Hostname: page-component-7c8c6479df-xxrs7 Total loading time: 0 Render date: 2024-03-19T05:45:43.775Z Has data issue: false hasContentIssue false

Fruit and vegetable intake and breast cancer prognosis: a meta-analysis of prospective cohort studies

Published online by Cambridge University Press:  03 April 2017

Chen Peng
Affiliation:
Department of Medical Statistics and Epidemiology, School of Public Health, Sun Yat-Sen University, Guangzhou 510080, People’s Republic of China
Wei-Ping Luo
Affiliation:
Department of Medical Statistics and Epidemiology, School of Public Health, Sun Yat-Sen University, Guangzhou 510080, People’s Republic of China Department of Prevention and Health Care, Guangzhou First People’s Hospital, Guangzhou Medical University, Guangzhou 510180, People’s Republic of China
Cai-Xia Zhang*
Affiliation:
Department of Medical Statistics and Epidemiology, School of Public Health, Sun Yat-Sen University, Guangzhou 510080, People’s Republic of China Guangdong Provincial Key Laboratory of Food, Nutrition and Health, Guangzhou 510080, People’s Republic of China
*
*Corresponding author: Professor C.-X. Zhang, fax +86 20 8733 0446, email zhangcx3@mail.sysu.edu.cn
Rights & Permissions [Opens in a new window]

Abstract

The effect of fruit and vegetable intake on breast cancer prognosis is controversial. Thus, a meta-analysis was carried out to explore their associations. A comprehensive search was conducted in PubMed, Web of Science, OVID, ProQuest and Chinese databases from inception to April 2016. The summary hazard ratios (HR) and 95 % CI were estimated using a random effects model if substantial heterogeneity existed and using a fixed effects model if not. Subgroup analyses and sensitivity analyses were also performed. In total, twelve studies comprising 41 185 participants were included in the meta-analysis. Comparing the highest with the lowest, the summary HR for all-cause mortality were 1·01 (95 % CI 0·72, 1·42) for fruits and vegetables combined, 0·96 (95 % CI 0·83, 1·12) for total vegetable intake, 0·99 (95 % CI 0·89, 1·11) for cruciferous vegetable intake and 0·88 (95 % CI 0·74, 1·05) for fruit intake; those for breast cancer-specific mortality were 1·05 (95 % CI 0·77, 1·43) for total vegetable intake and 0·94 (95 % CI 0·69, 1·26) for fruit intake; and those for breast cancer recurrence were 0·89 (95 % CI 0·53, 1·50) for total vegetable intake and 0·98 (95 % CI 0·76, 1·26) for cruciferous vegetable intake. This meta-analysis found no significant associations between fruit and vegetable intake and breast cancer prognosis.

Type
Full Papers
Copyright
Copyright © The Authors 2017 

Breast cancer is the most prevalent cancer among women; it is increasing globally and is the most common cause of cancer-related death in women on all continents( Reference Yue 1 ). Breast cancer affects about 12 % of women worldwide( Reference McGuire, Brown and Malone 2 ). It was reported that 1·7 million new cases were diagnosed worldwide in 2012, accounting for about 25 % of total cancer cases in women( Reference Torre, Bray and Siegel 3 ). Moreover, breast cancer is the leading cause of cancer deaths in women. It has the highest mortality of any cancer in women (12·9 per 100 000), accounting for approximately 14·7 % of all cancer-related mortalities among women worldwide( Reference Forman and Ferlay 4 ). However, advances in early detection and in types of therapies and their application have resulted in prolonged survival among women diagnosed with breast cancer( Reference Brewster and Helzlsouer 5 ). Thus, it is very important to research the prognostic factors of breast cancer given the increasing number of breast cancer survivors.

Fruits and vegetables may offer potential protective effects against breast cancer occurrence and prognosis. Various biochemical compositions found in a diet high in fruits and vegetables kill breast cancer cells in vitro and prevent and suppress breast cancer progression in various preclinical animal models( Reference Reuben, Gopalan and Petit 6 ). Chatterjee et al. ( Reference Chatterjee, Roy and Janarthan 7 ) found that carotenoids, consisting largely of α-carotene, β-carotene and β-cryptoxanthin, are typical constituents of orange-, red- and yellow-coloured fruits and green vegetables. β-Ionone, an end-ring analogue of β-carotenoid, inhibits 7,12-dimethylbenz-(α)anthracene (DMBA)-induced mammary carcinogenesis by inhibiting cell proliferation and inducing apoptosis( Reference Liu, Sun and Dong 8 ). Liu et al. ( Reference Liu, Dong and Sun 9 ) also documented the chemopreventive effects of varied doses of dietary β-ionone on the development and growth of DMBA-induced rat mammary tumours and the biologically relevant plasma antioxidant status. Cruciferous vegetables are also major sources of glucosinolate-derived bioactive compounds such as isothiocyanates, which have been shown in animal and in vitro studies to inhibit cancer growth and progression( Reference Nechuta, Caan and Chen 10 ). Sulforaphane is an isothiocyanate that elicits both pro-apoptotic and anti-proliferative properties( Reference Fimognari, Lenzi and Hrelia 11 ). Azarenko( Reference Azarenko 12 ) found that sulforaphane stabilised microtubules in breast cancer cells by suppressing microtubule dynamic instability at concentrations that inhibited cell proliferation and induced mitotic arrest. It has also been reported that high consumption of brassicaceous vegetables is regularly associated with low cancer risk, with its extracts regulating the progression of cancer through anti-inflammatory effects, effects on signal transduction, epigenetic effects and modulation of the colonic microflora( Reference Ferguson and Schlothauer 13 ). At present, interest in a number of fruits high in polyphenols has been heightened because of their reported chemopreventive and chemotherapeutic potential( Reference Turrini, Ferruzzi and Fimognari 14 ). Many fruits such as pomegranate have been shown to exert anticancer activity, which is generally attributed to their high polyphenol content( Reference Seeram, Adams and Henning 15 ). Therefore, high consumption of fruits and vegetables might play an important role in breast cancer prognosis.

To clarify the effects of the consumption of fruits and vegetables on the prognosis of breast cancer survivors, several studies have investigated the association of fruit and vegetable intake with all-cause mortality or breast cancer-specific mortality or recurrence( Reference Nechuta, Caan and Chen 10 , Reference Pan 16 Reference Buck, Zaineddin and Vrieling 26 ). The majority of these studies have reported a non-significant inverse association of fruit and vegetable intake with all-cause mortality( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ), and two studies also found this same non-significant association with breast cancer recurrence( Reference Nechuta, Caan and Chen 10 , Reference Saxe, Rock and Wicha 25 ). Moreover, among survivors of early-stage breast cancer, adoption of a diet that was very high in vegetables, fruits and fibre and low in fat did not reduce additional breast cancer events or mortality during a 7·3-year follow-up period( Reference Pierce, Natarajan and Caan 27 ). However, two studies reported an inverse association of fruit and vegetable intake with all-cause mortality( Reference Pan 16 , Reference McEligot, Largent and Ziogas 21 ), and one study reported that vegetable intake is inversely associated with breast cancer recurrence( Reference Thomson, Rock and Thompson 18 ). Hauner et al. ( Reference Hauner, Janni and Rack 28 ) found that a low-fat diet rich in fruits, vegetables and fibre seems to be weakly associated with a better prognosis. To our knowledge, the findings on the effects of fruit and vegetable intake on breast cancer prognosis are not conclusive. Therefore, a meta-analysis of prospective cohorts is required to quantitatively evaluate the overall effect of fruit and vegetable intake on the prognosis of breast cancer.

