Dietary Protein Sources and Incidence of Breast Cancer: A Dose-Response Meta-Analysis of Prospective Studies

Protein is important to the human body, and different sources of protein may have different effects on the risk of breast cancer. Thus, we conducted a meta-analysis to investigate the association between different dietary protein sources and breast cancer risk. PubMed and several databases were searched until December 2015. Relevant articles were retrieved according to specific searching criteria. Forty-six prospective studies were included. The summary relative risk (RR) for highest versus lowest intake was 1.07 (95% confidence interval (CI) 1.01–1.14, I2 = 34.6%) for processed meat, 0.92 (95% CI 0.84–1.00, I2 = 0%) for soy food, 0.93 (95% CI 0.85–1.00, I2 = 40.1%) for skim milk, and 0.90 (95% CI 0.82–1.00, I2 = 0%) for yogurt. Similar conclusions were obtained in dose-response association for each serving increase: total red meat (RR: 1.07; 95% CI 1.01–1.14, I2 = 7.1%), fresh red meat (RR: 1.13; 95% CI 1.01–1.26, I2 = 56.4%), processed meat (RR: 1.09; 95% CI 1.02–1.17, I2 = 11.8%), soy food (RR: 0.91; 95% CI 0.84–1.00, I2 = 0%), and skim milk (RR: 0.96; 95% CI 0.92–1.00, I2 = 11.9%). There was a null association between poultry, fish, egg, nuts, total milk, and whole milk intake and breast cancer risk. Higher total red meat, fresh red meat, and processed meat intake may be risk factors for breast cancer, whereas higher soy food and skim milk intake may reduce the risk of breast cancer.


Introduction
Protein is important to the human body. Protein is involved in the constitution of human tissues and the regulation of various physiological functions. It is essential for body growth and development, as well as the transport of many important substances, and the provision of biological energy. Protein deficiency can induce many types of diseases, such as retarded growth and development, fatigue, and nutritional edema, and it can even be life threatening [1,2].
No food contains exclusively protein; we must always intake different sources of protein from a variety of compounds. Different sources of protein might have different effects on breast cancer risk. For example, carcinogenic byproducts-such as heterocyclic amines and polycyclic aromatic hydrocarbons formed during the high-temperature cooking of red meat [3,4]-can increase the risk of breast cancer. However, evidence from prospective cohort studies remains controversial [5][6][7][8]. Thus, in this study, we wanted to investigate the association between the consumption of different sources of dietary protein and the risk of breast cancer. We also intend for this article to serve as a reference for Table 1. Search Strategies for meta-analysis.

Topic
Search Strategy Breast Cancer "Breast Neoplasms" OR "Breast Cancer" Study "Cohort" OR "Nested Case-control" OR "Case-cohort" OR "Prospective"

Poultry and Risk of Breast Cancer
Highest versus lowest category analysis. Eleven cohort studies [5][6][7][8]17,21,24,34,41,50,55] were included in the analysis, including 19,400 cases among 726,947 participants. The summary RR for Dose-response analysis of total red meat, fresh red meat, processed meat, soy foods, and skim milk intake and the risk of breast cancer. The solid line represents estimated relative risks (RRs), and dashed lines are their 95% confidence interval (CI). As shown in the figure, total red meat, fresh red meat, and processed red meat are related to an increased risk of breast cancer, whereas soy food and skim milk are related to a decreased risk of breast cancer.

Nuts and Risk of Breast Cancer
Highest versus lowest category analysis. Three cohort studies [7,8,47] investigated the association between nut intake and breast cancer risk, including 4506 cases among 148,807 participants. The summary RR for highest versus lowest was 0.96 (95% CI 0.88-1.06), with no heterogeneity, I 2 = 0.0% (See Table 2). No publication bias was observed by Begg's test (p = 1.000) or Egger's test (p = 0.461).
Dose-response analysis. Three cohort studies [7,8,47] were included in the dose-response analysis. The summary RR per serving/day was 0.96 (95% CI 0.84-1.09) with no heterogeneity, I 2 = 0.0% (See Table 2). No publication bias was observed by Begg's test (p = 0.100) or Egger's test (p = 0.955). We did not observe a linear association between nut intake and risk of breast cancer.

Skim Milk and Risk of Breast Cancer
Highest versus lowest category analysis. Eight studies from seven cohorts [31,32,41,50,52,53,55] reported an association between skim milk intake and risk of breast cancer, including 16,664 cases and 586,726 participants. The summary RR for highest versus lowest was 0.93 (95% CI 0.85-1.00) with moderate heterogeneity, I 2 = 40.1% (see Figure 4C and Table 2). No publication bias was observed by Begg's test (p = 0.266) or Egger's test (p = 0.616).

