Fruit, Vegetable, and Animal Food Intake and Breast Cancer Risk by Hormone Receptor Status

Abstract

Background

The effects of diet on breast cancer are controversial and whether the effects vary with hormone receptor status has not been well investigated. This study evaluated the associations of dietary factors with risk for breast cancer overall and by hormone receptor status of tumors among Chinese women.

Methods

The Shanghai Breast Cancer Study, a large, population-based, case-control study, enrolled 3,443 cases and 3,474 controls in 1996–1998 (phase I) and 2002–2004 (phase II); 2,676 cases had ER and PR data. Dietary intake was assessed using a validated, quantitative, food frequency questionnaire (FFQ). Odds ratios (ORs) and 95% confidence intervals (95% CI) were derived from multivariate, polychotomous, unconditional logistic regression models.

Results

Total vegetable intake was inversely related to breast cancer risk, with an adjusted OR for the highest quintile of 0.80 (95% CI = 0.67–0.95; P trend=0.02). Reduced risk was also related to high intake of allium vegetables (P trend = 0.01) and fresh legumes (P trend = 0.0008). High intake of citrus fruits and rosaceae fruits were inversely associated with breast cancer risk (P trend = 0.003 and P trend = 0.004, respectively), although no consistent association was seen for total fruit intake. Elevated risk was observed for all types of meat and fish intake (all P trend <0.05), while intakes of eggs and milk were associated with a decreased risk of breast cancer (both P trend <0.05). There was little evidence that associations with dietary intakes varied across the four tumor subtypes or between ER+/PR+ and ER−/PR− tumors (P for heterogeneity >0.05).

Conclusion

Our results suggest that high intake of total vegetables, certain fruits, milk, and eggs may reduce the risk of breast cancer, while high consumption of animal-source foods may increase risk. The dietary associations did not appear to vary by ER/PR status.

Introduction

Breast cancer is a hormone-related malignancy, and diet may influence the risk of cancer through its effects on hormone levels, growth factors, and anti-oxidation (1–4). The role of dietary factors, which are potentially modifiable, in the cause of breast cancer has been extensively investigated (1, 5–6). However, the results have been inconsistent (1, 5–7). In 2007, the World Cancer Research Fund concluded in their publication “Food, Nutrition, and the Prevention of Cancer: A Global Perspective” that the scientific data on the relation of breast cancer and dietary intake were too limited to reach a conclusion (2).

Evidence is accumulating to suggest that the effects of hormone-related factors, such as parity and BMI, are modified by estrogen receptor (ER) and progesterone receptor (PR) status (8). It has been suggested that associations between dietary consumption and breast cancer risk may depend on menopausal status (9–10) or the ER and PR status of the breast cancer (6, 9). However, results from the few published epidemiologic studies addressing the effects of food groups on breast cancer by ER and PR status have been inconsistent (9, 11–18).

We evaluated associations between dietary intakes and breast cancer and investigated whether the dietary associations varied by hormone receptor status using data from the Shanghai Breast Cancer Study (SBCS), a large-scale, population-based, case-control study.

Subjects and Methods

Study population

The SBCS is a two-stage study, including 3,443 breast cancer cases and 3,474 healthy controls, who were recruited in two phases, 1996–1998 (phase I or SBCS I) and 2002–2005 (phase II or SBCS II), in Shanghai, China. Detailed study methods have been described elsewhere (19). Eligibility criteria for the study participants were as follows: permanent resident of urban Shanghai, aged 25–70 years, and no prior history of any cancer. Through the rapid case-ascertainment system of the Shanghai Cancer Registry, 1,602 eligible breast cancer cases were identified during phase I and 2,388 cases were identified during phase II. Controls, 1,724 for phase I and 2,724 for phase II, were randomly selected from women in the Shanghai Resident Registry and frequency-matched to cases by age in 5-year intervals. Of eligible cases and controls, 3,443 (86.3%) and 3,474 (78.1%) participated in the study. All participants provided written informed consent, and the study was approved by the Institutional Review Boards at each of the institutes involved.

Data Collection

All study participants were interviewed in person by trained interviewers using a structured questionnaire that covered detailed information on demographic characteristics and known or suspected risk factors, including dietary information, reproductive factors, menopausal history, and family history of breast cancer in first-degree relatives. Each participant was measured for weight, height, and circumferences of the waist and hips at the time of the interview according to standard procedures. Body mass index (BMI) was calculated from these data. Dietary information was collected using a food-frequency questionnaire (FFQ) that accounts for over 85% of foods consumed in Shanghai. The validity and reliability of the quantitative FFQ used in this study was assessed through two FFQ surveys and one 24-h dietary recall survey (20). The FFQ included 76 food items/groups, a total of 30 fresh vegetables, 8 fruits, 19 animal foods, and 2 staple foods. For each food item, each participant was first asked to report her usual frequency of intake (per day, week, month, year or never), followed by a question on the amount consumed in liang (1 liang = 50 g) per unit of time in the previous 5-year period, ignoring recent changes in usual dietary intake. For seasonal foods, the study participants were asked to describe their consumption based on market availability and total months per year. Average daily intakes were estimated by calculating the percentage of months that a food was on the market over a 1-year period.

Individual fruits and vegetables were categorized into 11 food groups: total vegetables, dark-green vegetables, carrots and tomatoes, cruciferous vegetables, fresh legumes, allium vegetables, melons, other vegetables, total fruits, citrus fruits, and rosaceae fruits. Meats were classified into the following four food groups: red meat, poultry, fresh-water fish, and marine fish. Eggs and milk were also included in the analysis. Intake for each food group was obtained by summing the grams per day for each food item in the group.

Detailed information on cancer estrogen receptor (ER) and progesterone receptor (PR) status was obtained from medical records. For patients whose ER/PR status could not be obtained from medical records, double immunohistochemical stains for PR and ER-alpha (ERα)/ER-beta (ERβ) were conducted at the Vanderbilt Molecular Epidemiology Laboratory. Overall, information on ER/PR status was ascertained for 2,676 (77.7%) of the 3,443 breast cancer cases. ER and PR status was categorized as positive or negative.

