Effect of the Intake of Isoflavones on Risk Factors of Breast Cancer—A Systematic Review of Randomized Controlled Intervention Studies

Epidemiological studies suggest that high intake of soy isoflavones may protect against breast cancer, but causal relationships can only be established by experimental trials. Thus, we aimed to provide a systematic review of randomized controlled trials (RCTs) on the effect of an isoflavone intake on risk factors of breast cancer in healthy subjects. After a systematic literature search in PubMed, 18 different RCTs with pre- and/or postmenopausal women were included and investigated for details according to the PRISMA guideline. In these studies, isoflavones were provided by soy food or supplements in amounts between 36.5–235 mg/d for a period of 1–36 months. Breast density, estrogens including precursors, metabolites, estrogen response such as length of menstrual cycle, and markers of proliferation and inflammation were considered. However, in most studies, differences were not detectable between isoflavone and control/placebo treatment despite a good adherence to isoflavone treatment, irrespective of the kind of intervention, the dose of isoflavones used, and the duration of isoflavone treatment. However, the lack of significant changes in most studies does not prove the lack of effects as a sample size calculation was often missing. Taking into account the risk of bias and methodological limitations, there is little evidence that isoflavone treatment modulates risk factors of breast cancer in pre- and postmenopausal women. Future studies should calculate the sample size to detect possible effects and consider methodological details to improve the study quality.


