Direct effects

The cholesterol-lowering effect of soya protein was first demonstrated clinically in 1967⁽Reference Hodges, Krehl and Stone²⁴⁾, but not until 1995 did the hypocholesterolaemic effects of soya protein receive widespread recognition. In that year, a meta-analysis of the clinical data by Anderson et al. ⁽Reference Anderson, Johnstone and Cook-Newell²⁵⁾, which included thirty-four trials and thirty-eight different comparisons, found that soya protein lowered LDL-cholesterol (LDL-C) by 12·9 %. The reduction in LDL-C was attributed to the protein or components (isoflavones) associated with soya protein and not to the low saturated fat content of the soyabean, since in nineteen trials the fatty acid content of the soya and control diets was similar. And 4 years later, on the basis of their own analysis, the US FDA approved a health claim for soyafoods and CHD⁽¹³⁾. In the subsequent year, the American Heart Association (AHA) recognised the ability of soya protein to lower cholesterol and endorsed the use of soyafoods for people with elevated cholesterol levels⁽Reference Erdman²⁶⁾.

However, only 2 years following approval of the health claim, the US Adult Treatment Panel III of the National Cholesterol Education Program chose not to endorse the hypocholesterolaemic effects of soya protein, citing inconsistent data and the lack of dose–response information⁽²⁷⁾. Of greater significance, in their 2006 position paper, the AHA, while continuing to endorse the use of soyafoods for heart-healthy diets because of their ‘high content of polyunsaturated fats, fibre, vitamins and minerals, and low content of saturated fat … ’ concluded that earlier research indicating that ‘soya protein has clinically important favourable effects as compared to other proteins has not been confirmed’⁽Reference Sacks, Lichtenstein and Van Horn¹⁷⁾. Furthermore, the FDA announced in December 2007 their intention to re-evaluate the evidence in support of the health claim, although there is no indication that this re-evaluation has been undertaken because the data are no longer supportive of the claim.

Given the gravitas of the AHA, and their pronounced change in sentiment about the efficacy of soya protein, it is instructive to consider the evidence upon which their reversal is based. Briefly, in an analysis of twenty-two studies, the AHA estimated that soya protein lowered LDL-C only by approximately 3 %⁽Reference Sacks, Lichtenstein and Van Horn¹⁷⁾. In their view, this degree of reduction was insufficient to warrant a health claim. However, the AHA did not actually conduct a formal statistical meta-analysis of the data. When this was recently done, Jenkins et al. ⁽Reference Jenkins, Mirrahimi and Srichaikul⁶⁾ found that the AHA had underestimated the hypocholesterolaemic effects of soya protein. Rather than lowering LDL-C by approximately 3 %, Jenkins et al. ⁽Reference Jenkins, Mirrahimi and Srichaikul⁶⁾ estimated that soya protein lowered LDL-C by 4·3 %. Furthermore, when the analysis was limited to eleven of the twenty-two studies which provided evidence that the control and soya-containing diets were nutritionally matched, soya protein was found to lower LDL-C by 5·2 %. This estimate is in line with the results of other recently published meta-analyses⁽Reference Zhan and Ho^28–Reference Anderson and Bush³²⁾ (Table 1) and is similar to the cholesterol-lowering effects of soluble fibre, which also has an FDA health claim⁽Reference Brown, Rosner and Willett³³⁾. Although the consensus estimate (4–6 %) of the magnitude of the hypocholesterolaemic effect of soya protein is much more modest than that initially reported by Anderson et al. ⁽Reference Anderson, Johnstone and Cook-Newell²⁵⁾, the decrease is still relevant at both population and patient levels.

Table 1 Change (%) in circulating lipid levels in response to soya protein: results from recently published meta-analyses

↓ , Decrease; LDL-C, LDL-cholesterol; ↑ , increase; HDL-C, HDL-cholesterol; NR, not reported.

* Refers to subject numbers for LDL-C measurements.

† Refers to parallel studies only.