Statistical analysis

Intakes of food groups (fruits, vegetables, and animal foods) were subdivided into five categories according to the quintile distributions of dietary intakes among controls. Milk, eggs, and fresh-water fish were grouped into 3 categories based on the distribution among controls. All variables were categorized separately for phase I and II study participants.

Unconditional logistic regression analysis was used to derive odds ratios (ORs) and 95% confidence intervals (95% CI) for associations of dietary groups with breast cancer for all types of breast cancer combined and by ER/PR status. Total energy intake was adjusted for in the multivariate analysis using the standard method in which total energy intake is included in a multivariate risk model along with the nutrient of interest (21). Covariates adjusted for included age (continuous), education level (categories), family history of breast cancer in first-degree relatives (yes/no), history of breast fibroadenoma (yes/no), regular exercise (yes/no), study phase (I and II), BMI (continuous), age at menarche (categories), menopausal status (yes/no), and parity (categories). We also mutually adjusted for total meat, total fruit, and total vegetable intake in the analysis. Tests for trend were conducted by treating categorical variables as ordinal values in the models. We did not find that menopausal status modified dietary associations with breast cancer risk. Therefore, the analysis of associations between dietary intake and breast cancer risk was based on all breast cancer cases and controls. Additional analyses stratified by the time interval between diagnosis and interview were also conducted.

Effect modification was evaluated using the log likelihood ratio test to compare logistic models with and without the interaction term. To evaluate whether the observed risk estimated for subtype tumors was different across the four breast cancer subtypes (ER+/PR+, ER−/PR−, ER+/PR−, and ER−/PR+) or between ER+/PR+ and ER−/PR− tumors, multivariable polychotomous unconditional logistic regression was used. Wald χ² tests were employed to evaluate heterogeneity across cancer subtypes. All analyses were performed using SAS version 9.1 (SAS Institute, Cary, NC, USA). All statistical tests were based on two-sided probability, and a significance level of P < 0.05.

Results

Among 2,676 cases with both dietary information and hormone receptor status available, 1,409 (52.7%) were ER+/PR+, 712 (26.6%) were ER−/PR−, 301 (11.2%) were ER+/PR−, and 254 (9.5%) were ER−/PR+. ER or PR status was missing for 767 cases. Table 1 presents mean intakes for total energy and each food group for controls, cases, and by each subtype of breast cancer. Compared with controls, women with breast cancer had higher intake of red meat, poultry, fresh-water fish, and marine fish and lower intake of allium vegetables and fresh legumes. Dietary intakes did not differ across the four breast cancer subtype groups with ER/PR information. Compared with cases with ER/PR data, those with no ER/PR data had lower intake of milk, total vegetables, allium vegetables, fresh legumes, and melon and higher intake of rice.

Table 1.

Intake of selected food groups among controls and cases by ER and PR status, the Shanghai Breast Cancer Study I and II

	Controls	Cases	ER+/PR+	ER−/PR−	ER+/PR−	ER−/PR+	ER/PR unknown	P1	P2
	(n=3,474)	(n=3,443)	(n=1,409)	(n=712)	(n=301)	(n=254)	(n=767)
Total energy [mean ± SD (kcal/d)]	1754.0±437.3	1762.9±435.6	1762.2±427.7	1753.3±451.8	1756.9±385.1	1770.7±433.9	1773.0±454.2	0.39	0.94
Staple foods [mean ± SD (g/d)] 3	319.3±90.8	317.3±88.3	314.4±86.5	315.0±89.6	313.7±78.3	317.4±86.4	326.2±94.2	0.37	0.96
Total meat [mean ± SD (g/d)]	126.9±72.1	143.1±79.3	145.8±82.6	140.2±77.7	146.7±72.3	143.7±73.4	139.1±78.8	<0.0001	0.44
Red meat 4	56.5±39.5	61.4±43.4	61.7±44.6	61.5±45.0	61.0±36.7	60.9±40.8	61.3±43.3	<0.0001	0.99
Poultry 5	15.8±17.3	18.5±20.9	18.6±20.9	18.6±19.9	20.0±21.5	18.5±20.7	17.8±21.4	<0.0001	0.76
Fresh-water fish	16.8±20.6	19.8±23.9	20.3±23.9	19.0±22.9	21.5±23.4	19.2±24.1	19.0±24.7	<0.0001	0.40
Marine fish	22.7±27.9	26.2±31.7	27.2±32.8	25.1±28.2	27.9±31.7	27.3±32.3	24.3±32.4	<0.0001	0.45
Eggs [mean ± SD (g/d)]	28.9±21.5	28.7±23.7	28.6±26.7	28.9±22.2	28.7±21.1	27.6±20.7	29.0±31.1	0.71	0.87
Milk [mean ± SD (g/d)]	92.2±98.9	88.9±97.7	92.5±98.9	92.2±99.5	91.1±92.6	89.5±94.6	78.3±96.3	0.17	0.97
Total vegetables [mean ± SD (g/d)]	308.5±181.0	304.0±173.8	307.0±176.8	303.3±165.4	320.8±163.6	323.5±183.2	286.3±175.3	0.29	0.25
Dark-green vegetables 6	92.2±65.2	90.1±62.4	89.0±58.9	89.4±61.4	94.2±62.5	92.8±64.8	90.5±68.3	0.18	0.48
Carrots and tomatoes	38.3±47.1	39.2±48.7	40.4±52.7	38.4±44.4	38.0±37.7	42.7±48.3	37.1±49.0	0.47	0.54
Allium vegetables 7	10.6±13.5	9.8±11.1	10.1±11.7	10.1±10.3	10.6±11.4	9.5±10.0	8.8±11.0	0.007	0.73
Cruciferous vegetables 8	93.8±65.6	93.6±66.7	92.0±63.1	92.1±62.1	97.2±66.8	101.2±71.7	94.0±75.0	0.91	0.12
Fresh legumes 9	40.3±41.6	38.4±36.3	40.6±37.5	40.3±37.0	41.9±36.6	39.7±35.0	31.0±32.8	0.04	0.90
Melons 10	47.5±46.7	48.7±46.0	49.7±47.3	47.9±43.3	52.7±42.9	53.3±51.1	44.4±45.1	0.31	0.27
Other vegetables 11	39.5±34.2	38.0±32.4	37.8±31.7	38.5±34.0	40.5±29.1	39.6±31.5	36.2±33.8	0.06	0.54
Total fruit [mean ± SD (g/d)] 12	220.6±159.1	226.1±161.6	229.9±163.0	222.1±150.9	222.6±182.7	242.1±171.2	218.9±156.4	0.15	0.34
Citrus fruit 13	19.7±25.1	19.5±27.7	19.8±27.0	18.4±26.0	21.5±35.3	22.2±32.9	18.1±25.1	0.67	0.19
Rosaceae fruit 14	56.1±55.7	54.2±55.4	55.6±57.6	53.9±53.1	52.0±60.2	58.7±55.2	51.5±51.5	0.17	0.50