Methods

Literature search

A comprehensive search of PubMed, Web of Science, OVID, ProQuest and Chinese databases from June 1955 to April 2016 was conducted. The structured search strategy included the following format of key terms: [(vegetable OR fruit OR diet OR nutrition)] AND [(breast cancer OR breast neoplasm OR breast carcinoma)] AND [(prognosis OR disease progression OR disease-free survival OR mortality OR survival rate OR survival analysis OR medical futility OR treatment outcome OR treatment failure OR causes of death OR fatal outcome OR recurrence)]. We also carried out manual searches of the reference lists of eligible articles and pertinent reviews. Only full-length, original articles were considered, and no attempt was made to include abstracts or unpublished results. No language restrictions were imposed. The titles and abstracts of all identified articles were screened by one investigator (C. P.) for eligibility. Two investigators (C. P. and W.-P. L.) independently reviewed the full texts of the remaining articles to identify eligible studies, with differences in opinion resolved by consensus.

Study selection criteria

Studies were eligible for the analysis if (i) the study design was a prospective cohort study; (ii) data related to the dietary consumption of fruits and vegetables were available; (iii) prognostic indicators were restricted to recurrence, all-cause mortality and breast cancer-specific mortality( Reference Li, Lu and Wang 29 ); and (iv) the study provided OR, relative risk (RR) or hazard ratio (HR) and CI data associating fruit and vegetable intake with breast cancer prognosis( Reference Aune, Chan and Vieira 30 Reference Gandini, Merzenich and Robertson 32 ). When multiple publications covered the same population, only the most recently published report was included in the analysis. Case reports, reviews, articles without full text, animal studies and in vitro studies were also excluded.

Data extraction and quality assessment

Two reviewers (C. P. and W.-P. L.) independently reviewed and extracted the data. Any discrepancies between the two reviewers were resolved by discussion until a consensus was reached. The study characteristics were recorded using a standardised data collection form, which included the name of the first author, publication year, country or region, study design, number of cases, length of follow-up, dietary assessment, exposure or outcome, comparison, OR, RR or HR from the most fully adjusted model for the highest v. the lowest fruit and vegetable intakes and their corresponding 95 % CI; confounders were adjusted for in the multivariate analysis.

The quality of the included studies, including selection of the study population, the comparability of the study groups and the outcome or exposure assessment, was assessed on the basis of the Newcastle–Ottawa Quality assessment scale (NOS) for assessing the quality of non-randomised studies in meta-analysis( Reference Zeng, Liu and Chen 33 ). Studies were considered to be of high quality if they scored seven or higher of a possible nine points( Reference Li, Yang and Cao 34 , Reference Chak, Rutherford and Steinmaus 35 ).

Statistical analysis

Summary HR were calculated for recurrence and death (all-cause and breast cancer-specific mortality) using generic inverse variance. The generic inverse variance method in RevMan can be used for ratio measures such as HR or RR( 36 ). High v. low meta-analysis was performed with ratio data from the most fully adjusted model for the highest v. the lowest fruit and vegetable intakes. HR is used to represent ratio measures of effect, including RR. Adjusted HR estimates are summarised using a fixed or random effects model. The random effects model was used to consider both within-study and between-study variations. Assessment of heterogeneity among the studies was performed by using Q and I 2 statistics. P-values of <0·05 or I 2-values over 50 % indicate substantial heterogeneity( Reference Higgins, Thompson and Deeks 37 ). If substantial heterogeneity exists, the random effects model is appropriate; otherwise, the fixed effects model is preferred( Reference DerSimonian and Laird 38 ). Pre-specified stratified analyses were performed to assess the effects of the various study characteristics. Subgroup analyses were carried out by comparing summary HR in each stratum, including menopausal status, number of patients, countries, timing of dietary intake and the use of tamoxifen, on outcomes. Because of the small number of studies reporting on recurrence or breast cancer-specific mortality, the subgroup analyses were mainly carried out for the association between fruit and vegetable intake and all-cause mortality.

Population-based study cohorts were defined as highly representative( Reference Wells, Shea and O’Connell 39 ). In contrast, studies that selected groups of users, for example, nurses and volunteers, and studies with no description of the derivation were categorised as poorly representative( Reference Wells, Shea and O’Connell 39 ). To assess the stability of the findings, sensitivity analyses to rule out poorly representative results from a single study in the meta-analysis were carried out by excluding each study individually and using both fixed and random effects models to evaluate the robustness of the results. Tests for publication bias were performed by constructing and assessing funnel plots. A symmetric inverted funnel shape arises from a ‘well-behaved’ data set, in which publication bias is unlikely( Reference Sterne and Egger 40 ).

Results

Results of the literature search

According to the search criteria, 4493 records were identified, of which 96 % were excluded from the meta-analysis after reviewing the title and abstract. After reviewing the full text of the remaining 146 articles, twelve articles( Reference Nechuta, Caan and Chen 10 , Reference Pan 16 Reference Buck, Zaineddin and Vrieling 26 ) were included in the meta-analysis. The most common reasons for exclusion were lack of data on fruit and vegetable intake, breast cancer recurrence, all-cause mortality or breast cancer-specific mortality. A flow chart of the procedure used to select studies is shown in Fig. 1.

Fig. 1 Flow chart of the selection of studies included in the meta-analyses. WOS, Web of Science; HR, hazard ratio; RR, relative risk.

Characteristics of studies

The characteristics and quality scores of the included studies are summarised in Table 1. In total, twelve included studies involving 52 962 women were identified, which included 29 295 breast cancer patients from eleven studies( Reference Nechuta, Caan and Chen 10 , Reference Pan 16 Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ) and 23 667 women from one study( Reference Sauvaget, Nagano and Hayashi 23 ). The studies had three outcomes: all-cause mortality in ten studies( Reference Nechuta, Caan and Chen 10 , Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Dal Maso, Zucchetto and Talamini 19 Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ), breast cancer-specific mortality in four studies( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Dal Maso, Zucchetto and Talamini 19 , Reference Sauvaget, Nagano and Hayashi 23 , Reference Buck, Zaineddin and Vrieling 26 ) and recurrence in three studies( Reference Nechuta, Caan and Chen 10 , Reference Thomson, Rock and Thompson 18 , Reference Saxe, Rock and Wicha 25 ). The twelve studies included prospective cohort studies published between 1999 and 2013. In total, seven studies were conducted in the USA( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Thomson, Rock and Thompson 18 , Reference Pierce, Stefanick and Flatt 20 Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 , Reference Saxe, Rock and Wicha 25 ), one in China( Reference Pan 16 ), one in Japan( Reference Sauvaget, Nagano and Hayashi 23 ), one in Germany( Reference Buck, Zaineddin and Vrieling 26 ), one in Italy( Reference Dal Maso, Zucchetto and Talamini 19 ) and one in both the USA and China( Reference Nechuta, Caan and Chen 10 ).

Table 1 Characteristics of included studies and quality score

HR, hazard ratio; NOS, the Newcastle–Ottawa Quality assessment scale; CWLS, the Collaborative Women’s Longevity Study; Q, quartiles; RNK, Rhein-Neckar-Karlsruhe; ER/PR, oestrogen receptor/progesterone receptor; PACE, Prospective Analysis of Case–control studies on Environmental factors and health study group; TNM, tumour node metastasis; LIBCSP, the Long Island Breast Cancer Study Project; NHS, The Nurses’ Health Study; IRB, the Internal Review Board of the University of California, Irvine; ABCPP, the After Breast Cancer Pooling Project; SBCSS, the Shanghai Breast Cancer Survival Study; LACE, the Life After Cancer Epidemiology Study; WHEL, the Women’s Healthy Eating and Living Study; MET, metabolic equivalent tasks; LSS, the Life Span Study; IRBSM, Institutional Review Board of the School of Medicine, University of Michigan.