Yogurt and Risk of Breast Cancer
Highest versus lowest category analysis. Seven studies from five cohorts [31,44,51,53,55] explored an association between yogurt consumption and breast cancer risk, with 6793 cases among 225,057 participants. The summary RR for highest versus lowest was 0.90 (95% CI 0.82-1.00) with no heterogeneity, I 2 = 0.0% (see Figure 4D and Table 2). No publication bias was observed by Begg's test (p = 0.764) or Egger's test (p = 0.77).
Dose-response analysis. Only three cohorts [44,53,55] were eligible for the dose-response analysis. The summary RR per 200 g/day was 0.87 (95% CI 0.72-1.06) with no heterogeneity, I 2 = 0.0% (See Table 2). No publication bias was observed by Begg's test (p = 1.000) or Egger's test (p = 0.488). We did not observe a linear association between yogurt intake and risk of breast cancer.

Subgroup and Sensitivity Analysis
Subgroup analysis and meta-regression were conducted for both highest versus lowest analysis (Supplementary Materials Table S3) and dose-response analysis (Table 3). For most sources of dietary protein exposure, meta-regression and subgroup analysis did not exhibit any substantial change in the summary relative risk. For processed meat and skim milk, the results of subgroup analysis were consistent with summary analysis; however, there was a weak association in some strata (p < 0.1). For soy food and total red meat, we did not conduct the subgroup analysis, due to no evidence of heterogeneity in the summary analysis (I soyfood 2 = 0.0%, I total red meat 2 = 7.1%). Given that only three cohort studies investigated the association between nut and yogurt intake and risk of breast cancer, we also did not conduct the meta-regression and subgroup analysis for nuts and yogurt.
In sensitivity analysis, we removed one study at a time and calculated overall homogeneity and effect size. We confirmed that the summary results were reliable for most sources of dietary protein (data not shown).