¹From the T-test (continuous variables), P for cases versus controls.

²From the ANOVA test (continuous variables), P for comparison across the four known receptor statuses.

³Staple foods: rice, noodles, steamed bread, and other wheat foods.

⁴Red meat: pork, pork chops, pig’s feet, beef, lamb, liver, organ meat (except liver);

⁵Poultry: chicken, duck, or goose;

⁶Dark-green vegetables: bok choy, spinach, fresh green peppers, garlic shoots, chives, scallions, Chinese celery;

⁷Allium vegetables: garlic, head of garlic, onions, chives, scallions, garlic shoots;

⁸Cruciferous vegetables: bok choy, cabbage, napa cabbage, cauliflower, Chinese white turnip;

⁹Fresh legumes: fresh soybeans, fresh broad beans, yard long beans, green beans, hyacinth bean/snow peas;

¹⁰Melons: winter melon, cucumber, wax gourd;

¹¹Other vegetables: potato, eggplant, corn, ginger, lotus root, wild rice stems, asparagus lettuce, bamboo shoots;

¹²Total fruit: citrus fruit, rosaceae fruit, watermelon, grapes, bananas, other fruits (such as strawberries and cantaloupe);

¹³Citrus fruit: tangerines, oranges, grapefruit;

¹⁴Rosaceae fruit: apples, pears, peaches.

As shown in Table 2, total vegetable intake level was significantly and inversely associated with breast cancer risk. The adjusted OR and 95% CI for the highest quintile of intake of total vegetables compared with the lowest quintile of intake was 0.80 (0.67–0.95), with a P for trend =0.02. Food group analysis showed that intakes of allium vegetables, fresh legumes, and other vegetables (such as potato, eggplant, lotus root, and ginger) were inversely associated with breast cancer risk (P for trend= 0.01, 0.0008, and 0.002, respectively). Compared with women in the lowest quintiles of intake of allium vegetables and fresh legumes, women in the highest quintiles of intake had a 17% and 26% reduction in breast cancer risk, respectively (OR=0.83, 95%CI=0.71–0.98 for allium vegetables and OR=0.74, 95%CI=0.63–0.87 for fresh legumes). No significant associations were found for intakes of dark-green vegetables, carrots and tomatoes, cruciferous vegetables, or melons with breast cancer risk. Higher intakes of citrus fruit and rosaceae fruit were inversely associated with breast cancer risk, and ORs were 0.77 (P for trend=0.003) and 0.84 (P for trend=0.004), respectively, for the highest quintiles versus the lowest. Total fruit intake, on the other hand, did not show a clear association with breast cancer risk. We found that, in general, the above dietary associations did not differ across the four breast cancer subtypes, with the possible exceptions of total vegetable intake (P for heterogeneity =0.05) and melon intake (P for heterogeneity =0.02). There was no significant heterogeneity between ER+/PR+ and ER−/PR− breast tumors (all P for heterogeneity >0.05).

Table 2.

Association of overall risk for breast cancer and subtypes defined by joint ER/PR status with intake levels of vegetables and fruits, SBCS I and II