The number of participants in each study ranged from 149 to 23 667, and the median length of follow-up ranged from 3 to 18 years. Most of the individual studies adjusted for a wide range of potential confounders, including age, BMI, stage of disease, age at diagnosis, hormone-replacement therapy use, alcohol use and physical activity. Most studies used a validated FFQ to assess the consumption of fruits and vegetables; eleven studies were of high quality (NOS score ≥7), with an average NOS score of 8·0.

All-cause mortality

As shown in Fig. 2(a), three included studies( Reference Dal Maso, Zucchetto and Talamini 19 , Reference Pierce, Stefanick and Flatt 20 , Reference Fink, Gaudet and Britton 22 ) with four groups of patients investigated the association between the highest v. the lowest intake of fruits and vegetables combined and all-cause mortality among breast cancer patients. The summary HR (highest v. lowest) from the three studies for all-cause mortality was 1·01 (95 % CI 0·72, 1·42) with moderate heterogeneity (P for heterogeneity=0·09, I 2=54 %). No significant association was found between fruit and vegetable intake and all-cause mortality among the breast cancer patients. The funnel plot was symmetrical, and there was no evidence of publication bias.

Fig. 2 Forest plots of observational studies investigating the association of all-cause mortality with (a) fruit and vegetable intake, (b) total vegetable intake, (c) cruciferous vegetable intake and (d) fruit intake. Highest v. lowest intake.

In all, seven included studies( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ) with eight groups of patients were eligible for the meta-analysis of the association between highest v. lowest total vegetable intake and all-cause mortality among breast cancer patients. The summary HR (highest v. lowest) from the eight groups of breast cancer patients for all-cause mortality was 0·71 (95 % CI 0·43, 1·17) with evidence of high heterogeneity (P for heterogeneity<0·00001, I 2=91 %). However, in the funnel plot, one study( Reference Pan 16 ) deviated far from the vertical axis, and an asymmetrically inverted funnel shape indicates potential publication bias. Therefore, as shown in Fig. 2(b), this study( Reference Pan 16 ) was excluded, and the summary HR (highest v. lowest) from the remaining six studies was 0·96 (95 % CI 0·83, 1·12) with no significant heterogeneity (P for heterogeneity=0·11, I 2=42 %). No significant association was found between total vegetable intake and all-cause mortality among the breast cancer patients. There was no evidence of small-study bias from the funnel plot.

As shown in Fig. 2(c), three included studies( Reference Nechuta, Caan and Chen 10 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Fink, Gaudet and Britton 22 ) with four groups of patients investigated the association between highest v. lowest cruciferous vegetable intakes and all-cause mortality. The summary HR (highest v. lowest) for all-cause mortality was 0·99 (95 % CI 0·89, 1·11) with no significant heterogeneity (P for heterogeneity=0·84, I 2=0 %). No significant association was found between cruciferous vegetable intake and all-cause mortality among the breast cancer patients. The funnel plot was symmetrical, and there was no evidence of publication bias.

As shown in Fig. 2(d), six included studies( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) with seven groups of patients investigated the association between highest v. lowest fruit intake and all-cause mortality. The summary HR (highest v. lowest) from the six studies for all-cause mortality was 0·88 (95 % CI 0·74, 1·05) with no heterogeneity detected (P for heterogeneity=0·11, I 2=43 %). Thus, no significant association was found between fruit intake and all-cause mortality. The funnel plot was symmetrical, and there was no evidence of publication bias.

Breast cancer-specific mortality

As shown in Fig. 3(a), three included studies( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Sauvaget, Nagano and Hayashi 23 , Reference Buck, Zaineddin and Vrieling 26 ) investigated the association between highest v. lowest total vegetable intake and breast cancer-specific mortality. The summary HR (highest v. lowest) for breast cancer-specific mortality was 1·05 (95 % CI 0·77, 1·43), with no evidence of heterogeneity (P for heterogeneity=0·82, I 2=0 %). Thus, no significant association was found between total vegetable intake and breast cancer-specific mortality. The funnel plot was symmetrical, and there was no evidence of publication bias.

Fig. 3 Forest plots of observational studies investigating the association of breast cancer-specific mortality with (a) total vegetable intake and (b) fruit intake. Highest v. lowest intake.

As presented in Fig. 3(b), three included studies( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Sauvaget, Nagano and Hayashi 23 , Reference Buck, Zaineddin and Vrieling 26 ) investigated the association between highest v. lowest fruit intake and breast cancer-specific mortality. The summary HR (highest v. lowest) for breast cancer-specific mortality was 0·94 (95 % CI 0·69, 1·26), with no evidence of heterogeneity (P for heterogeneity=0·54, I 2=0 %). Thus, no significant association was found between fruit intake and breast cancer-specific mortality. The funnel plot was symmetrical, and there was no evidence of publication bias.

Breast cancer recurrence

Two included studies( Reference Thomson, Rock and Thompson 18 , Reference Saxe, Rock and Wicha 25 ) investigated the association between highest v. lowest total vegetable intake and breast cancer recurrence. The summary HR (highest v. lowest) from these two studies for breast cancer recurrence was 0·89 (95 % CI 0·53, 1·50) with substantial heterogeneity shown (P for heterogeneity=0·003, I 2=89 %) in Fig. 4(a). There was no evidence of small-study bias from the funnel plot.

Fig. 4 Forest plots of observational studies investigating the association of breast cancer recurrence with (a) total vegetable intake and (b) cruciferous vegetable intake. Highest v. lowest intake.

As presented in Fig. 4(b), two included studies( Reference Nechuta, Caan and Chen 10 , Reference Thomson, Rock and Thompson 18 ) investigated the association between highest v. lowest cruciferous vegetable intake and breast cancer recurrence. The summary HR (highest v. lowest) was 0·98 (95 % CI 0·76, 1·26) with moderate heterogeneity (P for heterogeneity=0·05, I 2=73 %). The funnel plot was symmetrical, and there was no evidence of publication bias.

Subgroup analyses

Subgroup analyses of the associations of fruit and vegetable intake with all-cause mortality and breast cancer recurrence are summarised in Table 2.

Table 2 Summary of subgroup analyses of the effects of fruit and vegetable intake on all-cause mortality and breast cancer recurrence (Hazard ratios (HR) and 95 % confidence intervals)

HR, hazard ratio; BC, breast cancer.

* P-values were obtained using test for subgroup differences to compare each group.

Four included studies investigated the association between total vegetable intake( Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) or fruit intake( Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) and all-cause mortality stratified by menopausal status, respectively. No significant association was found between total vegetable intake or fruit intake and all-cause mortality when stratified by menopausal status. The summary HR of total vegetable intake were 0·92 (95 % CI 0·76, 1·12) for those who were postmenopausal and 1·18 (95 % CI 0·71, 1·95) for those who were premenopausal. Those of fruit intake were 0·88 (95 % CI 0·72, 1·08) for postmenopausal patients and 0·90 (95 % CI 0·41, 1·97) for premenopausal patients.