Introduction
Breast cancer is the most frequent type of cancer in women globally. Each year, about 2.3 million women worldwide develop breast cancer, and about 685,000 people were estimated to die from breast cancer in 2020 [1]. Women from Europe and North America are particularly concerned as the age-adjusted incidence rate of breast cancer is approximately 2-4 times higher than in Asia [2].
Breast cancer is favored by race, ethnicity, family history of cancer, genetic variants and mutations of genes modulating DNA repair. Age at menarche, parity, as well as age at first pregnancy affect the risk by modulating the long-term sex hormone levels. Physical inactivity is a modifiable risk factor as well as diet. Whereas certain foods and food ingredients increase the risk of breast cancer (e.g., alcohol), others like soy or isoflavones seems to be protective [3].
More than 25 years ago, a population-based case-control study revealed a higher breast cancer rate in women of Chinese, Japanese, and Filipino ethnicities migrating to USA and Hawaii than in the countries of origin, which approximated in Asian-Americans born in the West to the U.S. white rate. However, the risk increased among Asian immigrants in the U.S. over several generations, suggesting that the Western lifestyle substantially increased the breast cancer risk [4].
In contrast to the Western diet, the Asian diet is traditionally rich in foods produced from soybeans as main ingredients. As soy products are the major dietary source of isoflavones, the intake of isoflavones in Asia (China: 6.2-75.7 mg/d; Japan: 22.6-54.3 mg/d) is much higher than in Europe (0.37-4.5 mg/d) and in USA (0.73-3.3 mg/d) despite considerable variation between individual studies [5]. A meta-analysis mainly derived from case-control studies has shown that a high (≥20 mg/d) and moderate isoflavone intake (~10 mg/d) by consumption of soy food reduces the risk of breast cancer in Asia and Asian American populations by 29% and 12%, respectively, compared to a low isoflavone intake (≤5 mg/d). This effect was dose-dependent (risk reduction about 16% per 10 mg of isoflavones intake per day). Moreover, it could be observed in both pre-and postmenopausal women. However, in Western populations, high vs. low isoflavone intake (≥0.8 mg/d vs. ≤0.15 mg/d) did not affect the breast cancer risk [6]. In contrast, a recently published meta-analysis of 16 prospective cohort studies (six from Asia, 10 from Western countries) did not find an association between high and moderate vs. low isoflavone intake and the risk of breast cancer. However, if consumption of soy food was considered, high vs. low intake of soy foods was associated with a 13% lower risk to develop breast cancer. In addition, moderate consumption of soy food was also associated with a 25%-28% reduced breast cancer risk if the duration of the follow-up lasted ≥10 years, and if the study was adjusted for smoking status, alcohol intake, and for hormone replacement therapy [7].
Isoflavones such as genistein and daidzein may protect against breast cancer through certain mechanisms. First, isoflavones may affect the hormone levels in breast and ovaries by modulating the activity of steroidogenic enzymes (e.g., aromatase, 3-and 17β-hydroxysteroid dehydrogenase), thereby reducing the conversion of estrogen precursors (androgens) to estrogens and the dehydrogenation of estrone (E 1 ) to estradiol (E 2 ). Second, isoflavones may alter estrogen metabolism away from cancerous metabolites (16-α-hydroxyl metabolites) towards 2-hydroxy estrogen metabolites with lower estrogen activity. Third, due to structural similarities to human 17β-estradiol, isoflavones can bind to estrogen receptors (ER), preferentially to ER-β, which suppresses the transcription of many genes involved in cell growth and inflammation, thereby diminishing the estrogenic effects induced by ERα [8,9].
As reviewed earlier, isoflavones may also prevent against breast cancer by other mechanisms. Binding of phytoestrogens to ER at the surface of cells might directly modulate the expression of genes by the inhibition of signaling pathways like Akt and MAPK stimulating cell growth, proliferation, and survival, while activating proapoptotic genes like Bcl-2, p53, caspase-3, Bax, BRCA-1, and BRCA-2 [10]. By this mechanism, isoflavones may stimulate the synthesis of sex hormone binding globulin (SHBG), thereby reducing the free (active) E 2 in plasma [11].
Moreover, isoflavones exert antioxidant properties [12,13]. Since reactive oxygen species from exogenous and endogenous sources (estrogens) favor oxidative stress which in turn can stimulate inflammatory and proliferative pathways involved in the pathogenesis of breast cancer [14], the antioxidant and anti-inflammatory properties of isoflavones might also contribute to protection from breast cancer.
Hence, epidemiological studies suggest that high isoflavone intake may protect against breast cancer. This is supported by in vitro studies on biological mechanisms to explain how isoflavones can modulate steps involved in the development of breast cancer. However, the effectiveness of isoflavone treatment for the prevention of breast cancer can only be investigated by intervention studies with a randomized controlled design which allow cause-effect relationships between intervention and outcome [15]. Randomization balances patients' characteristics between the groups or treatments and enables attribution of any differences in outcome to intervention [16]. As breast cancer develops over years, the response to isoflavone intervention can only be evaluated by surrogate endpoint mark-Nutrients 2021, 13, 2309 3 of 32 ers, i.e., by risk factors for breast cancer that can easily be determined by non-invasive methods [17].
Breast density is an independent risk factor for breast cancer [18]. Risk factors in serum include free E 2 , insulin-like growth factor 1 (IGF-1), the ratio of IGF-1 to IGF binding protein 3 (IGFBP-3), single nucleotide polymorphisms, and breast intra-epithelial neoplasia. An increased breast density is associated with changes in the microenvironment (increased secretion of inflammatory molecules, cytokines, growth factors) that favor tumor growth [19], and with a reduced renal excretion of estrogens [20]. Estrogens stimulate the expression of genes involved in cell growth and inflammation [21]. Ki-67 antigen, expressed by proliferating cells, is an established molecular marker for breast cancer [22]. Nipple aspiration fluid (NAF) cytology is considered as predictor for breast cancer [23], with a high diagnostic specificity (0.97), but a low sensitivity (0.64) which limits diagnostic accuracy [24]. As hormonal stimulation of the breast tissue plays a considerable role in breast carcinogenesis and for the length of the menstrual cycle (MC), the latter might be relevant for breast cancer risk [25].
Hence, the aim of this systematic review was to investigate whether the intake of isoflavones by healthy subjects in RCTs may protect against breast cancer by consideration of breast density (main endpoint). In addition, further parameters associated with the risk of breast cancer (estrogens, growth factors, markers on inflammation, proliferation and apoptosis, NAF cytology, length of MC) were assessed. Moreover, the risk of bias (RoB) and the imprecision of the included studies was judged to evaluate the evidence for preventive effects.

Materials and Methods
This systematic review was conducted according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [26].