‡ 7·70 mg/l increase.

§ 62·60 mg/l decrease.

The FDA established 25 g/d soya protein as the threshold intake required to lower LDL-C because most trials used at least this much protein, not because lower amounts were shown not to be efficacious. There is, in fact, evidence suggesting that fewer than 25 g/d soya protein is needed to lower cholesterol⁽Reference Harland and Haffner^29, Reference Messina³⁴⁾. Whether exceeding 25 g/d produces larger decreases in LDL-C is unclear. In the overall statistical model in the initial meta-analysis by Anderson et al. ⁽Reference Anderson, Johnstone and Cook-Newell²⁵⁾, the amount of soya protein did not make an impact on cholesterol reduction although when adjusted for baseline cholesterol, it was estimated that 25 and 50 g/d soya protein decreased serum cholesterol concentrations by 89 and 174 mg/l, respectively. Zhan & Ho reported that LDL-C decreased to a larger extent in response to soya protein providing ≥ 80 mg/d isoflavones⁽Reference Zhan and Ho²⁸⁾. In agreement, Reynolds et al. ⁽Reference Reynolds, Chin and Lees³⁰⁾ found significant correlations between the reduction in LDL-C and soya protein (r − 0·57, P < 0·001) and isoflavone (r − 0·48, P < 0·01) intake. However, Weggemans & Trautwein⁽Reference Weggemans and Trautwein³¹⁾ found that there was no dose–response relationship between soya-associated isoflavones and changes in LDL-C. Furthermore, in the meta-regression analysis by Harland & Haffner⁽Reference Harland and Haffner²⁹⁾, no dose–response relationship between soya protein intake in the range of 15–40 g/d and LDL-C reduction was observed. This finding is particularly noteworthy because this range of soya protein intake more closely reflects the amounts likely to be consumed by free-living populations than the ranges in the other meta-analyses (Table 1). In support of Weggemans & Trautwein⁽Reference Weggemans and Trautwein³¹⁾ and Harland & Haffner⁽Reference Harland and Haffner²⁹⁾, the results of the recent meta-analysis by Anderson & Bush⁽Reference Anderson and Bush³²⁾ also support the notion that neither dose of soya protein (range, 15–50 g/d) nor isoflavones (range, 22–185 mg/d) affect the LDL-C-lowering response.

Participants in the studies in the meta-analyses listed in Table 1 were roughly equally divided between men and women; however, there is some evidence indicating that there are sex differences in the response to soya protein. For example, Zhan & Ho⁽Reference Zhan and Ho²⁸⁾ found that LDL-C reductions in response to soya protein for men and women were − 0·30 and − 0·14 mmol/l, respectively, although it is not clear whether their analysis controlled for potentially confounding variables such as baseline cholesterol levels. Nevertheless, in agreement, Harland & Haffner⁽Reference Harland and Haffner²⁹⁾ found greater LDL-C reductions in men than in women.

Indirect effects

In addition to the direct hypocholesterolaemic effect of soya protein, as a result of differences in fatty acid content, Jenkins et al. ⁽Reference Jenkins, Mirrahimi and Srichaikul⁶⁾ using US National Health and Nutrition Examination Survey III population survey data, estimated that when soyafoods replace commonly consumed sources of animal protein in the diet, LDL-C is reduced by 3–6 %. There was a 4 % reduction in LDL-C when 24 g soya protein – an amount similar to the 25 g/d established by FDA as the threshold intake for cholesterol reduction – replaced a comparable amount of animal protein. Obviously, the displacement effect can vary greatly, as it depends entirely on the types of soyafoods consumed and the types of foods that are replaced.

