Soy intake and risk of endocrine-related gynaecological cancer: a meta-analysis

Abstract

Background  Epidemiology studies have reported associations between soy intake and the risk of endocrine-related gynaecological cancers. However, to date there have been no quantitative meta-analyses reported regarding this topic.

Objectives  We investigated the quantitative associations between soy food intake and the risk of endometrial cancer and ovarian cancer by a meta-analysis of case–control studies and cohort studies.

Search strategy  We searched MEDLINE (PubMed), EMBASE and the Cochrane Library during October 2008 using common keywords related to soy intake and endometrial or ovarian cancer. Two evaluators independently reviewed and selected articles, based on predetermined selection criteria.

Selection criteria  Included studies met all of the following criteria: (1) a case–control study or cohort study (to date, no randomized controlled trials have been reported); (2) investigated the associations between ‘soy or soy product intake’ and ‘endometrial cancer’ or ‘ovarian cancer’; (3) reported outcome measures with adjusted odds ratios (OR) or relative risks (RR) and 95% confidence intervals (CI).

Data collection and analysis  We investigated the associations between the overall soy intake (highest versus lowest intake) and the risk of endocrine-related gynaecological cancers (endometrial or ovarian cancer) as the main analysis. We also performed subgroup analyses by type of cancer (endometrial or ovarian), type of study design (case–control or cohort) and type of soy intake (soy foods or soy constituents).

Main results  Out of 477 articles that met our initial criteria, a total of seven epidemiology studies consisting of five case–control studies and two cohort studies were included in the final analyses. Compared with the lowest soy intake, the OR for the highest soy intake was 0.61 (95% CI, 0.53–0.72) of all endocrine-related cancers among seven studies; 0.70 for endometrial cancer (95% CI, 0.57–0.86) and 0.52 for ovarian cancer (95% CI, 0.42–0.66) in the fixed-effects meta-analyses. The subgroup analyses by study design showed similar findings among the case–control studies (OR, 0.62; 95% CI, 0.53–0.73) and the cohort studies (OR, 0.57; 95% CI, 0.36–0.90).

Author’s conclusions  The results of the current study showed protective effects of soy intake on the risk for endocrine-related gynaecological cancers. Additional larger prospective studies are now needed.

Among gynaecological cancers, endometrial cancer and ovarian cancer, the so called endocrine-related cancers, are known to be affected by hormonal and reproductive events.¹ The risk factors for endometrial cancer include nulliparity, early menarche, late menopause, obesity and postmenopausal estrogen replacement therapy.^2–5 Although the aetiology of ovarian cancer has not been determined, higher parity and oral contraceptive use have been reported to reduce the risk of ovarian cancer.⁶

Some plant-derived compounds possessing estrogen activity have been reported to be related inversely to the risk of endometrial cancer^7,8 or to have anti-tumour activity in ovarian cancer.^9,10 Phytoestrogens are nonsteroidal chemicals with a low molecular weight, similar to estrogens.¹¹ Among several phytoestrogens, the isoflavones have been vigorously investigated; they are abundant in soybeans and have the ability to bind the estrogen receptors α and β.¹² Isoflavones can act as weak plant-derived estrogens reducing the activity of animal sources of estrogen by several proposed mechanisms. They are thought to inhibit enzymes that synthesize estrogen and to enhance the production of the sex hormone-binding globulin.^13,14 Among isoflavones, genistein, daidzein and coumestrol have been recently reported to be natural selective estrogen-receptor modulators (SERM).¹⁵

According to reported epidemiological studies, such as case–control studies and cohort studies, soy foods or soy constituents have been shown to have properties that appear to prevent the potential development of hormone-dependent cancers. However, to date there have been no quantitative meta-analyses reported regarding this topic.

In the current study, we investigated the quantitative associations between soy food intake and the risk of hormone-related gynaecological cancers by a meta-analysis of case–control studies and cohort studies.

Methods

Literature search

We searched MEDLINE (PubMed) (1968 to August 2008), EMBASE (1977 to August 2008) and the Cochrane Library (1953 to August 2008) using common keywords related to soy intake and endometrial or ovarian cancer. The keywords were as follows: ‘soy’, ‘soybeans’, ‘tofu’, ‘miso’, ‘natto’, ‘soy protein’, ‘phytoestrogen’, ‘isoflavones’, ‘genistein’ and ‘daidzein’, for exposure factors; ‘endometrial cancer’, ‘uterine cancer’ and ‘ovarian cancer’ for outcome factors. We also reviewed the bibliographies of relevant articles to locate additional publications. The language of the publication was not restricted.