Literature Search Strategy
A systematic literature search was performed in PubMed for human intervention studies that investigated the effects of isoflavone intake on risk factors of breast cancer. For the literature search, the following combinations of keywords were used: "breast cancer soy", "breast cancer isoflavones prevention", "breast cancer phytoestrogen prevention", "breast cancer red clover", "mammary carcinoma soy", "mammary carcinoma isoflavones prevention", "mammary carcinoma phytoestrogen prevention", "mammary carcinoma red clover", "breast cancer isoflavones primary prevention", "breast cancer phytoestrogen primary prevention", "breast cancer phytoestrogen primary prevent *". Filters were applied with regard to study type (RCTs), language (German, English), and species (humans). The database search was performed by two reviewers (L.F., E.S.) up to 31 July 2020 for relevant studies that were published as original contributions or short communications. Additional databases beyond PubMed were not used for literature search as for other clinical topics, the search in PubMed has shown a higher specificity than Google Scholar, and a comparable sensitivity, suggesting that PubMed is an optimal tool for biomedical research [27]. Paid databases such as Scopus and Web of Science cover further areas of research (e.g., health, life and social sciences, technology) and have shown to provide additional records, but mostly without relevance for biomedical questions [28].

Inclusion and Exclusion Criteria
Studies were included if (1) they investigated the effect of an isoflavone intake on risk factors of breast cancer, e.g., breast density, estrogen metabolites, length of MC, markers of proliferation, volume and cytology of NAF, tyrosine kinase activity and expression of genes related to proliferation, apoptosis, and estrogenic effects. Further inclusion criteria were (2) a randomized controlled study design, (3) isoflavone treatment by consumption of soy foods or supplements, (4) a control treatment with restrictions on soy consumption or isoflavone intake or an isoflavone-free placebo if isoflavones were provided as supple-Nutrients 2021, 13, 2309 4 of 32 ments (e.g., capsules, tablets, soy protein powder). Exclusion criteria comprised (1) the investigation of subjects already suffering from breast cancer, (2) treatment of the subjects with oral contraceptives, and (3) in placebo-controlled studies, the use of a low-isoflavone intake instead of an isoflavone-free placebo treatment.

Study Selection, Data Extraction, and Assessment of Study Quality
Two independent reviewers (L.F., E.S.) identified relevant studies according to the predefined eligibility criteria. All records were checked for duplicates. Duplicates were removed and the remaining records were screened by title and/or abstract to exclude records that did not meet the inclusion criteria. For the remaining records, the full-text articles of potentially relevant studies were checked for eligibility, based on the abovementioned criteria. Any discrepancies in the study selection process were discussed between both reviewers (L.F., E.S.) and if necessary with S.E., until a consensus was reached. Finally, eligible trials were included in this review.
Both review authors (L.F., E.S.) extracted relevant data from the included studies independently by using a self-made Excel template: study design, details on intervention (kind/amount of isoflavones, application form) and on control or placebo treatment, participants (sample size, demographic data, criteria of eligibility), country in which the study was performed, and risk factors of breast cancer. Moreover, parameters on the bioavailability of isoflavones and on the compliance with intervention (e.g., urinary isoflavone excretion) were considered. Moreover, studies were checked for sample size calculation (prospectively performed, biomarker chosen, sufficient subjects available for statistical evaluation) and checked for industry funding. Discrepancies in data extraction were discussed between both reviewers and, if necessary, with S.E.
Afterwards, the RoB of included RCTs was independently assessed by two authors (L.F., S.E.) using the Cochrane risk of bias tool [29] considering the following criteria: (1) generation of randomization list before the study and using an adequate randomization method (selection bias), (2) allocation concealment from participants and investigators, (3) blinding of participants and investigators (performance bias), (4) blinding of outcome assessment until completion of statistical evaluation (detection bias), (5) completeness of outcome data, reporting number and reasons of dropout for each group/treatment, application of intention-to-treat analysis or statistical models to consider missing values (attrition bias), (6) registration of the study protocol, reporting full endpoints and outcomes according to registration (reporting bias), (7) considering potential confounder from diet (other risk of bias) by investigation of food consumption, restrictions on soy or isoflavone intake, and by considering the compliance with intervention. The latter was investigated by urinary isoflavone excretion, by a diary where the intake of foods or supplements were documented, or as ratio of ingested capsules to the number of capsules that should have been ingested during intervention. The overall RoB was assessed within and across the studies by using the Cochrane risk of bias tool [29]. Each publication was checked for the registration number of the study protocol and each study for registration at clinicaltrials.gov. Again, a self-made Excel template was used to check these criteria for each study. Discrepancies in RoB assessment occurred if details on randomization, blinding of investigators and allocation concealment remained unclear. A closer look on the study design and/or on the results could be helpful. If the procedure to consider missing values was not clearly described, the statistical analysis was illuminated in detail. Discrepancies were also resolved through discussion.