As an oil seed, soyabeans derive approximately 40 % of their energy from fat, which is much higher than that for other legumes⁽Reference Messina³⁵⁾ with the exception of peanuts, another oil seed legume. The predominant fatty acid in soya oil, as in many vegetable oils, is the essential n-6 PUFA linoleic acid, which accounts for about 55 % of the total fat content⁽Reference Wu, Rodgers and Marshall^3, Reference Slavin, Kenworthy and Yu³⁶⁾. The soyabean has a moderate amount of oleic acid (approximately 29 %) and a low amount of saturated fat (approximately 12 %). Importantly, unlike many other vegetable oils, soya oil is comprised of approximately 6 % α-linolenic acid, the essential n-3 fatty acid⁽Reference Wu, Rodgers and Marshall^3, Reference Slavin, Kenworthy and Yu³⁶⁾.

Although conventional thinking has been that replacing cholesterol-raising saturated fats with n-6 PUFA will lower cholesterol and CHD risk⁽Reference Calder³⁷⁾, a recent comprehensive analysis of the clinical data found that to lower CHD, saturated fat needs to be replaced with a mix of n-6 and n-3 PUFA⁽Reference Ramsden, Hibbeln and Majchrzak³⁸⁾. In fact, according to Ramsden et al. ⁽Reference Ramsden, Hibbeln and Majchrzak³⁸⁾, replacement of saturated fat with n-6 PUFA increases risk⁽Reference Ramsden, Hibbeln and Majchrzak³⁸⁾. Replacing saturated fat with refined carbohydrate also increases risk, whereas the effect of replacement with monounsaturated fat is unclear⁽Reference Ramsden, Hibbeln and Majchrzak^38–Reference Astrup, Dyerberg and Elwood⁴¹⁾. Therefore, the fatty acid profile of soyabeans would appear to be ideally suited for lowering blood cholesterol. On the other hand, the high n-6:n-3 fatty acid ratio (approximately 9:1) has led to concern that soyafoods might elicit a pro-inflammatory response. However, this fear does not appear to be justified⁽Reference Harris, Mozaffarian and Rimm⁴²⁾.

Linoleic acid is converted in vivo to arachidonic acid, the fatty acid from which pro-inflammatory eicosanoids are produced; however, plasma and tissue levels of arachidonic acid are very tightly regulated⁽Reference Rett and Whelan⁴³⁾. Further, it is now recognised that some eicosanoids produced from arachidonic acid are anti-inflammatory and possess other desirable attributes⁽Reference Calder⁴⁴⁾. To this point, in a cross-sectional study of 915 men aged 40–49 years, both serum linoleic acid and arachidonic acid were found to be inversely related to plasma levels of plasminogen activator inhibitor-1. Because plasminogen activator inhibitor-1 is a primary inhibitor of plasminogen activators, it has an anti-fibrinolytic function.

Finally, given the desirable fatty acid profile of and high-quality protein⁽Reference Rand, Pellett and Young^1, Reference Hughes, Ryan and Mukherjea²⁾ provided by soyafoods, it is not surprising that in research diets shown to dramatically lower blood LDL-C levels, soyafoods have played an important role⁽Reference Jenkins, Kendall and Faulkner^45, Reference Jenkins, Kendall and Marchie⁴⁶⁾. If one assumes that there is a 4 % reduction in LDL-C resulting from a direct effect of consuming 25 g/d soya protein and a 4 % reduction due to displacement effects, in theory, CHD risk will be reduced by 8–16 % when soyafoods replace conventional sources of protein in Western diets⁽Reference Law, Wald and Thompson^47, Reference Law, Wald and Wu⁴⁸⁾. Obviously, only long-term clinical trials can definitively show that incorporating soyafoods into the diet reduces coronary events.