Selection criteria

We included epidemiology studies that met all of the following criteria: (1) a case–control study or cohort study (to date, no randomized controlled trials have been reported); (2) investigated the associations between ‘soy or soy product intake’ and ‘endometrial cancer’ or ‘ovarian cancer’; (3) reported outcome measures with adjusted odds ratios (OR) or relative risks (RR) and 95% confidence intervals (CI). If data were duplicated or shared in more than one study, the first published study was included in the analysis.

Selection of relevant studies

Two of the authors (Myung SK, Ju W) independently evaluated the eligibility of all studies retrieved from the databases as per the above selection criteria. Disagreements between evaluators were resolved by discussion. We extracted the following data from the studies included in the final analyses: study name (together with the name of first author and year of publication), journal name, country and design, study period (in years), participants (sample size, mean age and ethnicity of cases/controls), measure of soy intake, adjusted OR or RR with 95% CI and adjustments.

Main and subgroup analyses

We investigated the associations between the overall soy intake (highest versus lowest intake) and the risk of endocrine-related gynaecological cancers (endometrial or ovarian cancer) as the main analysis. We also performed subgroup analyses by the type of cancer (endometrial or ovarian), type of study design (case–control or cohort), type of soy intake (soy foods or soy constituents) and dosage of soy intake (lower, moderate and highest).

Statistical analyses

To compute a pooled OR with 95% CI, we used the adjusted ORs and 95% CIs of the highest soy intake compared with the lowest soy intake. We examined the heterogeneity in the results across studies using the Higgins I², which measures the percentage of total variation across studies.¹⁶ The I² was calculated as follows:

• 

where Q is Cochran’s heterogeneity statistic and df the degrees of freedom. Negative values of I² were set at zero; the I² results are between 0% (no observed heterogeneity) and 100% (maximal heterogeneity).¹⁶ An I² value >50% was considered to indicate substantial heterogeneity.

The pooled OR with 95% CI was calculated on the basis of both the fixed- and random-effects models: one assumes that the studies included in the meta-analysis had the same effect size; the other assumes that the studies included in the meta-analysis had varying effect sizes across the studies. When substantial heterogeneity was not observed, the pooled estimate calculated based on the fixed-effects model was reported; the Woolfe’s method (inverse variance method) was used in the fixed-effects model. When substantial heterogeneity was observed, the pooled estimate calculated based on the random-effects model was reported. The DerSimonian and Laird method for calculating summary measures was used in the random-effects models.¹⁷ In the current study, summary measures were mostly reported based on both the fixed- and random-effects model.

Regarding a dose–response relationship, to test for linear trend, a weighted linear regression was performed to model the natural logarithm (LN) of OR for the risk of endometrial cancer or ovarian cancer as a function of qualitatively described soy intake (lower = 1, moderate = 2 and highest = 3) using the inverse variance calculated from CI of each category; the standard error (SE) for the LN of the OR was estimated as {LN(upper limit)-LN(lower limit)}/2 × 1.96, and the inverse variance as 1/SE².

We evaluated publication bias regarding the studies included in the final analysis using Begg’s funnel plot and Egger’s test. If publication bias existed, Begg’s funnel plot would be asymmetrical or the P-value would be <0.05 by Egger’s test. We used the Stata SE version 10.0 software package (StataCorp, College Station, TX, USA) for the statistical analysis.

Results

Identification of relevant studies

Figure 1 shows a flow diagram of how we identified relevant studies. A total of 477 articles were identified by searching three databases and hand-searching relevant bibliographies. We excluded 165 duplicated articles and an additional 295 articles that did not satisfy the selection criteria. We reviewed the full texts of the remaining 17 articles. Among these, ten articles were excluded because of the following: not relevant to our analysis (n = 5);^18–22 included totally within another article (n = 2);^23,24 shared an identical study population with other articles (n = 2);^25,26 insufficient data (n = 1).²⁷ The remaining seven studies (five case–control studies^28–32 and two cohort studies^33,34) were included in the final analysis.

Figure 1.  Flow diagram for identification of relevant studies.