Study Selection and Study Characteristics
After a systematic literature search in PubMed, 162 records were identified. After removing duplicates, 78 records remained and were screened by title and/or by abstract. In total, 36 records were excluded after screening as being not relevant for the question addressed by this review. The remaining 42 records were assessed for eligibility by the full text. Thirteen intervention studies were excluded as they were not randomized (n = 2), had no adequate control/placebo treatment (n = 3) or did not address the question of the review (n = 8). Finally, 29 records were considered to be eligible .
However, with regard to country, participants, study design, intervention and study protocol registration, it becomes obvious that the results of Maskarinec [48] also provided results which were partly obtained from study B and study C, respectively. Since 14 publications of Maskarinec and co-workers [34][35][36][37][38][40][41][42][43][44][45][46][47][48] were obtained from three different trials, the 29 records included in the present review described the results of 18 different RCTs. A flow diagram of the identification and selection of the studies is shown in Figure 1.

Studies with Premenopausal Women
Eighteen RCTs were conducted with premenopausal women. Details are shown in Table 1.
Nagata et al. [32] investigated the effect of soymilk consumption (400 mL/d; 109 mg/d isoflavones) for 3 MC in addition to a regular diet. The study was done in parallel group design in Japan. Soymilk consumption decreased E 1 in serum after 3 MC compared to baseline and compared to control treatment. The changes in E 2 , SHBG and in the length of MC were not different between both treatments.
Zittermann et al. [39] provided soy-containing cookies (52 mg isoflavones/d) and soy-free cookies (placebo), respectively, daily for 1 MC in a crossover study. The urinary excretion of genistein and daidzein increased, but this was not accompanied by different concentrations of E 1 , E 2 (total, free), follicle-stimulating hormone (FSH), SHBG, and progesterone in serum between both treatments.
Duncan et al. [31] investigated whether supplementation of isoflavones in doses of 2.0 mg/kg BW/d and 1.0 mg/kg BW/d by soy protein induces hormonal changes compared to a protein powder providing only traces of isoflavones (0.15 mg/kg BW/d; control). In this crossover study, each intervention was done for 3 MC plus 9 days, with a 3-week-washout between two interventions. In the midfollicular phase, E 1 was lower after an isoflavone intake of 2.0 mg/kg BW/d vs. 1.0 mg/kg BW/d. No differences could be observed in E 1 in early follicular, periovulatory, and midluteal phase. In the periovulatory phase, luteinizing hormone (LH) and FSH were lower after medium isoflavone intake vs. control. In other phases, no differences could be observed. Estrone sulfate (E 1 S) and progesterone were not modulated by any treatment in any phase of the MC. Dehydroepiandrosterone sulfate (DHEAS) reached higher concentrations in plasma after high vs. medium isoflavone intake, but testosterone, androstenedione, dehydroepiandrosterone (DHEA), SHBG, and prolactin and the length of MC, follicular and luteal phase in MC 2 and 3 were comparable between all interventions.
Brown et al. [30] provided 40 mg/d isoflavones with soy protein in addition to a Western diet and used a Western diet free from soy protein as control. In this crossover study, each intervention was done for 2 MC. Changes in serum E 1 , E 2 , E 1 S, progesterone, testosterone, androstenedione, DHEA, DHEAS, SHBG, prolactin (mid-follicular and midluteal phase), FSH (mid-follicular phase), and LH (mid-luteal phase) were not detectable. The excretion of hormonal metabolites such as 2-(OH)E 1 and 16α-(OH)E 1 in 48-h-urine and the 2-(OH)E 1 -to-16α-(OH)E 1 -ratio was also comparable and the length of MC not different between all treatments.
Kumar et al. [33] provided 40 mg/d isoflavones by soy protein or an isoflavonefree milk protein (placebo) for 3 MC. Changes were not different between both treatments with regard to E 1 , free and total E 2 , SHBG, and with regard to the length of each MC and each follicular phase. Nevertheless, the length for 3 MC was extended by isoflavone vs. placebo treatment.
The first RCT of Maskarinec et al. (study A), a double-blind, placebo-controlled trial with parallel group design, provided either 100 mg/d isoflavones by tablets or a placebo (maltodextrine) for 12 months to premenopausal women. Supplementation of isoflavones increased their urinary excretion already after 1 month vs. placebo and remained increased up to the end the study [34,35]. Nevertheless, the serum concentration of E 1 , E 1 S, E 2 , SHBG, FSH, LH, and progesterone remained unchanged. Estrogen metabolites such as 16α-(OH)E 1 , 2-(OH)E 1 , the 2-(OH)E 1 -to-16α-(OH)E 1 -ratio [34,35] and estrone-3-glucuronide [34] were not different after both treatments 1, 3, 6, and 12 months of intervention. Differences in the length of MC [35] and in mammographic parameters [37] were not detectable.
The third RCT of the same working group (study C), a crossover study, provided two servings of soy foods daily for 6 months in addition to a low soy diet. The latter served as control. Both interventions were separated by a 1-month-washout [42][43][44][45][46][47][48]. Again, urinary excretion of isoflavones increased by the consumption of soy foods [43,47,48], but without changes in the estrogens' concentration in serum [42], NAF [42], and urine [45]. Urinary excretion of most estrogenic metabolites was not different between both treatments except of E 1 S [48], 4-(OH)E 1 [45], and the 2-(OH)E 1 -to-16α-(OH)E 1 -ratio [45]. NAF volume [43,48] and the cytological classification of mammary epithelial cells from NAF [47] were comparable. The 15-F 2 -isoprostanes-to-creatinine-ratio in urine increased by consumption of soy foods vs. control, also with consideration of the compliance, but significance was failed after excluding participants with very low creatinine levels [46].
If statistical analysis separated between pre-and postmenopausal women, the same results were obtained as for all women with regard to the changes in the plasma concentration of genistein, equol, E 2 , FSH, progesterone, SHBG, E 2 /SHBG, and the expression of the above mentioned genes from mammary epithelial cells [57].

