Soy isoflavone consumption and colorectal cancer risk: a systematic review and meta-analysis

Colorectal cancer (CRC) is one of the most predominant solid carcinomas in Western countries. However, there is conflicting information on the effects of soy isoflavone on CRC risk. Therefore, we performed a meta-analysis to assess the association between soy isoflavone consumption and CRC risk in humans using PubMed, Embase, Web of Science, and Cochrane Library databases. A total of 17 epidemiologic studies, which consisted of thirteen case-control and four prospective cohort studies, met the inclusion criteria. Our research findings revealed that soy isoflavone consumption reduced CRC risk (relative risk, RR: 0.78, 95% CI: 0.72–0.85; I2 = 34.1%, P = 0.024). Based on subgroup analyses, a significant protective effect was observed with soy foods/products (RR: 0.79; 95% CI: 0.69–0.89), in Asian populations (RR: 0.79; 95% CI: 0.72–0.87), and in case-control studies (RR: 0.76; 95% CI: 0.68–0.84). Therefore, soy isoflavone consumption was significantly associated with a reduced risk of CRC risk, particularly with soy foods/products, in Asian populations, and in case-control studies. However, due to the limited number of studies, other factors may affect this association.


Statistical analysis.
We assessed the association between soy isoflavone consumption and CRC risk using the reported RRs. Soy isoflavones were defined as soy foods, soy products, isoflavones, tofu, soy milk, miso, natto, genistein, daidzein, and flavonols. When adjusted and crude RRs were provided, the most adjusted RRs were extracted.
We used HR and OR to evaluate CRC risk. HR and OR were considered to be approximations to RR, because CRC is a rare outcome in humans. Pooled RRs and 95% CIs were estimated on the basis of the most adjusted RRs or ORs for the highest versus lowest soy isoflavone intake.
We used I 2 and Q statistics to assess possible homogeneity of RRs across studies, which is a quantitative measure of inconsistency among studies 31 . Pooled ORs and 95% CIs were calculated using a random effects model 32 .
To estimate cancer site-specific and ethnicity-specific effects, subgroup analyses were performed by geographic area, study type, anatomical subsite, gender, and soy isoflavone type. Additionally, a sensitivity analysis was conducted to investigate the effect of a single study on the overall risk estimate. This allowed us to estimate whether the results could have been significantly affected by a single study.
Data analyses were performed with STATA version 13.0. Statistical significance was set at P ≤ 0.05. Egger's and Begger's regression models were used to evaluate potential publication bias 31 . All reported P values were from two-sided statistical tests.
Thirteen studies assessed the association between soy product consumption and CRC risk, while nine studies evaluated the association between isoflavone consumption and CRC risk. Among them, six studies separately presented findings for men and women, and two studies separately reported results for risk of rectal and colon cancers. Twelve studies were conducted in Asia and five in non-Asia countries ( Table 2). Data from both men and women were individually extracted. Different soy food types were evaluated in these studies; some studies assessed more than one type of soy food. Therefore, we used the risk estimate that was the most representative of overall soy consumption and the soy food item that was the most commonly consumed. In descending order, the most common soy food or products were tofu (bean curd), soy beans, soy milk, and miso soup (soy paste soup).
The analysis of the 17 studies yielded a combined risk estimate of 0.77 (95% CI, 0.72-0.82; P = 0.024) with a heterogeneity value (I 2 ) of 34.1% (Fig. 3). However, the results from the 17 studies were inconsistent. Nine studies reported that soy isoflavone intake was associated with a significant reduction in CRC risk, whereas other studies reported no association. Six studies reported that soy isoflavone intake was associated with a significant reduction in CRC risk in both men and women, three studies reported a significant reduction in CRC risk only in women, and other studies reported no association in women or men. We conducted a sensitivity analysis (Fig. 4) and meta regulation test (Fig. 5). The sensitivity analysis revealed that the publication dates were similar. The geographical area was associated with ~44.3% heterogeneity reduction across the studies. No publication bias was detected (Figs 6 and 7) based on Egger's and Begger's regression models 32 .

Discussion
We analyzed 17 epidemiological studies that assessed the association between soy isoflavone consumption and CRC risk in humans. The findings revealed that the consumption of soy isoflavones was associated with a 23% reduction in CRC risk. CRC is caused by environmental (e.g., diet and lifestyle) and genetic factors 34 . When stratified by geographical area, a significant protective effect of soy isoflavone consumption was observed in Asian populations, which are likely to be attributed to their lifestyle habits and overall health. Ecological and immigration studies have shown that differences in CRC risk among populations are largely attributed to environmental factors, such as eating habits. Asian populations have higher intakes of soy isoflavones than Western populations 35 . The consumption of Western diets, which are high in fat and calories, is associated with an increased incidence in CRC. Dietary fat increases the secretion of bile acids, which directly damage the intestinal mucosa, stimulate epithelial hyperplasia, and increase CRC risk 36 . On the other hand, the frequency of physical activity is lower in Asian populations than in American or European populations. Regular physical activity is a protective factor against CRC, because it reduces random motions of the intestine and stimulates bowel movements. Additionally, physical activity promotes the secretion of prostaglandins, which stimulate peristalsis and cleansing and reduce the contact time between the intestinal mucosa and carcinogens 37,38 . When stratified by study design, a significant protective effect of soy isoflavone intake was observed with case-control studies, which could be attributed to higher recall rates and greater selection bias in these types of studies. When stratified by soy foods/products and soy isoflavones, a significant protective effect was observed with soy foods/products, probably due to a limited number of studies focused on soy isoflavones.
Epidemiological and animal studies have found that the consumption of dietary soy decreases the incidence of certain tumors, including those of the colon and rectum [39][40][41][42][43] . The three main soy isoflavone aglycones are genistein, daidzein, and glycitein 43 . The mechanism by which soy protects against the development of CRC remains unclear. It has been reported that in CRC, there is a reduced expression of estrogen receptor-β (ER-β) expression 44 . Dietary isoflavones increase ER-β expression, but reduce ER-α expression in the colon of female rats 45 . In CRC patients, ER gene expression is either diminished or absent 46 .
Our meta-analysis had some limitations. First, only studies written in English were included. Second, most studies used FFQs as the main dietary assessment method. Recall bias may have affected the results. Additionally, it was challenging to predict the effect of misclassification of case-control studies on the results. Third, certain confounding factors were not adjusted in the evaluated studies, e.g., family history of CRC, smoking, and alcohol consumption, which are important risk factors of CRC [47][48][49] . Fourth, we failed to evaluate a dose-response relationship between soy isoflavone consumption and CRC risk.
There was heterogeneity across the studies in terms of soy isoflavone consumption, which is not surprising considering the differences in the study designs, soy types, and gender. Additionally, differences in geographic  area may have contributed to the heterogeneity results; most of the studies were conducted in Asia, where the consumption of soy is high. Moreover, while some studies were adjusted for age, gender, and family history of CRC in the calculation of risk estimates, not all parameters were considered. The measurement units varied among the studies. Sensitivity analysis was performed by sequentially omitting one single study to assess the effect of each study on the overall results (Fig. 2). The Egger's funnel plot revealed a P value > 0.05; the shape of the Begger's funnel plot seemed symmetrical. There was no significant evidence for publication bias in our meta-analysis (P > 0.05).