Food group	Controls	All cases		ER+/PR+		ER−/PR−		ER+/PR−		ER−/PR+		P*	P**
		n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)
Total vegetables***
Q1 (< 166.92 g/d )	693	711	1.0(reference)	290	1.0(reference)	142	1.0(reference)	50	1.0(reference)	41	1.0(reference)
Q2 ( < 233.17 g/d )	692	657	0.88(0.75–1.02)	269	0.86(0.70–1.05)	147	0.99(0.76–1.28)	55	1.03(0.69–1.54)	44	1.04(0.66–1.62)
Q3 (< 309.49 g/d )	693	765	1.00(0.86–1.17)	323	1.00(0.82–1.22)	160	1.04(0.81–1.35)	73	1.32(0.89–1.95)	61	1.45(0.95–2.22)
Q4 (< 419.46 g/d )	694	657	0.85(0.73–1.00)	275	0.84(0.68–1.03)	135	0.89(0.68–1.17)	62	1.15(0.76–1.73)	52	1.22(0.78–1.91)
Q5 (≥ 419.46 g/d)	692	633	0.80(0.67–0.95)	247	0.71(0.57–0.90)	124	0.81(0.60–1.09)	60	1.13(0.73–1.75)	54	1.24(0.77–1.98)
P trend			0.02		0.01		0.13		0.46		0.26	0.05	0.68
Dark-green vegetables ***
Q1 (< 40.24 g/d )	693	734	1.0(reference)	304	1.0(reference)	138	1.0(reference)	59	1.0(reference)	54	1.0(reference)
Q2 ( < 65.43 g/d )	694	681	0.93(0.80–1.09)	267	0.86(0.71–1.06)	162	1.17(0.91–1.51)	52	0.85(0.57–1.26)	39	0.72(0.47–1.11)
Q3 (< 93.54 g/d )	690	694	0.93(0.79–1.09)	285	0.92(0.75–1.12)	144	1.03(0.79–1.34)	63	1.06(0.73–1.55)	56	1.02(0.69–1.51)
Q4 (< 134.37 g/d )	694	642	0.86(0.73–1.00)	280	0.90(0.74–1.10)	123	0.88(0.67–1.16)	63	1.02(0.70–1.48)	53	0.97(0.65–1.45)
Q5 (≥ 134.37 g/d)	693	672	0.93(0.79–1.09)	268	0.87(0.71–1.07)	141	1.04(0.80–1.36)	63	1.11(0.75–1.62)	50	0.92(0.61–1.39)
P trend			0.20		0.31		0.50		0.42		0.82	0.57	0.92
Carrots and tomatoes ***
Q1 (< 8.37 g/d )	694	678	1.0(reference)	277	1.0(reference)	123	1.0(reference)	48	1.0(reference)	46	1.0(reference)
Q2 ( < 18.19 g/d )	681	687	1.01(0.86–1.18)	272	0.97(0.79–1.19)	168	1.34(1.04–1.74)	54	1.05(0.70–1.58)	51	1.11(0.73–1.68)
Q3 (< 30.71 g/d )	695	683	0.95(0.82–1.11)	282	0.96(0.78–1.17)	147	1.14(0.88–1.50)	78	1.47(1.00–2.15)	46	0.95(0.62–1.47)
Q4 (< 57.05 g/d )	702	674	0.95(0.81–1.11)	268	0.91(0.74–1.12)	136	1.07(0.81–1.40)	67	1.24(0.83–1.85)	52	1.10(0.72–1.68)
Q5 (≥ 57.05 g/d)	692	701	1.00(0.85–1.17)	305	1.06(0.86–1.31)	134	1.08(0.81–1.43)	53	1.06(0.69–1.63)	57	1.18(0.77–1.81)
P trend			0.71		0.78		0.76		0.53		0.46	0.81	0.63
Allium vegetables ***
Q1 (< 3.00 g/d )	660	656	1.0(reference)	267	1.0(reference)	125	1.0(reference)	51	1.0(reference)	39	1.0(reference)
Q2 ( < 5.32 g/d )	520	553	1.03(0.87–1.21)	255	1.13(0.91–1.40)	118	1.10(0.83–1.45)	53	1.14(0.76–1.72)	32	1.03(0.63–1.67)
Q3 (< 8.49 g/d )	894	961	1.07(0.92–1.24)	373	1.04(0.86–1.27)	195	1.18(0.92–1.51)	86	1.22(0.84–1.76)	92	1.82(1.23–2.70)
Q4 (< 14.74 g/d )	682	638	0.92(0.78–1.08)	254	0.89(0.72–1.10)	145	1.10(0.84–1.44)	55	0.96(0.64–1.43)	50	1.24(0.80–1.92)
Q5 (≥ 14.74 g/d )	708	615	0.83(0.71–0.98)	255	0.83(0.67–1.03)	125	0.91(0.69–1.21)	55	0.94(0.63–1.42)	39	0.89(0.55–1.42)
P trend			0.01		0.02		0.60		0.55		0.86	0.57	0.23
Crucifer vegetables ***
Q1 (< 38.97g/d )	693	701	1.0(reference)	285	1.0(reference)	141	1.0(reference)	60	1.0(reference)	44	1.0(reference)
Q2 ( < 65.35 g/d )	692	673	0.98(0.84–1.14)	281	0.99(0.81–1.21)	143	1.02(0.79–1.32)	46	0.76(0.51–1.14)	50	1.18(0.77–1.80)
Q3 (< 95.05g/d )	692	680	0.97(0.83–1.13)	286	0.99(0.81–1.22)	139	0.98(0.76–1.27)	66	1.09(0.75–1.59)	44	1.02(0.66–1.57)
Q4 (< 137.88 g/d )	693	794	0.99(0.85–1.16)	276	0.97(0.79–1.19)	155	1.10(0.85–1.42)	59	0.96(0.66–1.41)	53	1.24(0.81–1.89)
Q5 (≥ 137.88 g/d)	694	675	0.97(0.83–1.14)	276	0.96(0.78–1.18)	130	0.94(0.72–1.23)	69	1.17(0.80–1.70)	61	1.44(0.95–2.18)
P trend			0.81		0.63		0.90		0.23		0.10	0.20	0.82
Fresh legumes ***
Q1 (< 13.84 g/d )	689	752	1.0(reference)	316	1.0(reference)	142	1.0(reference)	58	1.0(reference)	49	1.0(reference)
Q2 ( < 23.61 g/d )	694	691	0.89(0.77–1.04)	269	0.83(0.68–1.01)	165	1.14(0.89–1.47)	57	0.94(0.64–1.38)	52	1.07(0.71–1.61)