Six included studies with seven groups of patients were eligible for the meta-analysis of the association between total vegetable intake( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ) or fruit intake( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) and all-cause mortality stratified by number of patients. Three studies included 4441( Reference Beasley, Newcomb and Trentham-Dietz 17 ), 1982( Reference Holmes, Stampfer and Colditz 24 ) and 2563( Reference Buck, Zaineddin and Vrieling 26 ) breast cancer patients for total vegetable intake, and three studies included 1264( Reference Pan 16 ), 4441( Reference Beasley, Newcomb and Trentham-Dietz 17 ) and 2563( Reference Buck, Zaineddin and Vrieling 26 ) cases for fruit intake, whereas three other studies with four groups of patients included 516( Reference McEligot, Largent and Ziogas 21 ), 376( Reference Fink, Gaudet and Britton 22 ), 834( Reference Fink, Gaudet and Britton 22 ) and 149( Reference Saxe, Rock and Wicha 25 ) breast cancer patients for total vegetable or fruit intake. Stratified by number of patients, the summary HR of total vegetable intake were 1·05 (95 % CI 0·78, 1·42) for studies with more than 1000 patients and 0·89 (95 % CI 0·66, 1·22) for those with less than 1000 patients. The summary HR of fruit intake were 0·85 (95 % CI 0·51, 1·42) for those with more than 1000 patients and 0·88 (95 % CI 0·68, 1·13) for those with less than 1000 patients.

Six included studies with seven groups of patients were eligible for the meta-analysis of the association between total vegetable intake( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Holmes, Stampfer and Colditz 24 Reference Buck, Zaineddin and Vrieling 26 ) or fruit intake( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) and all-cause mortality stratified by timing of dietary intake information. Four studies( Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) were based on information on diet collected before diagnosis or dietary habits before diagnosis and three studies collected after diagnosis( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Holmes, Stampfer and Colditz 24 ). The summary HR of total vegetable intake were 0·95 (95 % CI 0·75, 1·20) for those whose information on diet was collected before diagnosis( Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) and 1·05 (95 % CI 0·60, 1·85) for those collected after diagnosis( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Holmes, Stampfer and Colditz 24 ). The summary HR of fruit intake were 0·86 (95 % CI 0·71, 1·05) for those collected before diagnosis( Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) and 0·82 (95 % CI 0·29, 2·36) for those collected after diagnosis( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 ).

Six included studies( Reference Pan 16 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Fink, Gaudet and Britton 22 , Reference Saxe, Rock and Wicha 25 , Reference Buck, Zaineddin and Vrieling 26 ) were eligible for the meta-analysis of the association between fruit intake and all-cause mortality stratified by countries. No significant association was found between fruit intake and all-cause mortality among breast cancer patients in the studies from the USA and the study from Germany, respectively. In the study data from the USA, the summary HR for all-cause mortality was 0·98 (95 % CI 0·79, 1·21). Of the single study from Germany and that from China, only the HR from the study in China indicated statistical significance.

Two included studies( Reference Nechuta, Caan and Chen 10 , Reference Thomson, Rock and Thompson 18 ) investigated the association between cruciferous vegetable intake and breast cancer recurrence stratified by the use of tamoxifen. No significant association was found between cruciferous vegetable intake and breast cancer recurrence when stratified by tamoxifen use. In the group of patients taking tamoxifen, the summary HR was 1·12 (95 % CI 0·88, 1·43), whereas in the group of patients not taking tamoxifen the summary HR was 0·83 (95 % CI 0·53, 1·29).

Sensitivity analyses

Sensitivity analyses to examine the effects of outliers were performed by using both fixed and random effects models and excluding each study individually and re-calculating the summary HR. Most of the results showed that the summary HR of each meta-analysis remained statistically non-significant. Only the sensitivity analysis excluding one study( Reference Beasley, Newcomb and Trentham-Dietz 17 ) yielded a different result, showing that the summary HR for the association of fruit intake with all-cause mortality was 0·82 (95 % CI, 0·68, 0·99) with no heterogeneity (P for heterogeneity=0·30, I 2=17 %), which may indicate the instability of the finding of the association between fruit intake and all-cause mortality (data not shown).

Discussion

The studies included in this meta-analysis examined four types of fruit and vegetable intakes including fruits and vegetables combined, total vegetable intake, cruciferous vegetable intake and fruit intake. The studies had three outcomes, including all-cause mortality, breast cancer-specific mortality and breast cancer recurrence. In this meta-analysis, no significant association was found between fruits and vegetables combined, total vegetable intake, cruciferous vegetable intake or fruit intake and all-cause mortality, breast cancer-specific mortality or breast cancer recurrence, respectively.

To our knowledge, this is the first meta-analysis to assess a possible association between fruit and vegetable intake and breast cancer prognosis. The ability of this meta-analysis to evaluate the effect of fruit and vegetable intake on breast cancer prognosis might be limited because the types of fruit and vegetable intake or the outcome indicators of breast cancer prognosis were investigated differently among the included studies. Some studies mainly investigated the association of total vegetable or fruit intake with all-cause mortality or breast cancer-specific mortality( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference McEligot, Largent and Ziogas 21 , Reference Sauvaget, Nagano and Hayashi 23 , Reference Holmes, Stampfer and Colditz 24 , Reference Buck, Zaineddin and Vrieling 26 ), whereas others focused only on breast cancer recurrence( Reference Nechuta, Caan and Chen 10 , Reference Thomson, Rock and Thompson 18 , Reference Saxe, Rock and Wicha 25 ). Only one included study evaluated the association between fruits and vegetables combined( Reference Dal Maso, Zucchetto and Talamini 19 ) or cruciferous vegetable intake( Reference Beasley, Newcomb and Trentham-Dietz 17 ) and breast cancer-specific mortality, respectively. Similarly, only one or no included study investigated the association between fruit intake( Reference Saxe, Rock and Wicha 25 ) or fruits and vegetables combined and breast cancer recurrence, respectively. Thus, the types of fruit and vegetable intake and the outcomes of breast cancer prognosis were inconsistent; therefore, this meta-analysis mainly evaluated the summary analysis on all-cause mortality.

The dose–response relationship of fruit and vegetable intake with all-cause mortality might be impossible to estimate in the present meta-analysis due to inconsistent assessment. Some studies used servings per day( Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Dal Maso, Zucchetto and Talamini 19 , Reference Pierce, Stefanick and Flatt 20 , Reference McEligot, Largent and Ziogas 21 , Reference Holmes, Stampfer and Colditz 24 ) or quartiles with gram per day( Reference Nechuta, Caan and Chen 10 ), and others indicated a rough assessment criteria( Reference Pan 16 , Reference Thomson, Rock and Thompson 18 , Reference Sauvaget, Nagano and Hayashi 23 ). Furthermore, some studies assessed the intake with quartiles( Reference Nechuta, Caan and Chen 10 , Reference Beasley, Newcomb and Trentham-Dietz 17 , Reference Holmes, Stampfer and Colditz 24 ) and some with tertiles( Reference McEligot, Largent and Ziogas 21 , Reference Buck, Zaineddin and Vrieling 26 ). Most of the included studies indicated the HR of intake with highest v. lowest, but one study reported the HR with the highest quartile as the reference( Reference Dal Maso, Zucchetto and Talamini 19 ). Although these factors constrained the ability to find a dose–response relationship and might slightly have influenced the summary HR investigated in this meta-analysis, sensitivity analyses were carried out. Most of them confirmed the robustness of the findings. Only the sensitivity analysis excluding one study( Reference Beasley, Newcomb and Trentham-Dietz 17 ) investigating the relationship between fruit intake and all-cause mortality yielded different results, which showed that the excluded one( Reference Beasley, Newcomb and Trentham-Dietz 17 ) might be poorly representative. However, no heterogeneity was detected on the summary HR for the association of fruit intake with all-cause mortality, which indicated that there is no difference among the included studies. Thus, this study( Reference Beasley, Newcomb and Trentham-Dietz 17 ) was included for the meta-analysis.

