in our study did not change the study results.

For some subgroup analyses, particularly those by HER2 status and mixed ER/PR status, the statistical power of the study was low. We had to analyze the data using tertile categorization which prevented us from investigating more extreme soy food intake in these specific groups of women. Additionally, some measurement errors, with the ER, PR, and HER2 status information obtained from multiple hospitals and a long lag time for recalling adolescent soy food intake, may have biased our results towards the null.

The strengths of our study include its prospective design, large sample size, and low lost to follow-up rate. Adult soy intake was assessed using a validated, culturally appropriate FFQ designed to capture usual soy intake in our population. Multiple dietary assessments (at baseline and at the 1^st follow-up survey) improved our soy intake measurement.

In conclusion, we found that soy food intake was inversely associated with breast cancer risk, particularly in premenopausal women. The association between soy food intake and breast cancer risk may be modified by menopausal status and hormone receptor status.

Novelty and Impact: Epidemiological evidence is inconsistent and limited with regard to whether the soyfood and breast cancer risk association differ by hormone receptors and HER2 status. In a large cohort study, we found that soyfood intake was associated with both ER/PR positive and negative breast cancer risk but the association differed by menopausal status. No modification by HER2 status was observed. These results suggest that soyfood may influence breast cancer risk via multiple mechanisms.

Abbreviations

ER

estrogen receptor

PR

progesterone receptor

HER2

human epidermal growth factor-2

BMI

body mass index