Characteristics of studies included in the final analysis

The seven studies included in the final analysis had 169 051 participants; 3516 cases and 165 535 controls. The mean age of the subjects was 54.1 (range, 18–85 years). Table 1 shows the general characteristics of the studies included in the final analysis. Three studies^28,29,31 involved endometrial cancers, and four studies involved ovarian cancers.^30,32–34

Table 1.   Characteristics of the studies included in the final analysis (n = 7)

Study (reference)	Journal	Country; design (Study name)	Study period (years)	Participants (mean age); ethnicity	Measure of soy intake	Adjusted OR or RR (95% CI)	Adjustment
PCC, population-based case–control study; HCC, hospital-based case–control study; PCS, prospective cohort study; OR, odds ratio; RR, relative risk; CI, confidence interval.
Case–control studies
1. Goodman et al.^28	Am J Epidemiol	USA; PCC	1985–1993 (8)	332 endometrial cancer cases (58.5 years) and 511 age- and ethnicity-matched controls (57.1 years); Japanese, Caucasian, Native Hawaiian, Filipino and Chinese)	Legumes (g/day): >39.0 versus <8.5	0.51 (0.31–0.86)	Pregnancy history, birth control pill use, estrogen use, history of diabetes mellitus, Quetelet’s index (kg/m^2) and total calories.
2 Horn-Ross, et al.^29	J Natl Cancer Inst	USA; PCC	1996–1999	500 endometrial cancer cases and 470 age- and ethnicity-matched controls; African American, Latina and white women	Isoflavone (μg/day): ≥2726 versus <1150	0.59 (0.37–0.93)	Age, race/ethnicity, daily caloric intake, age at menarche, parity, use of oral contraceptives and hormone replacement therapy and body mass index.
3. Zhang et al.^30	Nutr Cancer	China; PCC	1999–2000 (1)	254 ovarian cancer cases and 652 age-matched controls; Chinese	Soy foods (g/day): ≥136.4 versus ≤47.0	0.50 (0.31–0.82)	Age at diagnosis, education, area of residence, body mass index, tobacco smoking, alcohol consumption, tea drinking, physical activity, age at menarche, parity, menopausal status, hormone replacement therapy, oral contraceptive use, ovarian cancer in first-degree relatives, and total energy intake.
4. Xu et al.^31	Am J Epidemiol	China; PCC (The Shanghai Endometrial Cancer Study: SECS)	1997–2003	1042 endometrial cancer cases and 1035 age-matched controls	Soy protein (g/day): >13.3 versus <7.1	0.8 (0.6–1.0)	Age, education, menopausal status, years of menstruation, number of pregnancies, diagnosis of diabetes, alcohol consumption, body mass index, physical activity, energy intake, tea consumption and total fruit and vegetable intake
5. Rossi et al.^32	Int J Cancer	Italy; HCC	1992–1999 (7)	1301 ovarian cancer cases and 2411 controls	Isoflavone (μg/day): >32.5 versus <12.8	0.51 (0.37–0.69)	Age, study center, education, year of interview, parity, oral contraceptive use and family history of ovarian or breast cancer or both in first-degree relatives.
Cohort studies
1. Chang et al.^33	Am J Epidemiol	USA; PCS (The California Teachers Study)	1995–2003 (8)	280 ovarian cancer cases among 97275 women	Isoflavone (mg/day): >3 versus <1	0.56 (0.33–0.96)	Race, total energy intake, parity, oral contraceptive use, strenuous exercise, wine consumption and menopausal status/hormone therapy use
2. Sakuchi et al.^34	Nutr Cancer	Japan; PCS (The Japan Collaborative Cohort study: JACC)	1988–2003 (15)	77 ovarian cancer death cases among 63541 women	Soybean curd (times/week): Almost every day versus ≤1–2	0.61 (0.26–1.45)	Age, menopausal status, number of pregnancies, history of sex hormone use, BMI, physical activity and education.

The selected studies were published between 1997 and 2008, spanning 11 years. The studies were conducted in the following countries: the US (n = 3),^28,29,33 China (n = 2),^30,31 Italy (n = 1),³² Japan (n = 1).³⁴ Among five case–control studies, four studies^28–31 were population-based case–control studies and the other one³² was a hospital-based case–control study. The range of enrollment/study periods was 1985–2003.

Soy intake and the risk of endocrine-related gynaecological cancer

As shown in Figure 2, soy intake was significantly associated with a reduced risk of endocrine-related gynaecological cancers (endometrial cancer and ovarian cancer) in the fixed-effects model meta-analysis of all seven studies (pooled OR, 0.61; 95% CI, 0.53–0.72), without significant heterogeneity (I² = 12.1%) and in the random-effects model meta-analysis (pooled OR, 0.61; 95% CI, 0.51–0.72). No publication bias was observed in the selected studies (Begg’s funnel plot was symmetrical; Egger’s test, P for bias = 0.23). In addition, Figure 2 shows the preventive effects of soy intake on endometrial cancer (n = 3) and ovarian cancer (n = 4): the pooled ORs of the soy intake population were 0.70 (95% CI, 0.57–0.86; fixed-effects model) and 0.67 (95% CI, 0.51–0.88; random-effects model) for endometrial cancer and 0.52 (95% CI, 0.42–0.66; both fixed-effects and random-effects models) for ovarian cancer.