The findings of this meta-analysis reported a non-significant association between fruit and vegetable intake and breast cancer prognosis, probably because the consumption of fruits and vegetables may not be a strong dietary determinant of breast cancer mortality but has an indirect effect on breast cancer prognosis( Reference Hill 41 , Reference Verhoeven, Assen and Goldbohm 42 ). Many related articles regarding the association of fruit and vegetable intake with breast cancer prognosis mainly focus on the association between fruit and vegetable intake and weight change among breast cancer patients( Reference Erickson 43 Reference Chlebowski 47 ). Thomson et al.( Reference Thomson, Rock and Giuliano 48 ) indicated minimal short-term changes in body weight in response to a high-vegetable, low-fat diet in breast cancer survivors. However, several studies have implicated obesity as being positively associated with breast cancer mortality( Reference Donegan, Hartz and Rimm 49 Reference McTiernan, Irwin and Vongruenigen 53 ). Stokes et al.( Reference Stokes 54 ) indicated that an unhealthy body weight is an important modifiable risk factor for the recurrence of breast cancer. Hauner & Hauner( Reference Hauner and Hauner 55 ) found that obesity is a potent risk factor for both cancer development and prognosis, increasing the risk for overall and breast cancer-specific mortality by approximately 30 %. On the whole, it is suggested that a healthy dietary pattern that increases fruit and vegetable intake and reduces dietary fat intake and is associated with modest weight loss may influence breast cancer prognosis. Thus, it is likely that the association of fruit and vegetable intake with breast cancer prognosis should be investigated with weight change among breast cancer patients.

Subgroup analyses stratified by menopausal states, number of patients and use of tamoxifen showed no different effects in each group. First, hormone-related factors such as late menopause or use of postmenopausal hormone therapy have been known to be associated with an increased risk of breast cancer( Reference Key, Verkasalo and Banks 56 ). Bernstein & Ross( Reference Bernstein and Ross 57 ) found that endogenous sex hormones play a role in the development of breast cancer. Further, of the natural forms of oestrogen in women, oestrone (E1) is predominately found in postmenopausal women, whereas oestradiol (E2) is the major form in women of reproductive age( Reference Liu, Huang and Wang 58 ). In addition, the clinical implications of oestrogen receptor and progesterone receptor are different between premenopausal and postmenopausal patients with breast cancer( Reference Xiao, Tao and Duan 59 ). Some constituents mainly present in fruits and vegetables, such as phyto-oestrogens, may either have similar effects as oestrogen (agonistic) or block oestrogen’s effects (antagonistic)( Reference Jung, Yu and Lau 60 ). These compounds structurally resemble oestrogen and exert their effects primarily through binding to the oestrogen receptor( Reference Turner, Agatonovic-Kustrin and Glass 61 ). However, not all included studies so far published had enough information on hormonal receptor breast cancer subtypes. Therefore, analysis stratified by hormonal receptor status cannot be conducted. The findings of the subgroup analyses stratified by menopausal status also did not show any statistical significance. More epidemiological studies are needed to further investigate the influence of sex hormone levels on which fruit and vegetable intake may impact in promoting breast cancer. Second, analysis stratified by the number of patients was conducted to determine potential sources of heterogeneity and to see whether this factor influences the results. No significant difference in the subgroup was found. Therefore, the number of patients did not seem to be relevant to explain variability between studies, which confirmed the robustness of the effects of total fruit or vegetable intake on all-cause mortality. Finally, tamoxifen is an important oestrogen receptor antagonist used successfully for the treatment and prevention of breast cancer( Reference Williams, Twaddle and Churchwell 62 ). One included study( Reference Thomson, Rock and Thompson 18 ) suggested that baseline vegetable intake may be associated with a reduction in the risk of breast cancer recurrence, particularly for those using tamoxifen, whereas another study( Reference Nechuta, Caan and Chen 10 ) found no significance in the presence or absence of tamoxifen. However, existing mechanistic evidence suggests a synergistic anti-carcinogenic action of combinational interventions with selected bioactive constituents in cruciferous vegetables and both tamoxifen and oestrogen receptor-α ( Reference Malejka-Giganti, Parkin and Bennett 63 Reference Ramirez and Singletary 65 ). Thus, these findings encourage future investigations of the possible effects of fruits or vegetables on breast cancer prognosis.

In the subgroup analysis stratified by countries, the HR in the single study from China indicated a significant association between fruit intake and all-cause mortality. However, no significant association was indicated by the summary HR in the studies from the USA or in the study from Germany. Although breast cancer mortality rates have steadily increased in recent years, one study also found that east Asian countries have a lower rate, which may be associated with their low-fat diet and higher intake of fruits and vegetables when compared with their Western counterparts( Reference Torre, Bray and Siegel 3 ). However, epidemiological studies evaluating the association of fruit and vegetable intake with breast cancer prognosis in Asia were limited in that only two included studies were from Asian countries( Reference Pan 16 , Reference Sauvaget, Nagano and Hayashi 23 ) in this meta-analysis. Thus, more epidemiological studies examining regional differences in diet are expected to explore the association between fruit intake and all-cause mortality among breast cancer patients, especially in Asian countries.

No significant association was found between total vegetable or total fruit intake and all-cause mortality when stratified by timing of collection of dietary intake information. Some researchers suggested that antioxidant supplements provided during treatment could repair cellular oxidative damage to cancer cells caused by treatments such as radiation therapy and chemotherapy( Reference Labriola and Livingston 66 , Reference Lamson and Brignall 67 ), which probably explain why fruit and vegetable intake following diagnosis may not be beneficial to breast cancer patients. However, Velentzis et al.( Reference Velentzis, Keshtgar and Woodside 68 ) reported that patients were likely to make significant changes to their diet and adopt healthier food choices after diagnosis of breast cancer, potentially decreasing their risk of breast cancer recurrence and reducing the risk of developing other co-morbidities including diabetes and heart disease, which probably account for all-cause mortality. Izano et al. ( Reference Izano, Fung and Chiuve 69 ) indicated that healthy dietary choices after breast cancer diagnosis did not change the risk of breast cancer death and recurrence but were associated with a reduced risk of non-breast cancer mortality. Studies included in the present meta-analysis also indicated that high post-diagnostic fruit and vegetable intake decreased the risk of mortality following breast cancer( Reference Holmes, Stampfer and Colditz 24 , Reference Saxe, Rock and Wicha 25 ). Moreover, it is possible that pre-diagnosis diets may have reflected taste or appetite changes resulting from breast cancer( Reference Saxe, Rock and Wicha 25 ). Some studies indicated that pre-diagnosis diet associated with risk of breast cancer may be related to progression following the diagnosis of breast cancer and among the determinants of tissue characteristics that influence prognosis( Reference Rohan, Hiller and McMichael 70 , Reference Russo and Russo 71 ). Dal et al.( Reference Dal Maso, Zucchetto and Talamini 19 ) found that women eating <4 servings of fruits and vegetables/d before diagnosis showed higher risk for all-cause and breast cancer mortality compared with those who consumed 6 or more servings/d. Therefore, to pay more attention to post-diagnosis dietary changes and pre-morbid dietary habits in relation to breast cancer prognosis, more studies are needed to evaluate fruit and vegetable intake before and after diagnosis and its effects on breast cancer prognosis.