Q3 (< 35.66 g/d )	690	647	0.84(0.72–0.98)	249	0.77(0.63–0.94)	136	0.93(0.72–1.21)	67	1.10(0.75–1.59)	51	1.07(0.71–1.63)
Q4 (< 58.27 g/d )	696	727	0.89(0.76–1.04)	323	0.92(0.76–1.12)	129	0.85(0.65–1.11)	63	0.96(0.65–1.41)	59	1.16(0.77–1.75)
Q5 (≥ 58.27 g/d )	695	606	0.74(0.63–0.87)	247	0.70(0.57–0.87)	136	0.91(0.70–1.20)	55	0.86(0.57–1.27)	42	0.81(0.52–1.27)
P trend			0.0008		0.01		0.11		0.58		0.53	0.85	0.75
Melons ***
Q1 (< 15.33 g/d )	692	653	1.0(reference)	263	1.0(reference)	148	1.0(reference)	30	1.0(reference)	44	1.0(reference)
Q2 ( < 27.25 g/d )	694	653	0.95(0.82–1.11)	279	1.01(0.82–1.24)	131	0.85(0.65–1.10)	57	1.80(1.14–2.85)	47	1.03(0.67–1.59)
Q3 (< 43.34 g/d )	694	725	1.05(0.90–1.22)	284	1.00(0.81–1.22)	151	0.97(0.75–1.25)	75	2.34(1.50–3.65)	53	1.13(0.74–1.72)
Q4 (< 68.48 g/d )	693	664	0.97(0.83–1.14)	277	1.00(0.81–1.23)	129	0.86(0.67–1.12)	73	2.41(1.53–3.77)	41	0.90(0.58–1.42)
Q5 (≥ 68.48 g/d )	691	728	1.04(0.88–1.22)	301	1.03(0.83–1.27)	149	0.97(0.74–1.27)	65	2.10(1.32–3.34)	67	1.40(0.92–2.12)
P trend			0.64		0.88		0.84		0.002		0.21	0.02	0.77
Other vegetables ***
Q1 (< 14.36 g/d )	693	717	1.0(reference)	308	1.0(reference)	138	1.0(reference)	47	1.0(reference)	39	1.0(reference)
Q2 ( < 24.20 g/d )	693	694	0.92(0.79–1.07)	275	0.83(0.68–1.01)	154	1.05(0.81–1.36)	63	1.23(0.83–1.84)	53	1.33(0.87–2.06)
Q3 (< 36.89 g/d )	690	723	0.94(0.80–1.09)	293	0.88(0.72–1.07)	155	1.05(0.81–1.36)	66	1.28(0.86–1.91)	53	1.31(0.84–2.02)
Q4 (< 58.82 g/d )	693	648	0.83(0.71–0.97)	272	0.78(0.64–0.96)	137	0.91(0.70–1.20)	59	1.13(0.75–1.71)	60	1.46(0.94–2.25)
Q5 (≥ 58.82 g/d )	695	641	0.78(0.66–0.92)	256	0.69(0.55–0.85)	124	0.80(0.60–1.07)	65	1.22(0.80–1.85)	47	1.05(0.66–1.68)
P trend			0.002		0.001		0.09		0.52		0.74	0.06	0.47
Total fruits ****
Q1 (< 95.99 g/d )	693	701	1.0(reference)	283	1.0(reference)	132	1.0(reference)	64	1.0(reference)	47	1.0(reference)
Q2 ( < 161.83 g/d )	693	660	0.89(0.76–1.04)	261	0.87(0.70–1.06)	162	1.16(0.90–1.50)	56	0.77(0.53–1.13)	51	1.00(0.66–1.52)
Q3 (< 228.30 g/d )	692	654	0.84(0.72–0.98)	259	0.80(0.65–0.99)	133	0.92(0.70–1.20)	57	0.74(0.51–1.09)	48	0.86(0.56–1.32)
Q4 (< 317.69 g/d )	693	643	0.81(0.69–0.95)	267	0.83(0.67–1.02)	134	0.92(0.70–1.21)	66	0.83(0.57–1.22)	38	0.65(0.41–1.02)
Q5 (≥ 317.69 g/d)	693	765	1.00(0.85–1.17)	334	1.09(0.88–1.34)	147	1.05(0.80–1.39)	57	0.75(0.50–1.13)	68	1.13(0.74–1.72)
P trend			0.69		0.52		0.69		0.33		0.82	0.59	0.42
Citrus fruits ****
Q1 (< 2.33 g/d )	678	701	1.0(reference)	277	1.0(reference)	146	1.0(reference)	48	1.0(reference)	51	1.0(reference)
Q2 ( < 7.66 g/d )	683	673	0.87(0.75–1.02)	271	0.86(0.70–1.06)	146	0.92(0.71–1.19)	66	1.22(0.82–1.80)	50	0.88(0.58–1.32)
Q3 (< 16.39 g/d )	700	720	0.91(0.78–1.06)	282	0.86(0.70–1.06)	159	0.95(0.74–1.23)	66	1.15(0.77–1.70)	47	0.78(0.51–1.19)
Q4 (< 30.40 g/d )	686	675	0.86(0.74–1.01)	302	0.94(0.77–1.16)	132	0.80(0.62–1.04)	51	0.88(0.58–1.34)	53	0.89(0.59–1.33)
Q5 (≥ 30.40 g/d )	717	654	0.77(0.66–0.90)	272	0.78(0.63–0.96)	125	0.71(0.54–0.93)	69	1.09(0.73–1.62)	51	0.74(0.49–1.13)
P trend			0.003		0.09		0.01		0.77		0.21	0.54	0.26
Rosaceae fruits ****
Q1 (<12.51 g/d )	690	712	1.0(reference)	290	1.0(reference)	149	1.0(reference)	64	1.0(reference)	43	1.0(reference)
Q2 ( < 30.60 g/d )	696	736	0.96(0.83–1.12)	283	0.90(0.74–1.11)	158	1.01(0.78–1.30)	63	0.87(0.60–1.27)	55	1.18(0.78–1.80)
Q3 (< 54.88 g/d )	692	700	0.89(0.76–1.04)	291	0.90(0.73–1.10)	149	0.91(0.70–1.18)	70	0.90(0.62–1.30)	50	1.02(0.66–1.57)
Q4 (< 91.13 g/d )	693	627	0.80(0.68–0.94)	269	0.85(0.69–1.05)	110	0.68(0.52–0.90)	55	0.73(0.49–1.08)	53	1.07(0.69–1.64)
Q5 (≥ 91.13 g/d)	693	648	0.84(0.71–0.98)	271	0.85(0.69–1.06)	142	0.89(0.68–1.17)	48	0.61(0.40–0.93)	51	0.99(0.63–1.55)
P trend			0.004		0.14		0.06		0.02		0.74	0.43	0.52