Discussion
This meta-analysis suggested that higher intake of soy food and skim milk could decrease the risk of breast cancer, and that higher intake of processed meat may increase the incidence of breast cancer. Dose-response analysis revealed the summary risk of breast cancer decreased by 9% for soy food, decreased by 4% for skim milk, and increased by 9% for processed meat. For total red meat and fresh red meat, the dose-response analysis suggests that increased intake of these foods could increase the risk of breast cancer, but the results of summary RR for highest versus lowest intake failed to reveal an increased risk. For yogurt, we found it could decrease risk of breast cancer, and the summary RR for highest versus lowest intake was 0.90 (95% CI 0.82-1.00). However, the result was not consistent with the result from dose-response analysis. Only three cohorts were eligible for the dose-response analysis for yogurt. More studies are needed to investigate the association between yogurt intake and risk of breast cancer. We also found a null association between poultry, fish, egg, nut, total milk, and whole milk consumption and risk of breast cancer.
Meat is an important source of protein. NHANES data indicated that approximately 58% of the meat consumed was red meat and that 22% of the total meat consumed was processed meat in the USA [61]. Evidence from prospective studies of red and processed meat consumption and risk of breast cancer was inconsistent [7,8,53,55,57], and the mechanisms involved require further investigation. Several mechanisms were important in explaining the association between red and processed meat intake and risk of breast cancer: (1) carcinogenic byproducts-such as heterocyclic amines and polycyclic aromatic hydrocarbons-formed during high temperature cooking of red meat [4,62]; (2) fat, heme iron, and the animal sugar molecule N-glycolyneuraminic acid were enriched in red meat, which could promote inflammation, oxidative stress, and tumor formation [63][64][65]; (3) in a few countries, hormone residues of the exogenous hormones used to stimulate the growth of beef cattle has also been suggested as a risk factor of breast cancer [66].
Soy food is an essential part of many people's diets in Asian countries. An appropriate method of calculating the amount of protein in soy food is not available, and only one study listed soy protein as an observational item. Thus, we did not extract soy protein from the original data. Soy food contains numerous fibers and phytoestrogens, which can arrest cell cycle, induce apoptosis, and inhibit angiogenesis. These mechanisms might support the notion that soy food intake was negatively associated with breast cancer incidence.
Milk is another important source of protein. Many prospective cohort studies have suggested a null association between total milk consumption and risk of breast cancer [50,51,53,55]. A recent meta-analysis of prospective studies regarding milk consumption and breast cancer incidence found that the intake of total milk was not independently associated with increasing breast cancer risk [64]. They found skim milk could decrease the risk of breast cancer; however, only four studies of skim milk were included in the meta-analysis. Milk contains fat, calcium, vitamin D, conjugated linoleic acids (CLAs), etc. In vitro studies have suggested that calcium, vitamin D, and CLA exert anticarcinogenic effects, such as inhibition of cell cycle progression, induction of apoptosis, inhibition of angiogenesis, and differentiation of mammary cells [67][68][69][70]. However, dietary fat has been thought to be a risk factor of breast cancer [71][72][73]. However, the mechanisms of dietary fat's influence on cancer risk were speculative, and we need more studies to prove it.
There were few studies that explored the association between yogurt intake and the incidence of breast cancer [31,44,51,53,55]. Most of them hold the idea that yogurt consumption did not impact the risk of breast cancer. However, a study of estrogen-and progesterone receptor-positive (ER+PR+) breast cancer revealed a protective linear risk trend with yogurt (HR = 0.89, 95% CI = 0.80-0.99) [53]. More studies are needed to investigate the association between yogurt intake and risk of breast cancer.
To our knowledge, this is the first meta-analysis to systematically and comprehensively explore the association between different dietary sources of protein and the risk of breast cancer. This is the first meta-analysis to explore the association between poultry, nuts, whole milk, skim milk, and yogurt consumption and breast cancer risk. Additionally, this is an updated meta-analysis to investigate the association between total red meat, fresh red meat, processed meat, fish, egg, soy food, and total milk intake and incidence of breast cancer. Although some previous meta-analyses [15,[74][75][76][77] have investigated the association between total red meat, fresh red meat, processed meat, fish, egg, soy food, and total milk intake and incidence of breast cancer, many of them were based on highest versus lowest intake. Since then, a number of large-scale prospective studies were performed. Our study was based on dose-response meta-analysis, and forty-six independent prospective cohort studies were collected. In addition, we also explored the sources of study heterogeneity and the influence of potential residual confounding factors through fifteen subgroup analysis and meta-regression. Recently, a meta-analysis of red meat consumption and breast cancer found that red meat intake was associated with increasing breast cancer risk, but the term "red meat" referred to total red meat in some studies, and fresh red meat in other studies [78]. In our study, we separately explored the association between total red meat (fresh red meat and processed meat), fresh red meat, and processed meat intake and risk of breast cancer.
Our meta-analysis had several strengths. Firstly, the assessment was based on prospective studies, which avoided the influence of recall and selection bias. Secondly, the large sample size (60,615 breast cancer cases and 2,749,307 participants) allowed us to quantitatively assess the association of different sources of protein consumption and breast cancer incidence. Thirdly, our results were based on dose-response meta-analysis, which would avoid some influence of potential residual confounding factors. Finally, we explored the sources of study heterogeneity and the influence of potential residual confounding factors through fifteen subgroup analyses and meta-regression.
However, this meta-analysis also had several limitations. First, the inherent problems of unmeasured or residual confounders in the included studies might confound the association. Although we included fifteen subgroup analyses and meta-regression to explore and balance the influence of potential residual confounding factors (Table 3), numerous unmeasured or residual confounders-such as ER, PR, and Her-2 status of the tumor-might confound the association. Only a few studies [7,8,17,47,48,53,55,59] explored the influence of ER, PR, and Her-2 status of the tumor on the association. In addition, most of the studies used a single Food Frequency Questionnaire and assumed the diet did not change over many years of follow-up. Thus, we compared the studies with different follow-up periods, and the results were consistent (Table 3). However, more high quality studies are needed to investigate the association between yogurt intake and breast cancer risk. Third, different studies used different units (such as servings and times). Although we converted these units into grams per day according to standard conversions from the Food Standards Agency [13] and other documents [14][15][16], this could still influence the association.

Conclusions
In conclusion, our results suggested that diets rich in soy food and skim milk were associated with decreased breast cancer risk. High intake of red and/or processed meat was associated with an increased risk of breast cancer. A null association was noted between poultry, fish, egg, nut, total milk, and whole milk consumption and risk of breast cancer. However, additional well-designed cohort or interventional studies and studies exploring the mechanisms in humans are needed to confirm the association.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/8/11/730/s1, Table S1: Characteristics of included studies. Table S2: Quality assessment of included studies. Table S3: Subgroup analysis of dietary protein sources intake and risk of breast cancer, highest versus lowest intake.