Although this meta-analysis provided no evidence that a higher fruit and vegetable intake decreased the risk of all-cause mortality among breast cancer patients, several studies indicated that dietary intakes of nutrients common in fruits and vegetables seem to be associated with a better prognosis among breast cancer patients. Fruits and vegetables are common sources of many candidate protective substances, including ascorbic acid and carotenoids( Reference Gandini, Merzenich and Robertson 32 ). Harris et al. ( Reference Harris, Orsini and Wolk 72 ) suggested that post-diagnosis dietary vitamin C intake was statistically significantly associated with a reduced risk of total mortality and breast cancer-specific mortality. Plasma total carotenoid concentration was reported to be inversely associated with breast cancer recurrence in a cohort of 1551 women who had been diagnosed with early-stage breast cancer( Reference Rock, Flatt and Natarajan 73 ). Dietary fibre was found to be inversely associated with overall mortality among breast cancer patients( Reference McEligot, Largent and Ziogas 21 ). Moreover, some phytochemicals such as monoterpenes, resveratrol and lignans present in fruits and vegetables are also protective( Reference Tsubura, Uehara and Kiyozuka 74 ). Some in vitro studies and animal models have also shown that nutrients from fruits and vegetables may inhibit the metastasis of breast cancer cells. Kim et al.( Reference Kim, Sehrawat and Singh 75 ) found that a non-toxic, small-molecule constituent of edible cruciferous vegetables (benzyl isothiocyanate) inhibits mammary cancer development in mouse mammary tumour virus-neu transgenic mice by causing epithelial tumour cell apoptosis. Koh et al.( Reference Koh, Hwang and Moon 76 ) reported the significant inhibitory effect of lycopene on the invasive and migratory phenotypes of two highly aggressive breast cancer cell lines. Xu et al.( Reference Xu, Bower and Wang 77 ) evaluated the effects of cyanidin-3-glucoside (C3G), an anthocyanin present in many fruits and vegetables, on ethanol-induced breast cancer cell migration/invasion and found that C3G blocks ethanol-induced activation of the ErbB2/cSrc/FAK pathway, which is necessary for cell migration and invasion.

There are some potential limitations to this meta-analysis. First, only twelve articles( Reference Nechuta, Caan and Chen 10 , Reference Pan 16 Reference Buck, Zaineddin and Vrieling 26 ) with inconsistency in the types of fruit and vegetable intake and the prognostic indicators were included. Thus, the number of included articles might be too small to yield reliable summary results of breast cancer-specific mortality or breast cancer recurrence. Second, this meta-analysis did not estimate the dose–response relationship of fruit and vegetable intake with breast cancer prognosis and subgroup analysis stratified by other factors, such as pre- and post-diagnosis diet, because of lack of data. Finally, two funnel plots for the subgroup analyses stratified by menopausal state showed some asymmetry, indicating the risk of publication bias. However, the present meta-analysis also has several strengths. As all the included studies were prospective cohorts, we have effectively avoided recall and selection bias. Our clear delineation of the research questions and selection criteria, the comprehensive search strategy used and the objective assessment of the quality of the studies may have increased the validity of the findings. Moreover, with one exception, all the studies were high-quality cohorts. All but one of the sensitivity analyses yielded similar results, thus suggesting the stability of the findings. Most funnel plots were symmetrical, indicating that the results were unlikely to be due to publication bias.

In conclusion, no significant associations were found between fruit and vegetable intake (fruits and vegetables combined, total vegetable intake, cruciferous vegetable intake and fruit intake) and breast cancer prognosis (all-cause mortality, breast cancer-specific mortality and breast cancer recurrence). More studies are needed to determine whether the consumption of fruits and vegetables plays an important role in improving breast cancer prognosis. Additional studies on weight change, regional differences in diet, pre-diagnosis diet and dietary changes made after diagnosis are needed to confirm the effects of fruit and vegetable intake on breast cancer prognosis.

Acknowledgements

The authors gratefully acknowledge the student helpers for literature search in this study.

This study was supported by Science and Technology Program of Guangzhou, China (no. 201510010151), and the National Natural Science Foundation of China (no. 81102188). The funders had no role in the design, analysis or writing of this article.

The authors’ responsibilities are as follows: C. P. conceived and designed the study, analysed the data, collected information and wrote the paper. W.-P. L. analysed the data and collected information. C.-X. Z. conceived and designed the study, analysed the data, collected information, supervised and contributed to writing of the paper.

The authors declare that there are no conflicts of interest.