Note: Missing values (<0.5%) were excluded from models. The cut-point value for each food group is the mean of phase I and phase II. Adjusted for total energy, age, education level, ever diagnosed with benign breast disease, first-degree family history of breast cancer, participation in regular exercise, BMI, study phase (I and II), age at menarche, menopausal status, parity, and total meat intake.

^*Test for heterogeneity of trend; P across four subtypes, ER+/PR+, ER−/PR−, ER+/PR−, and ER−/PR+; calculated using multivariable polychotomous logistic regression.

^**Test for heterogeneity of trend; P between ER+/PR+ and ER−/PR−;

^***Further adjustment for total fruit intake;

^****Further adjustment for total vegetable intake.

Statistically significant positive associations were observed for total meat, red meat, poultry, fresh-water fish, and marine fish intake (Table 3). The adjusted ORs for the highest versus lowest quintiles or tertiles of intake with overall breast cancer risk were 2.18 (95%CI: 1.82–2.61) for total meat, 1.45 (95% CI: 1.22–1.72) for red meat, 1.31 (95% CI: 1.12–1.53) for poultry, 1.39 (95% CI: 1.23–1.56) for fresh-water fish, and 1.19 (95% CI: 1.02–1.39) for marine fish. Egg and milk consumption were inversely associated with breast cancer risk, with 13% and 17% reductions in risk for the highest tertiles of intake compared with the lowest tertiles of intake. We found no statistically significant heterogeneity in the risk estimates for these food groups across the four breast cancer subtypes or between ER+/PR+ and ER−/PR− tumors (all heterogeneity tests: P ≥0.15). There were suggestions that meat intake may be associated with higher risk of the ER+/PR− subtype compared with the other three breast cancer subtypes.

Table 3.

Association of overall risk for breast cancer and subtypes defined by joint ER/PR status with intake levels of animal foods, SBCS I and II

Food group	Controls	All cases		ER+/PR+		ER−/PR−		ER+/PR−		ER−/PR+		P*	P**
		n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)	n	OR (95%CI)
Total meat
Q1 (< 70.04 g/d)	693	471	1.0(reference)	175	1.0(reference)	98	1.0(reference)	25	1.0(reference)	39	1.0(reference)
Q2 ( < 98.18 g/d)	692	592	1.28(1.09–1.52)	245	1.42(1.13–1.78)	138	1.40(1.05–1.86)	52	2.17(1.32–3.56)	34	0.84(0.52–1.35)
Q3 (< 128.83 g/d)	693	702	1.55(1.31–1.83)	289	1.70(1.35–2.13)	134	1.38(1.03–1.85)	68	2.94(1.81–4.79)	55	1.33(0.85–2.07)
Q4 (< 174.82g/d)	695	741	1.64(1.38–1.94)	307	1.79(1.42–2.26)	165	1.71(1.27–2.29)	75	3.23(1.97–5.29)	51	1.21(0.76–1.91)
Q5 (≥ 174.82g/d)	691	917	2.18(1.82–2.61)	388	2.41(1.89–3.08)	173	1.93(1.41–2.65)	80	3.98(2.37–6.67)	73	1.76(1.10–2.82)
P trend			<0.0001		<0.0001		<0.0001		<0.0001		0.005	0.15	0.26
Red meat
Q1 (< 26.34 g/d)	693	564	1.0(reference)	211	1.0(reference)	117	1.0(reference)	40	1.0(reference)	43	1.0(reference)
Q2 ( < 40.51 g/d)	694	600	1.07(0.91–1.25)	262	1.24(1.00–1.53)	113	0.95(0.72–1.27)	56	1.39(0.91–2.13)	45	1.03(0.67–1.60)
Q3 (< 57.56 g/d)	691	741	1.30(1.11–1.52)	298	1.36(1.10–1.69)	164	1.36(1.04–1.78)	60	1.49(0.97–2.27)	54	1.19(0.78–1.83)
Q4 (< 82.11 g/d)	692	713	1.25(1.07–1.47)	310	1.43(1.15–1.77)	140	1.19(0.90–1.57)	76	1.91(1.27–2.89)	51	1.12(0.72–1.73)
Q5 (≥ 82.11 g/d)	694	805	1.45(1.22–1.72)	323	1.51(1.20–1.90)	174	1.55(1.16–2.07)	68	1.81(1.15–2.84)	59	1.29(0.81–2.03)
P trend			<0.0001		0.0003		0.001		0.002		0.28	0.57	0.71
Poultry
Q1 (< 3.71 g/d)	675	582	1.0(reference)	220	1.0(reference)	118	1.0(reference)	49	1.0(reference)	49	1.0(reference)
Q2 ( < 7.48g/d)	700	638	1.04(0.89–1.22)	255	1.09(0.88–1.35)	136	1.07(0.81–1.40)	53	1.01(0.67–1.52)	46	0.88(0.58–1.34)
Q3 (< 13.25 g/d)	694	607	0.98(0.84–1.16)	278	1.19(0.96–1.48)	110	0.87(0.65–1.15)	53	1.01(0.67–1.52)	37	0.70(0.45–1.09)
Q4 (< 24.33 g/d)	677	743	1.21(1.03–1.41)	307	1.27(1.03–1.57)	158	1.25(0.96–1.63)	64	1.20(0.80–1.78)	57	1.04(0.69–1.56)
Q5 (≥ 24.33 g/d)	718	853	1.31(1.12–1.53)	344	1.35(1.09–1.67)	186	1.40(1.07–1.83)	81	1.46(0.99–2.15)	63	1.04(0.70–1.57)
P trend			0.0002		0.003		0.003		0.03		0.56	0.64	0.57
Fresh-water fish
T1 (< 6.64 g/d)	1370	1159	1.0(reference)	477	1.0(reference)	244	1.0(reference)	92	1.0(reference)	94	1.0(reference)
T2 (< 21.07 g/d)	914	875	1.17(1.03–1.32)	345	1.14(0.97–1.35)	179	1.16(0.93–1.43)	71	1.19(0.86–1.64)	65	1.08(0.78–1.51)
T3 (≥ 21.07 g/d)	1180	1389	1.39(1.23–1.56)	582	1.39(1.20–1.62)	285	1.36 (1.12–1.66)	137	1.60(1.20–2.14)	93	1.10(0.81–1.50)
P trend			<0.0001		<0.0001		0.002		0.001		0.52	0.32	0.84
Marine fish
Q1 (< 4.22g/d)	660	616	1.0(reference)	243	1.0(reference)	121	1.0(reference)	38	1.0(reference)	50	1.0(reference)
Q2 ( < 9.04 g/d)	682	606	0.93(0.79–1.09)	256	0.98(0.80–1.21)	125	0.96(0.73–1.27)	58	1.49(0.97–2.28)	31	0.56(0.35–0.89)
Q3 (< 20.97 g/d)	597	583	1.02(0.86–1.20)	239	1.02(0.82–1.26)	122	1.05(0.79–1.39)	59	1.71(1.12–2.63)	52	1.03(0.68–1.56)
Q4 (< 36.70 g/d)	696	655	0.97(0.82–1.13)	261	0.94(0.76–1.16)	153	1.11(0.85–1.45)	56	1.35(0.87–2.08)	48	0.79(0.52–1.21)
Q5 (≥ 36.70 g/d)	829	963	1.19(1.02–1.39)	405	1.25(1.02–1.52)	187	1.21(0.93–1.56)	89	1.92(1.28–2.88)	71	0.96(0.65–1.42)
P trend			0.01		0.05		0.07		0.009		0.60	0.44	0.79
Eggs ***
T1 (<16.55 g/d)	1380	1417	1.0(reference)	575	1.0(reference)	300	1.0(reference)	125	1.0(reference)	103	1.0(reference)
T2 (<43.70 g/d)	819	807	0.95(0.84–1.08)	150	1.01(0.85–1.20)	150	0.89(0.71–1.12)	66	0.95(0.69–1.32)	69	1.14(0.82–1.59)
T3 (≥ 43.70 g/d)	1265	1199	0.87(0.78–0.98)	258	0.91(0.78–1.06)	258	0.92(0.76–1.12)	109	0.90(0.68–1.19)	80	0.77(0.56–1.06)
P trend			0.02		0.23		0.40		0.49		0.11	0.78	0.93
Milk ***
T1 (<7.39 g/d)	1310	1371	1.0(reference)	546	1.0(reference)	274	1.0(reference)	113	1.0(reference)	103	1.0(reference)
T2 (<135.72 g/d)	811	784	0.85(0.74–0.97)	319	0.83(0.70–0.99)	164	0.86(0.69–1.07)	77	0.99(0.72–1.35)	53	0.71 (0.50–1.01)
T3 (≥ 135.72 g/d)	1343	1268	0.83(0.73–0.93)	539	0.85(0.73–0.99)	270	0.85(0.70–1.04)	110	0.80(0.60–1.08)	96	0.80(0.59–1.10)
P trend			0.001		0.04		0.13		0.14		0.18	0.97	0.91