Brief overview of isoflavones

The three soyabean isoflavones, genistein, daidzein and glycitein, and their respective glycosides account for approximately 50, 40 and 10 %, respectively, of total soyabean isoflavone content⁽Reference Murphy, Barua and Hauck⁸²⁾. In Japan, mean isoflavone intake ranges from about 30 to 50 mg/d (expressed as aglycone equivalents)⁽Reference Messina, Nagata and Wu⁸³⁾. In considering their possible impact on breast tissue, it is important to recognise that although isoflavones are classified as phyto-oestrogens, they are more accurately classified as selective ER modulators (SERM), a group which includes the breast cancer drug tamoxifen and the osteoporosis drug raloxifene⁽Reference Oseni, Patel and Pyle⁸⁴⁾. SERM have tissue-selective effects; depending upon the tissue in question, SERM can have oestrogenic effects, anti-oestrogenic effects or no effects at all on tissues affected by oestrogen. This will depend on the types and ratios of ER, the shape of the ligand–receptor complex and the types of co-activators and co-repressors in cells in a particular tissue⁽Reference Pike⁸⁵⁾. The tissue selectivity of isoflavones probably derives from their preferential binding to and transactivation of ERβ in comparison to the ERα⁽Reference Oseni, Patel and Pyle^84, Reference Reiter, Beck and Medjakovic⁸⁶⁾. These two receptors have different tissue distributions in the body and different functions. This is particularly evident in the breast where the activation of ERβ is thought to inhibit the stimulatory and proliferative effects of ERα activation⁽Reference Speirs, Carder and Lane⁸⁷⁾.

Although limited, there is considerable evidence supporting the classification of isoflavones as SERM. For example, Carmignani et al. ⁽Reference Carmignani, Pedro and Costa-Paiva⁸⁸⁾ found that the vasomotor symptom score of postmenopausal women was decreased over a 16-week period by approximately 33, 54 and 55 % in response to a placebo, soya protein containing 53 mg isoflavones, and hormone therapy (HT, oestrogen plus progestin), respectively. However, only HT increased the vaginal maturation index ( < 0·01), a measure of the extent to which the vaginal tissue has been exposed to oestrogenic molecules. Thus, the oestrogen-like effects of isoflavones were apparently sufficient to alleviate menopausal symptoms in a manner similar to HT but unlike HT, isoflavones did not exert an oestrogenic effect on the vagina.

Isoflavones, oestrogen and breast cancer: clinical evidence

It has been postulated that in the high-oestrogen environment of premenopausal women, isoflavones function as oestrogen antagonists and potentially decrease breast cancer risk, whereas in the low-oestrogen environment of postmenopausal women, they function as oestrogen agonists, and increase risk⁽Reference Bondesson and Gustafsson^89, Reference Penttinen-Damdimopoulou, Power and Hurmerinta⁹⁰⁾. However, although circulating oestradiol levels are 4-fold higher in pre- compared to postmenopausal women, breast tissue oestrogen concentrations are similar⁽Reference Pasqualini, Chetrite and Blacker^91, Reference Geisler⁹²⁾ because of the uptake of oestrogen by the breast from the circulation⁽Reference Haynes, Straume and Geisler⁹³⁾. Thus, the notion that the effect of isoflavones on breast tissue differs according to the oestrogenic environment may not be physiologically relevant. Furthermore, in the Women's Health Initiative trial, although oestrogen-plus-progestin use increased breast cancer incidence⁽⁹⁴⁾ and mortality⁽Reference Chlebowski, Anderson and Gass⁹⁵⁾, in hysterectomised women, oestrogen-only use actually decreased risk⁽Reference Anderson, Limacher and Assaf^96, Reference Stefanick, Anderson and Margolis⁹⁷⁾ and the protective effects were maintained for several years after cessation of therapy⁽Reference LaCroix, Chlebowski and Manson⁹⁸⁾. One caveat, however, is that the most recent analysis from the Women's Health Initiative trial suggests that the protective effects of oestrogen may not apply to women at increased risk of breast cancer⁽Reference Anderson, Chlebowski and Aragaki⁹⁹⁾.