Figure 2.  Soy intake and the risk of endocrine-related gynaecological cancers by type of cancer in a meta-analysis of epidemiologic studies (n = 7). *Fixed-Effects Model. OR, odd ratio; CI, confidence interval.

In the subgroup analyses, by study design, both case–control and cohort studies demonstrated the preventive effects of soy intake on hormone-related gynaecological cancers (Figure 3). In the fixed-effects model meta-analysis, the pooled ORs were 0.62 (95% CI, 0.53–0.73; n = 5) for the case–control studies and 0.57 (95% interval, 0.36 to 0.90) for the cohort studies. Regarding the type of soy intake, both soy foods (OR, 0.69; 95% CI, 0.58–0.83; n = 5) and isoflavones (OR, 0.53; 95% CI, 0.43–0.66; n = 4) had a preventive effect on hormone-related gynaecological cancers (Table 2). This preventive effect was also found in genistein and daidzein.

Figure 3.  Soy intake and the risk of endocrine-related gynaecological cancers by study design in a meta-analysis of epidemiologic studies (n = 7). *Fixed-effects model. OR, odd ratio; CI, confidence interval.

Table 2.   Relationships between soy intake and the risk of endometrial cancer or ovarian cancer by type of soy intake in subgroup meta-analyses

Category	No. studies	Summary OR (95% CI)	Model used	Heterogeneity, I^2
Soy foods	5	0.69 (0.58–0.83)	Fixed-effects	12.0%
		0.68 (0.55–0.83)	Random-effects
Isoflavones	4	0.53 (0.43–0.66)	Fixed-effects	0.0%
		0.53 (0.43–0.66)	Random-effects
Genistein	3	0.61 (0.47–0.81)	Fixed-effects	0.0%
		0.61 (0.47–0.81)	Random-effects
Daidzein	3	0.66 (0.50–0.86)	Fixed-effects	0.0%
		0.66 (0.50–0.86)	Random-effects
Glycitein	1	0.59 (0.35–0.97)	n.a	n.a
Coumestrol	1	1.10 (0.70–1.70)	n.a	n.a
Lignans	1	0.68 (0.44–1.10)	n.a	n.a

As shown in Table 3, a negative dose–response relationship was observed between soy intake and the risk of endometrial cancer or ovarian cancer in subgroup meta-analyses (P for trend = 0.025).

Table 3.   Dose-response relationship between soy intake and the risk of endometrial cancer or ovarian cancer in subgroup meta-analyses*

Category of soy intake (vs lowest)	No. studies	Summary OR (95% CI)	Heterogeneity, I^2	Model used	P for trend**
*Highest intake was defined as quintile 5, quartile 4 or tertile 3; moderate intake as quintile 4 or quartile 3; lower intake as quintile 2, quartile 2 or tertile 2, respectively, based on each study’s categorization. **A weighted linear regression was performed to model the natural logarithm of OR for the risk of endometrial cancer or ovarian cancer as a function of qualitatively described soy intake (lower = 1, moderate = 2 and highest = 3) using the inverse variance calculated from confidence intervals of each category; standard error = {LN(upper limit) − LN(lower limit)}/2 × 1.96; inverse variance = 1/(standard error × standard error).
Lower	7	0.94 (0.83–1.06)	34.1%	Fixed-effects	0.025
Moderate	5	0.77 (0.65–0.91)	45.3%	Fixed-effects
Highest	7	0.61 (0.53–0.72)	12.1%	Fixed-effects

Discussion

The results of the current meta-analysis found that participants with higher soy intake had a reduced risk for endocrine-related gynaecological cancers such as endometrial cancer and ovarian cancer, when compared with lower soy intake. These preventive effects were found regardless of the type of cancer (endometrial cancer or ovarian cancer), the type of study (case–control study or cohort study) and the type of soy intake (soy foods or soy constituents). Also, a negative dose–response relationship was observed between them.