References

1. Yue, Y (2011) Research progress on risk factors of breast cancer. Oncology Progress 5, 520523.Google Scholar
2. McGuire, A, Brown, JA, Malone, C, et al. (2015) Effects of age on the detection and management of breast cancer. Cancers 7, 908929.CrossRefGoogle ScholarPubMed
3. Torre, LA, Bray, F, Siegel, RL, et al. (2015) Global cancer statistics, 2012. CA Cancer J Clin 65, 87108.CrossRefGoogle ScholarPubMed
4. Forman, D & Ferlay, J (2014) The global and regional burden of cancer. In World Cancer Report 2014, pp. 1653 [BW Stewart and CP Wild, editors]. Lyon, France: International Agency for Research on Cancer.Google Scholar
5. Brewster, A & Helzlsouer, K (2001) Breast cancer epidemiology, prevention, and early detection. Curr Opin Oncol 13, 420425.CrossRefGoogle ScholarPubMed
6. Reuben, SC, Gopalan, A, Petit, DM, et al. (2012) Modulation of angiogenesis by dietary phytoconstituents in the prevention and intervention of breast cancer. Mol Nutr Food Res 56, 1429.CrossRefGoogle ScholarPubMed
7. Chatterjee, M, Roy, K, Janarthan, M, et al. (2012) Biological activity of carotenoids: its implications in cancer risk and prevention. Curr Pharm Biotechnol 13, 180190.CrossRefGoogle ScholarPubMed
8. Liu, J, Sun, XR, Dong, HW, et al. (2008) β-Ionone suppresses mammary carcinogenesis, proliferative activity and induces apoptosis in the mammary gland of the Sprague-Dawley rat. Int J Cancer 122, 26892698.CrossRefGoogle ScholarPubMed
9. Liu, JR, Dong, HW, Sun, XR, et al. (2010) Effects of beta-ionone on mammary carcinogenesis and antioxidant status in rats treated with DMBA. Nutr Cancer 62, 5865.CrossRefGoogle ScholarPubMed
10. Nechuta, S, Caan, BJ, Chen, WY, et al. (2013) Postdiagnosis cruciferous vegetable consumption and breast cancer outcomes: a report from the After Breast Cancer Pooling Project. Cancer Epidemiol Biomarkers Prev 22, 14511456.CrossRefGoogle ScholarPubMed
11. Fimognari, C, Lenzi, M & Hrelia, P (2008) Chemoprevention of cancer by isothiocyanates and anthocyanins: mechanisms of action and structure-activity relationship. Curr Med Chem 15, 440447.CrossRefGoogle ScholarPubMed
12. Azarenko, O (2009) Dietary isothiocyanates suppress microtubule dynamic instability in association with mitotic arrest and inhibition of tumor cell proliferation, pp. 206. University of California, Santa Barbara, Ann Arbor.Google Scholar
13. Ferguson, LR & Schlothauer, RC (2012) The potential role of nutritional genomics tools in validating high health foods for cancer control: broccoli as example. Mol Nutr Food Res 56, 126146.CrossRefGoogle ScholarPubMed
14. Turrini, E, Ferruzzi, L & Fimognari, C (2015) Potential effects of pomegranate polyphenols in cancer prevention and therapy. Oxid Med Cell Longev 2015, 938475.CrossRefGoogle Scholar
15. Seeram, NP, Adams, LS, Henning, SM, et al. (2005) In vitro antiproliferative, apoptotic and antioxidant activities of punicalagin, ellagic acid and a total pomegranate tannin extract are enhanced in combination with other polyphenols as found in pomegranate juice. J Nutr Biochem 16, 360367.CrossRefGoogle Scholar
16. Pan, ZT (2003) Study on survival and prognostic factors of female breast cancer in urban Shanghai. Masters thesis, Fudan University, Shanghai, China.Google Scholar
17. Beasley, JM, Newcomb, PA, Trentham-Dietz, A, et al. (2011) Post-diagnosis dietary factors and survival after invasive breast cancer. Breast Cancer Res Treat 128, 229236.CrossRefGoogle ScholarPubMed
18. Thomson, CA, Rock, CL, Thompson, PA, et al. (2011) Vegetable intake is associated with reduced breast cancer recurrence in tamoxifen users: a secondary analysis from the Women’s Healthy Eating and Living Study. Breast Cancer Res Treat 125, 519527.CrossRefGoogle ScholarPubMed
19. Dal Maso, L, Zucchetto, A, Talamini, R, et al. (2008) Effect of obesity and other lifestyle factors on mortality in women with breast cancer. Int J Cancer 123, 21882194.CrossRefGoogle ScholarPubMed
20. Pierce, JP, Stefanick, ML, Flatt, SW, et al. (2007) Greater survival after breast cancer in physically active women with high vegetable-fruit intake regardless of obesity. J Clin Oncol 25, 23452351.CrossRefGoogle ScholarPubMed
21. McEligot, AJ, Largent, J, Ziogas, A, et al. (2006) Dietary fat, fiber, vegetable, and micronutrients are associated with overall survival in postmenopausal women diagnosed with breast cancer. Nutr Cancer 55, 132140.CrossRefGoogle ScholarPubMed
22. Fink, BN, Gaudet, MM, Britton, JA, et al. (2006) Fruits, vegetables, and micronutrient intake in relation to breast cancer survival. Breast Cancer Res Treat 98, 199208.CrossRefGoogle ScholarPubMed
23. Sauvaget, C, Nagano, J, Hayashi, M, et al. (2003) Vegetables and fruit intake and cancer mortality in the Hiroshima/Nagasaki Life Span Study. Br J Cancer 88, 689694.CrossRefGoogle ScholarPubMed
24. Holmes, MD, Stampfer, MJ, Colditz, GA, et al. (1999) Dietary factors and the survival of women with breast carcinoma. Cancer 86, 826835.3.0.CO;2-0>CrossRefGoogle ScholarPubMed
25. Saxe, GA, Rock, CL, Wicha, MS, et al. (1999) Diet and risk for breast cancer recurrence and survival. Breast Cancer Res Treat 53, 241253.CrossRefGoogle ScholarPubMed
26. Buck, K, Zaineddin, AK, Vrieling, A, et al. (2011) Estimated enterolignans, lignan-rich foods, and fibre in relation to survival after postmenopausal breast cancer. Br J Cancer 105, 11511157.CrossRefGoogle ScholarPubMed
27. Pierce, JP, Natarajan, L, Caan, BJ, et al. (2007) Influence of a diet very high in vegetables, fruit, and fiber and low in fat on prognosis following treatment for breast cancer: the Women’s Healthy Eating and Living (WHEL) randomized trial. JAMA 298, 289298.CrossRefGoogle ScholarPubMed
28. Hauner, D, Janni, W, Rack, B, et al. (2011) The effect of overweight and nutrition on prognosis in breast cancer. Dtsch Arztebl Int 108, 795801.Google ScholarPubMed
29. Li, B, Lu, Y, Wang, L, et al. (2015) Folate intake and breast cancer prognosis. Eur J Cancer Prev 24, 113121.CrossRefGoogle ScholarPubMed
30. Aune, D, Chan, DS, Vieira, AR, et al. (2012) Fruits, vegetables and breast cancer risk: a systematic review and meta-analysis of prospective studies. Breast Cancer Res Treat 134, 479493.CrossRefGoogle Scholar
31. Smith-Warner, SA, Spiegelman, D, Yaun, SS, et al. (2001) Intake of fruits and vegetables and risk of breast cancer: a pooled analysis of cohort studies. JAMA 285, 769776.CrossRefGoogle ScholarPubMed
32. Gandini, S, Merzenich, H, Robertson, C, et al. (2000) Meta-analysis of studies on breast cancer risk and diet: the role of fruit and vegetable consumption and the intake of associated micronutrients. Eur J Cancer 36, 636646.CrossRefGoogle ScholarPubMed
33. Zeng, XT, Liu, H, Chen, X, et al. (2012) The fourth of meta analyses series: the quality assessment tool of observational studies. Chin J Evid Cardiovasc Med 4, 297299.Google Scholar
34. Li, Y, Yang, H & Cao, J (2011) Association between alcohol consumption and cancers in the Chinese population—a systematic review and meta-analysis. PLoS ONE 6, 14511453.Google ScholarPubMed
35. Chak, E, Rutherford, GW & Steinmaus, C (2009) The role of breast-feeding in the prevention of Helicobacter pylori infection: a systematic review. Clin Infect Dis 48, 430437.CrossRefGoogle ScholarPubMed
36. The Australasian Cochrane Centre (2008) Part 5 Data and analyses: using generic inverse variance. In Review Manager 5.3 Tutorial (revised in June 2014), pp. 33–35. [staff at the Australasian Cochrane Centre and the Thai Cochrane Network and the Information Management System team (now IKMD, the Cochrane Informatics and Knowledge Management Department), editors]. Melbourne, Australia. http://community.cochrane.org/tools/review-production-tools/revman-5/revman-5-download (accessed April 2016).Google Scholar
37. Higgins, JP, Thompson, SG, Deeks, JJ, et al. (2003) Measuring inconsistency in meta-analyses. BMJ 327, 557560.CrossRefGoogle ScholarPubMed
38. DerSimonian, R & Laird, N (1986) Meta-analysis in clinical trials. Control Clin Trials 7, 177188.CrossRefGoogle ScholarPubMed
39. Wells, GA, Shea, B, O’Connell, D, et al. (2016) The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm (accessed April 2016).Google Scholar