As shown in
Dropouts/missing data √ √ √ √ √ √ Dropouts/missing data reported 3 √ √ √ √ × √ Reasons for dropouts/missing data reported 4 × × × √ × √ Intention-to-treat analysis 5 × √ √ √ × − ×: No, not considered; √: yes, considered; ?: not clear, no details available, − irrelevant. 1 Investigated but results not given; 2 due to inclusion of subjects whose premenopausal status was not confirmed by hormone analysis; 3 separately for each group and dropout comparable between groups; 4 separately for each group or treatment and being comparable between groups; 5 missing data were imputed by appropriate statistical models.   Figure 6 shows the RoB assessment for individual studies with mixed groups of women on the basis of the criteria presented in Table 6. Three out of four RCTs used an adequate randomization method, thereby reducing the risk of selection bias [55,57,58]. For most studies, the risk of allocation concealment remains unclear [55][56][57]. Each study provided isoflavones by means of tablets and used an isoflavone-free placebo. Due to blind-  Figure 6 shows the RoB assessment for individual studies with mixed groups of women on the basis of the criteria presented in Table 6. Three out of four RCTs used an adequate randomization method, thereby reducing the risk of selection bias [55,57,58]. For most studies, the risk of allocation concealment remains unclear [55][56][57]. Each study provided isoflavones by means of tablets and used an isoflavone-free placebo. Due to blinding of participants, researcher and outcomes, the risks of performance bias and detection bias were low for all trials [55][56][57][58]. The risk of attrition bias was low [56], unclear [55], or high [57,58] as dropouts and underlying reasons were not always reported (if reported, not always separately for each group) and statistical methods to impute missing data were only applied in one study. The risk of other bias was different (low [56,57], unclear [55], high [58]) as confounders from diet and adherence to treatment were only partly considered. Within studies, overall RoB was unclear [55,56] or high [57,58]. RoB across RCTs was low for selection bias, performance bias and detection bias, high for attrition bias, and unclear for allocation concealment, reporting bias and other bias (Figure 7). Table 6. Criteria to assess the risk of bias in studies with a mixed group of pre-, peri-and postmenopausal women.