These findings from the Women's Health Initiative trial are generally supported by the existing epidemiological⁽Reference Warren^100–Reference Chen, Manson and Hankinson¹⁰⁴⁾ and clinical data⁽Reference Greendale, Reboussin and Slone^105–Reference Conner, Soderqvist and Skoog¹⁰⁸⁾. Granted, the extent to which the findings in women without a uterus can be applied to intact women is unclear; however, there is no reason why women without a uterus should be more or less sensitive to oestrogens. It is true that hysterectomy, even if the ovaries are not removed, may be associated with an earlier endocrine menopause. Such women are therefore more likely to resemble the postmenopausal state and yet are protected by oestrogen therapy.

Clinical studies in which the effects of isoflavone exposure on markers of breast cancer risk have been evaluated are supportive of safety regardless of whether isoflavones are derived from soyafoods or supplements⁽Reference Messina and Wood¹⁰⁹⁾. Higher circulating oestrogen levels, greater breast tissue density and greater in vivo breast cell proliferation, are seen as reflecting an increased cancer risk. To this point, tamoxifen decreases cell proliferation and breast tissue density and breast cancer risk, whereas HT increases these markers and breast cancer risk⁽Reference Conner, Skoog and Soderqvist^107, Reference Conner, Soderqvist and Skoog^108, Reference Boyd, Martin and Li^110–Reference Conner¹¹²⁾. In contrast to the effects of HT, isoflavones generally have no effect on any of these previously referred to markers⁽Reference Hooper, Ryder and Kurzer^113, Reference Huber, Imhof and Schmidt¹¹⁴⁾. One exception is a recently published 6-month study that found in response to very high-dose isoflavone supplementation (154 mg/d genistein), breast cell proliferation increased slightly in premenopausal women, although there were no effects on cell cytology and no effects at all in postmenopausal women, the latter group according to theory, being most vulnerable to isoflavones⁽Reference Khan, Chatterton and Michel¹¹⁵⁾. Further, the approximate 27 % increase in proliferation observed in premenopausal women is more than an order of magnitude less than the approximately 4- to 10-fold increased breast cell proliferation that occurs in response to HT⁽Reference Conner, Soderqvist and Skoog^108, Reference Murkes, Conner and Leifland¹¹⁶⁾.

Finally, it is worth noting that the clinical findings discussed previously, although supportive of safety, also argue against adult soya intake reducing breast cancer risk as was first proposed more than 20 years ago⁽Reference Messina and Barnes^117–Reference Setchell, Borriello and Hulme¹¹⁹⁾, even though epidemiological data show that among Asian women soya intake is associated with protection against breast cancer⁽Reference Wu, Yu and Tseng⁴⁾. Compelling, but still speculative, evidence indicates that to derive protection against breast cancer, soya intake must occur during childhood and/or adolescence⁽Reference Messina and Hilakivi-Clarke^9, Reference Messina and Wu¹⁰⁾.

Soya intake and the prognosis of breast cancer patients: epidemiology

Despite the lack of effect of isoflavone exposure on breast tissue, recent epidemiological evidence indicates that soya intake benefits breast cancer patients. Over the past 6 years, six prospective epidemiological studies, three from China⁽Reference Boyapati, Shu and Ruan^120–Reference Kang, Ansbacher and Hammoud¹²²⁾, two from the USA⁽Reference Guha, Kwan and Quesenberry^123, Reference Caan, Natarajan and Parker¹²⁴⁾ and one from Korea⁽Reference Woo, Park and Ro¹²⁵⁾, have evaluated the impact of soya consumption on the prognosis of breast cancer patients. One of these showed there to be no relationship between soya intake and disease-free survival⁽Reference Boyapati, Shu and Ruan¹²⁰⁾, four showed that soya intake improved prognosis⁽Reference Shu, Zheng and Cai^121, Reference Guha, Kwan and Quesenberry^123, Reference Caan, Natarajan and Parker^124, Reference Kang, Zhang and Wang¹²⁶⁾, whereas the Korean study found that the impact of soya depended upon the type of breast cancer in question⁽Reference Woo, Park and Ro¹²⁵⁾. That study⁽Reference Woo, Park and Ro¹²⁵⁾ and the largest and longest Chinese⁽Reference Shu, Zheng and Cai¹²¹⁾ and US⁽Reference Caan, Natarajan and Parker¹²⁴⁾ studies are briefly described next. When considering the Asian studies, it is worth noting that an argument has been made on the basis of animal studies, that breast tumours that develop in women who consumed soya when young may be different from and respond differently to tamoxifen, than breast tumours in women who did not consume soya early in life⁽Reference Hilakivi-Clarke, Andrade and Helferich¹²⁷⁾.