40. Sterne, JA & Egger, M (2001) Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemiol 54, 10461055.CrossRefGoogle ScholarPubMed
41. Hill, MJ (1997) Nutrition and human cancer. Ann N Y Acad Sci 833, 6878.CrossRefGoogle ScholarPubMed
42. Verhoeven, DT, Assen, N, Goldbohm, RA, et al. (1997) Vitamins C and E, retinol, beta-carotene and dietary fibre in relation to breast cancer risk: a prospective cohort study. Br J Cancer 75, 149155.CrossRefGoogle Scholar
43. Erickson, KD (2011) Diabetes and breast cancer: the women’s healthy eating & living study, pp. 96. University of California, Ann Arbor.Google Scholar
44. George, SAM (2010) Energy balance, risk of breast cancer, and prognosis after breast cancer: investigating the complex roles of diet and physical activity, pp. 108. Yale University, Ann Arbor.Google Scholar
45. Thomson, CA (1998) The effect of a high fruit and vegetable, low fat dietary intervention on immune function, DNA adduct formation, and body composition among breast cancer survivors, pp. 332. The University of Arizona, Ann Arbor.Google Scholar
46. Assaf, AR, Beresford, S, Risica, PM, et al. (2016) Low-fat dietary pattern intervention and health-related quality of life: the women’s health initiative randomized controlled dietary modification trial. J Acad Nutr Diet 116, 259271.CrossRefGoogle ScholarPubMed
47. Chlebowski, R (2007) Lifestyle change including dietary fat reduction and breast cancer outcome. J Nutr 137S, 233S235S.CrossRefGoogle Scholar
48. Thomson, CA, Rock, CL, Giuliano, AR, et al. (2005) Longitudinal changes in body weight and body composition among women previously treated for breast cancer consuming a high-vegetable, fruit and fiber, low-fat diet. Eur J Nutr 44, 1825.CrossRefGoogle ScholarPubMed
49. Donegan, WL, Hartz, AJ & Rimm, AA (1978) The association of body weight with recurrent cancer of the breast. Cancer 41, 15901594.3.0.CO;2-N>CrossRefGoogle ScholarPubMed
50. Reeves, MM, Terranova, CO, Eakin, EG, et al. (2014) Weight loss intervention trials in women with breast cancer: a systematic review. Obes Rev 15, 749768.CrossRefGoogle ScholarPubMed
51. Rodriguez, SFM, Aguilar, MA & Manuel-y-Keenoy, B (2013) [Influence of body weight on the prognosis of breast cancer survivors; nutritional approach after diagnosis]. Nutr Hosp 28, 18291841.Google Scholar
52. Champ, CE, Volek, JS, Siglin, J, et al. (2012) Weight gain, metabolic syndrome, and breast cancer recurrence: are dietary recommendations supported by the data? Int J Breast Cancer 2012, 506868.CrossRefGoogle Scholar
53. McTiernan, A, Irwin, M & Vongruenigen, V (2010) Weight, physical activity, diet, and prognosis in breast and gynecologic cancers. J Clin Oncol 28, 40744080.CrossRefGoogle ScholarPubMed
54. Stokes, SA (2014) Dietary behaviors and nutrition intervention preferences of breast cancer survivors, pp. 73. Clemson University, Ann Arbor.Google Scholar
55. Hauner, H & Hauner, D (2009) The impact of nutrition on the development and prognosis of breast cancer. Breast Care (Basel) 5, 377381.Google Scholar
56. Key, TJ, Verkasalo, PK & Banks, E (2001) Epidemiology of breast cancer. Lancet Oncol 2, 133140.CrossRefGoogle ScholarPubMed
57. Bernstein, L & Ross, RK (1993) Endogenous hormones and breast cancer risk. Epidemiol Rev 15, 4865.CrossRefGoogle ScholarPubMed
58. Liu, MM, Huang, Y & Wang, J (2012) Developing phytoestrogens for breast cancer prevention. Anticancer Agents Med Chem 12, 13061313.CrossRefGoogle ScholarPubMed
59. Xiao, JJ, Tao, M, Duan, WM, et al. (2009) Expressions of c-erbB-2, ER, PR and their prognostic correlation in premenopausal and postmenopausal patients with breast cancer. Cancer Res Prev Treat 36, 666668.Google Scholar
60. Jung, W, Yu, O, Lau, SM, et al. (2000) Identification and expression of isoflavone synthase, the key enzyme for biosynthesis of isoflavones in legumes. Nat Biotechnol 18, 208212.CrossRefGoogle ScholarPubMed
61. Turner, JV, Agatonovic-Kustrin, S & Glass, BD (2007) Molecular aspects of phytoestrogen selective binding at estrogen receptors. J Pharm Sci 96, 18791885.CrossRefGoogle ScholarPubMed
62. Williams, LD, Twaddle, NC, Churchwell, MI, et al. (2006) Quantification of tamoxifen and metabolites and soy isoflavones in human plasma using liquid chromatography with electrospray ionization tandem mass spectrometry. J Aoac Int 89, 11681173.CrossRefGoogle ScholarPubMed
63. Malejka-Giganti, D, Parkin, DR, Bennett, KK, et al. (2007) Suppression of mammary gland carcinogenesis by post-initiation treatment of rats with tamoxifen or indole-3-carbinol or their combination. Eur J Cancer Prev 16, 130141.CrossRefGoogle ScholarPubMed
64. Wang, TT, Milner, MJ, Milner, JA, et al. (2006) Estrogen receptor alpha as a target for indole-3-carbinol. J Nutr Biochem 17, 659664.CrossRefGoogle ScholarPubMed
65. Ramirez, MC & Singletary, K (2009) Regulation of estrogen receptor alpha expression in human breast cancer cells by sulforaphane. J Nutr Biochem 20, 195201.CrossRefGoogle ScholarPubMed
66. Labriola, D & Livingston, R (1999) Possible interactions between dietary antioxidants and chemotherapy. Oncology (Williston Park) 13, 10031008.Google ScholarPubMed
67. Lamson, DW & Brignall, MS (1999) Antioxidants in cancer therapy; their actions and interactions with oncologic therapies. Altern Med Rev 4, 304329.Google ScholarPubMed
68. Velentzis, LS, Keshtgar, MR, Woodside, JV, et al. (2011) Significant changes in dietary intake and supplement use after breast cancer diagnosis in a UK multicentre study. Breast Cancer Res Treat 128, 473482.CrossRefGoogle Scholar
69. Izano, MA, Fung, TT, Chiuve, SS, et al. (2013) Are diet quality scores after breast cancer diagnosis associated with improved breast cancer survival? Nutr Cancer 65, 820826.CrossRefGoogle ScholarPubMed
70. Rohan, TE, Hiller, JE & McMichael, AJ (1993) Dietary factors and survival from breast cancer. Nutr Cancer 20, 167177.CrossRefGoogle ScholarPubMed
71. Russo, J & Russo, IH (1994) Toward a physiological approach to breast cancer prevention. Cancer Epidemiol Biomarkers Prev 3, 353364.Google Scholar
72. Harris, HR, Orsini, N & Wolk, A (2014) Vitamin C and survival among women with breast cancer: a meta-analysis. Eur J Cancer 50, 12231231.CrossRefGoogle ScholarPubMed
73. Rock, CL, Flatt, SW, Natarajan, L, et al. (2005) Plasma carotenoids and recurrence-free survival in women with a history of breast cancer. J Clin Oncol 23, 66316638.CrossRefGoogle ScholarPubMed
74. Tsubura, A, Uehara, N, Kiyozuka, Y, et al. (2005) Dietary factors modifying breast cancer risk and relation to time of intake. J Mammary Gland Biol Neoplasia 10, 87100.CrossRefGoogle Scholar
75. Kim, SH, Sehrawat, A & Singh, SV (2013) Dietary chemopreventative benzyl isothiocyanate inhibits breast cancer stem cells in vitro and in vivo . Cancer Prev Res (Phila) 6, 782790.CrossRefGoogle ScholarPubMed
76. Koh, MS, Hwang, JS & Moon, A (2010) Lycopene inhibits proliferation, invasion and migration of human breast cancer cells. Biomol Ther (Seoul) 18, 9298.CrossRefGoogle Scholar
77. Xu, M, Bower, KA, Wang, S, et al. (2010) Cyanidin-3-glucoside inhibits ethanol-induced invasion of breast cancer cells overexpressing ErbB2. Mol Cancer 9, 114.CrossRefGoogle ScholarPubMed
Figure 0

Fig. 1 Flow chart of the selection of studies included in the meta-analyses. WOS, Web of Science; HR, hazard ratio; RR, relative risk.

Figure 1

Table 1 Characteristics of included studies and quality score

Figure 2

Fig. 2 Forest plots of observational studies investigating the association of all-cause mortality with (a) fruit and vegetable intake, (b) total vegetable intake, (c) cruciferous vegetable intake and (d) fruit intake. Highest v. lowest intake.

Figure 3

Fig. 3 Forest plots of observational studies investigating the association of breast cancer-specific mortality with (a) total vegetable intake and (b) fruit intake. Highest v. lowest intake.

Figure 4

Fig. 4 Forest plots of observational studies investigating the association of breast cancer recurrence with (a) total vegetable intake and (b) cruciferous vegetable intake. Highest v. lowest intake.

Figure 5

Table 2 Summary of subgroup analyses of the effects of fruit and vegetable intake on all-cause mortality and breast cancer recurrence (Hazard ratios (HR) and 95 % confidence intervals)