Discussion
The aim of this systematic review was to provide an overview on RCTs which investigated the effect of isoflavone intake on risk factors of breast cancer to evaluate the evidence for preventive effects in vivo taking into account the RoB of the studies considered. To the best of our knowledge, this is the first systematic review that provides a detailed picture on potential changes with regard to breast density, estrogen synthesis, estrogen metabolism and biological mechanisms that depend on estrogen response. For this, a variety of mammographic, functional, and laboratory parameters were considered, established risk factors as well as factors being associated with breast cancer risk.
In women with different menopausal states, changes in estrogen homeostasis by isoflavone treatment are also unlikely as the serum/plasma concentration of E 2 [55,57], progesterone [57], FSH [55,57,58], LH [55], SHBG [57], and SHBG/E 2 [57] remained unchanged. Parameters influenced by estrogens (breast density [55,58], growth factors [56], cytological classification of mammary epithelial cells [57], components of NAF [57]) were not different between both treatments although 10-times higher concentrations of genistein in NAF were achieved by isoflavone vs. placebo treatment [57]. Differences in the expression of genes related to proliferation, apoptosis and other estrogenic effects were not detectable between isoflavone and placebo group [57]. Therefore, isoflavone supplementation even in a large dose of 235 mg/d for 6 months does not modulate the expression of genes involved in the regulation of proliferation, apoptosis, and inflammation. If data were analyzed separately for pre-and postmenopausal women, differences between the subgroups were not detectable either [56,57]. It is well known that both, the estrogen concentration in serum/plasma and the expression of ER-β, are reduced in post-compared to premenopausal women. This in turn enhances proliferative and inflammatory response, thereby increasing the risk of breast cancer [62]. With regard to the mechanisms of isoflavones on estrogen synthesis, metabolism and estrogen response, effects by isoflavone intake were especially expected in postmenopausal women but were not found.
The sample size was calculated in four trials [53,55,57,58], but in two RCTs for vasomotor symptoms [53] and bone density [58] being not relevant for this review. The other used the Wolfe pattern [55] and Ki-67 LI [57] for sample size calculation. The number of cases included in the statistical evaluation of both trials was above the calculated sample size.
Moreover, pooling data as in meta-analyses increases the sample size and the probability to detect an effect by isoflavone treatment. A meta-analysis of eight RCTs published in 2010 investigated the impact of isoflavone-rich foods or supplements on breast density and related parameters. A small increase in breast density was detectable for premenopausal women, but not for postmenopausal women and all women.
This systematic review of RCTs investigates the response of isoflavone treatment to parameters which are considered as risk factor for breast cancer. These RCTs were described in detail and assessed for RoB to provide a clear picture on the effect induced by isoflavone intake in women with different menopausal states. Moreover, each study was checked for sample size calculation to evaluate imprecision.
Literature was only searched in PubMed as this has shown to be an optimal tool in the field of biomedical research, even with free access [27,28]. With regard to the results of previous investigations [27,28], an additional search in Google Scholar and in paid databases such as Scopus and Web of Science was unlikely to provide further records of relevance for this review. However, this remains speculative as records from other databases were not available for comparison. Thus, the restriction to literature search in PubMed might be a limitation.

Conclusions
Risk factors of breast cancer (breast density, estrogens and estrogen metabolites and further parameters related to estrogen response) did not change in most trials despite a good adherence to isoflavone treatment, independent of the kind of intervention, the dose of isoflavones used and the duration of isoflavone treatment. However, the lack of significant changes does not prove the lack of effects as a sample size calculation was missing in most studies. Taking into account the RoB and methodological limitations, there is little evidence that isoflavone treatment modulates risk factors of breast cancer in preand postmenopausal women.
Future studies should calculate the samples size based on existing results to allow clear conclusions and should report further methodological details to reduce RoB. A metaanalysis of RCTs is warranted to judge if an isoflavone intake might contribute to the prevention of breast cancer.