The Shanghai Breast Cancer Survival Study is a population-based cohort study that includes more than 5000 breast cancer survivors⁽Reference Shu, Zheng and Cai¹²¹⁾. During the median follow-up period of approximately 3·9 years, the hazard ratio (HR) plus 95 % CI associated with the highest soya protein intake quartile was 0·71 (95 % CI 0·54, 0·92) for total mortality and 0·68 (95 % CI 0·54, 0·87) for recurrence compared with the lowest intake quartile. Similar protective effects were associated with isoflavone intake. Soya intake was assessed at 6, 18, 36 and 60 months post enrolment and the fourth quartile intake cut-offs for soya protein and isoflavones were >15·31 g/d and >62·68 mg/d, respectively. Soya intake was protective in both pre- and postmenopausal women.

The US study by Caan et al. ⁽Reference Caan, Natarajan and Parker¹²⁴⁾ involved 2736 breast cancer survivors diagnosed between 1991 and 2000 with early-stage breast cancer, who were participants in the Women's Healthy Eating and Living study. During the median 7·3 years follow-up period, there were 448 new breast cancer events and 271 deaths. The results showed that as isoflavone intake increased, risk of death decreased. Women in the highest isoflavone intake category (cut-off, >16·3 mg/d; median 26·7 mg/d) had a 54 % reduction in risk of death (HR 0·46; 95 % CI 0·2, 1·05; P for trend = 0·02). The benefits of isoflavone intake were most obvious in tamoxifen users (HR 0·26; 95 % CI 0·0, 1·08; P for trend = 0·05). This finding is consistent with the observation that in the prospective Chinese epidemiological study by Kang et al. ⁽Reference Kang, Zhang and Wang¹²⁶⁾, soya intake improved the efficacy of the aromatase inhibitor anastrozole. The tamoxifen and anastrozole findings are especially noteworthy not only because of the possible implications for breast cancer patients on treatment, but because in ovariectomised athymic mice, genistein was shown to inhibit the efficacy of these drugs⁽Reference Ju, Doerge and Allred^78, Reference Ju, Doerge and Woodling⁷⁹⁾. Furthermore, both Chinese and US prospective studies show benefits to breast cancer patients of post-diagnosis soya intake, including those treated with tamoxifen. These findings help to address the concern that women who consume soya early in life develop tumours that differ from those who do not.

Finally, in a Korean prospective study by Woo et al. ⁽Reference Woo, Park and Ro¹²⁵⁾, among the 339 women aged 25–77 years, during the median follow-up period of 32·6 months (range 1·3–52·4 months), there were twenty-five recurrences. The adjusted HR for women in the third isoflavone intake tertile (cut-off, ≥ 14·6 mg/d) was 0·23 (P for trend, 0·01) for women with epidermal growth factor receptor-2 negative (HER2 − ) breast cancer, whereas for women with HER2+ breast cancer the adjusted HR (cut-off, ≥ 15·2 mg/d) was 3·85 (P for trend, 0·17). Only approximately 25 % of breast cancers are HER2+, but this type of breast cancer is more aggressive and is associated with a worse prognosis⁽Reference Hicks and Kulkarni¹²⁸⁾. Although these findings raise a cautionary note, no conclusions can be made on the basis of this study because of its limited size, as there were only eight recurrences among the eighty-three HER2+ breast cancer patients. The impact of post-diagnosis soya intake in HER2+ breast cancer patients in a much larger cohort is currently being evaluated.