Several biological mechanisms have been proposed regarding the preventive effects of phytoestrogens abundant in soy products. Estradiol-17β (E2) acts via the estrogen receptor (ER), which is composed of two subtypes, ERα and ERβ.³⁵ ERα is responsible for tissue proliferation and consequently for the development of malignancies.^36,37 ERβ, which was cloned from rat prostate in 1996,³⁵ appears to be involved in the differentiation of proliferating tissue or the antagonism of ERα activity.³⁸ Phytoestrogens show weak estrogenic activity via molecular binding to the estrogen receptor because isoflavones, especially genisteins, share a structural similarity with E2.³⁹ Genistein has been known to have a slightly higher affinity for recombinant human ERβ than for ERα.⁴⁰ This is one of the probable mechanisms explaining the reduced risk of endocrine-related gynaecological cancers observed in the current study.

Isoflavones have been shown to inhibit 17β-hydroxysteroid dehydrogenase (HSD) type 1, which reduces Estrone (E1) to E2.⁴¹ A laboratory study showed that the reduction of 17β-HSD type 5 that converts androstenedione to testosterone was inhibited by flavones.⁴² Genistein has been shown to stimulate the oxidative reaction of 17β-HSD type 2 that converts E2 to the less active E1 and testosterone to androstenedione.⁴³ Furthermore, phytoesterogens are able to enhance the production of sex hormone-binding globulin, thus reducing the circulating concentration of free active hormones.⁴⁴ In addition to the estrogen stimulating mechanisms, as described above, several other biological properties of phytoestrogens, which are generally detected in anti-tumour cytotoxic agents, have been suggested: inhibiting tumour angiogenesis,⁴⁵ cell proliferation,⁴⁶ tyrosine kinase⁴⁷ and topoisomerase II.⁴⁸

Our study has several strengths. Low heterogeneity was observed for most subgroup analyses as well as in the main analysis, and the pooled ORs with 95% CIs consistently showed preventive effects of soy intake. These findings indicate that our conclusion regarding the protective effects of soy intake, on endocrine-related gynaecological cancers, was a consistent finding with little bias. In addition, we used the adjusted data rather than unadjusted data. Even though the confounding variables used in individual studies were different, we could minimize the effects of confounding variables such as age, ethnicity, body mass index, social-economic status, history of hormonal replacement therapy or use of oral contraceptive.

Despite the strengths of the study, this study has several limitations. First, it does not provide the highest level of evidence. Among the seven studies included in the final analysis, five studies were case–control studies. In general, case–control studies have two main biases, recall bias and selection bias. We are not able to rule out that these biases caused spurious associations. Cohort studies or randomized controlled trials could overcome these problems with reduced bias. However, we included only two cohort studies in this study, because, to date, only two cohort studies and no randomized controlled trials, have been reported. Therefore, additional cohort studies and randomized controlled trials are needed to confirm the findings of preventive effects of soy intake on the risk of hormonal-related gynaecological cancers.

In addition, we can not suggest the optimal amount of soy food intake for the prevention of endocrine-related gynaecological cancers. The cut-off points for soy food intake and the types of soy foods varied across studies. Furthermore, the criterion for a sufficient amount of soy food that causes biochemical alterations in the human body has not yet been determined.

Moreover, the protective mechanisms of phytoestrogens, from the development of endocrine-related cancers, have not been entirely elucidated, thus caution should be taken in the interpretation of the epidemiology data. The in vitro effect of phytoestrogens on cell growth can change; it may inhibit or stimulate depending on the concentration. Isoflavones stimulated the growth of breast cancer cells at low concentration, but inhibited growth at a high concentration.^49,50 Even though the biology of breast cancer is different from that of endocrine-related gynaecological cancers, the potential for biphasic effects should be considered when isoflavones are studied in association with gynaecological cancers.

In summary, our meta-analysis of seven epidemiology studies shows statistical evidence linking soy intake to a reduced risk of endocrine-related gynaecological cancers, such as endometrial cancer and ovarian cancer, based on the currently published data. Our findings should be confirmed by large prospective cohort studies and/or randomized controlled trials providing a higher level of evidence.

Disclosure of interests

The authors declare that there are no conflicts of interest.

Contribution to authorship

W Ju and SK Myung were responsible for the initial plan, study design, statistical analysis and the conduct of the study. W Ju, SK Myung, HJ Choi and SC Kim were responsible for data collection, data extraction, data interpretation and manuscript drafting. W Ju, SK Myung and HJ Choi were responsible for data interpretation and manuscript drafting. W Ju is the guarantor for this paper and has full responsibility for this study.

Details of ethics approval

None.

Funding

None.

Acknowledgement

None.