Note: Missing values (<0.5%) were excluded from models. The cut-point value for each food group is the mean of phase I and phase II. Adjusted for total energy, age, education level, ever diagnosed with benign breast disease, first-degree family history of breast cancer, participation in regular exercise, BMI, study phase (I and II), age at menarche, menopausal status, parity, total vegetable intake, and total fruit intake.

^*Test for heterogeneity of trend; P across four subtypes, ER+/PR+, ER−/PR−, ER+/PR−, and ER−/PR+; calculated using multivariable polychotomous logistic regression.

^**Test for heterogeneity of trend; P between ER+/PR+ and ER−/PR−.

^***Further adjustment for total meat intake.

When stratifying by menopausal status, exercise (yes/no), BMI (categorized by the median), and study phase, we found no statistically significant heterogeneity in the risk estimates for food groups across the four breast cancer subtypes or between ER+/PR+ and ER−/PR− tumors for all stratified analyses (all P for heterogeneity >0.05; data not shown).

In this study, the median interval from diagnosis to interview was 126 days, and over 85% of cases were interviewed within 6 months after cancer diagnosis. When cases were stratified by the mean time interval (3 months) between diagnosis and interview and compared with all controls, the ORs for overall breast cancer risk in relation to dietary groups did not vary substantially. Moreover, the heterogeneity test between ER+/PR+ and ER−/PR− tumors remained non-significant (data not shown).

Discussion

This large case-control study showed that fruit, vegetable, milk, and egg consumption were each inversely associated with risk of breast cancer, while meat consumption was positively related to risk. In general, the associations of dietary factors with breast cancer did not vary by estrogen or progesterone receptor status or by menopausal status.

Fruits and vegetables are common sources of many nutritional compounds, such as dietary fiber, vitamins, minerals, and other bioactive compounds (phytochemicals), that may reduce the occurrence of breast cancer (5). However, the bioavailability of these compounds in fruits and vegetables is variable and their ultimate health effects are uncertain (1). In our study, we found that high intake of vegetables was associated with a decreased risk of breast cancer. Other studies, mostly conducted in Asian populations, have found the same protective effect (9–10, 22–25). A meta-analysis of 23 studies also concluded that vegetable intake reduced breast cancer risk by 20% to 25% (26). However, a pooled analysis of eight prospective studies conducted largely in Western populations did not find any protective effect for fruit or vegetable consumption (7).

Several factors have been put forth as possibly contributing to the variability in results across studies. These include: the types and levels of fruit and vegetable intake in different populations, the lack of variation in fruit and vegetable intake in Western populations, the detail and quality of the FFQs used in different studies, sample sizes, study design, and adjustment for potentially confounding variables (5, 7, 22). Analysis of fruit and vegetable groups has suggested that specific groups, such as dark-green leafy vegetables, carrots and tomatoes, cruciferous vegetables, and citrus fruits, may confer protection against breast cancer (9–10, 22, 25, 27). In our study, consumption of allium vegetables, fresh legumes, citrus fruits, and rosaceae fruits was inversely related to breast cancer risk. Fresh legumes, including fresh soybeans, fresh broad beans, yard long beans, and green beans, contain high levels of isoflavones, folate, B vitamins, and deguelin (1, 28–29), which have been previously related to reduced risk of breast cancer (30–32). Allium vegetables, such as onions, garlic, and scallions, are high in organosulfur compounds that have been shown to inhibit mutagenesis and DNA adduct formation and to have other anti-cancer effects (33). These two types of vegetables are consumed in high quantities in the Chinese population.

Residual or uncontrolled confounding may cover up the true association between consumption of food groups and breast cancer risk. High correlations between certain food groups can make it difficult to firmly establish the independent effects of any particular food group. Thus, the inconsistency or insufficiency of adjustment for potential confounding factors in previous published reports may also partly contribute to the controversial role of diet in breast cancer risk.

In our study, total meat intake was positively associated with breast cancer overall and with each of the four subtypes defined by hormone receptor status. However, previously published studies on the association between breast cancer and meat intake have been inconsistent and suggested a null association (17, 34–36). The positive association we found for red meat is consistent with the findings of some previous epidemiologic studies (14, 35–36), but not all (37–38). A recent meta-analysis (39) of the association between breast cancer risk and red meat consumption in premenopausal women found a summary relative risk of 1.24 (95% CI: 1.08–1.42). As for high consumption of poultry, the evidence is limited. Two previous studies observed a non-significant increase in risk for breast cancer (35, 40) associated with poultry consumption. In the latter study, breast cancer risk increased when chicken was consumed with the skin, suggesting that fat rather than muscle meat or some component produced during cooking may be the cause.

It was recently suggested by a review article (41) that the increase in risk associated with meat intake may not be a function of meat per se, but may reflect high intake of fat and/or be attributable to the carcinogens generated through various cooking and processing methods. Meat cooked at a high temperature may contain mammary-specific carcinogens, such as heterocyclic amines (HCAs) and polycyclic aromatic hydrocarbons (PAHs)(42–43). In our study population, the percentage of participants who used deep frying as a cooking method was 16.2 %. Meat is also a source of heme iron, a highly bio-available form of iron, which has been shown to enhance estrogen-induced tumor formation (44). Additionally, exogenous hormones administered to animals or hormones used during meat processing could be passed on to consumers of meat (45)(46).

We observed an increase in risk associated with high consumption of fresh-water fish and marine fish, which does not concur with other studies conducted in different populations where null (36–37, 47) or inverse (48–49) associations were found. Our findings are, however, consistent with several studies conducted in our study population, in which fish consumption has been related to an elevated risk of endometrial cancer (50), breast cancer (51), and colon cancer (52). Chemical pollution may partly explain the findings for fresh-water fish, since fresh-water fish raised in industrial areas may have high levels of methylmercury, dibenzofurans, organochlorine residues, and other chemicals, some of which are highly toxic and potent carcinogens (51, 53). Studies incorporating biomarkers of chemical exposure are needed.

Our finding of an inverse association with egg consumption is consistent with another case-control study carried out in China (23). A pooled analysis of European and North American cohort studies found that compared with women who did not eat eggs, breast cancer risk was slightly decreased among women who consumed <2 eggs per week, but slightly increased among women who consumed at least 7 eggs per week (≥1 egg per day) (36). The mean of egg intake in our population is about 3 eggs per week. In our study, milk consumption was associated with a reduced risk of breast cancer. However, a previous pooled analysis of cohort studies (36) found no relation between breast cancer and milk intake. A systematic review (54) of more than 40 case-control studies and 12 cohort studies, however, suggested that the rumenic, vaccenic, butyric, and branched-chain fatty acids, calcium, and vitamin D in milk might have the potential to protect against breast cancer. Compared with Western populations, consumption of milk and dairy products is low in our study population. Although we carefully adjusted for a wide range of variables that are related to dietary intake, we also cannot exclude confounding or selection bias as a likely explanation.

Few data are available on dietary components and breast cancer subtypes by hormone receptor status, and published reports are inconsistent. Two case-control studies found that the protective effect of fruit and vegetable intake was stronger for ER+ than for ER− tumors (9, 11). However, two prospective studies showed that fruit and vegetable intake were inversely associated with ER− breast cancer (12–13). In the Swedish Mammography Cohort (16) and NIH-AARP Diet and Health Study (17), the association of red meat intake with breast cancer did not differ by ER/PR status. The Nurses’ Health Study II, in contrast, found that higher intake of red meat was associated with a significantly increased risk of ER+/PR+ breast cancer, but not ER−/PR− breast cancer among premenopausal women (14). However, this study only included 167 patients with ER−/PR− breast cancer and had low statistical power to assess the etiology of this subtype of breast cancer. Another prospective study involving 218 ER+/PR+ and 93 ER−/PR− breast cancer patients reported that the association between whole grain products and risk of breast cancer did not differ according to hormone receptor status (18). In our study, the associations between food groups and breast cancer differed neither by hormone receptor status in general nor according to whether analyses were stratified by menopausal status, exercise participation, BMI, or study phase.

It has been hypothesized that the risk factors most closely associated with ER+/PR+ breast tumors mainly involve hormonal mechanisms related to estrogen and progesterone exposure, whereas the etiology of ER−/PR− breast cancer may be linked to non-hormonal mechanisms. Diet may play a comprehensive role in the prevention or risk of developing breast cancer via both estrogen-mediated and other non-estrogen related mechanisms. Our finding that dietary intake was linked to all ER and PR subtypes supports the multiple mechanism hypothesis.

Our study is the largest single study that has examined associations of food groups and breast cancer subtypes defined by ER and PR status. Additional strengths of our study include the population-based design, high response rates, and detailed information available on a wide range of potential confounders and dietary factors collected using a validated FFQ.

In our study, in order to minimize the influence of dietary changes after cancer diagnosis, we tried to interview patients as soon as possible after diagnosis, resulting in approximately 85% of cases being interviewed within the first 6 months after they were diagnosed with cancer. Furthermore, when we stratified analyses by the time interval between diagnosis and interview, we found no material influence on the dietary associations with breast cancer or subtypes. Although misclassification of dietary intake due to the nature of self-reported information taken from an FFQ is unavoidable, we found that the study controls had patterns of food intake similar to participants of the Shanghai Women’s Health Study (55), a population-based cohort study of 75,000 women aged 40–70 years who were recruited from the same urban Chinese population during the same time period as our study population. About 22% of cases lacked information on joint ER/PR status. Although this is lower than other case-control (9, 11) and prospective studies focusing on diet and breast cancer (56–57), potential selection bias cannot be completely ruled out.

In conclusion, we found that dietary intake was associated with breast cancer risk among Chinese women. In general, a plant-based diet was inversely related to breast cancer risk and a meat-based diet was associated with increased risk. The dietary associations we found did not vary significantly by ER/PR status, suggesting that the association between diet and breast cancer may be mediated through other mechanisms in addition to those related to estrogen.