Lignans and breast cancer risk in pre- and post-menopausal women: meta-analyses of observational studies

Phyto-oestrogens are plant compounds structurally similar to oestradiol, which have been proposed to have protective effects against breast cancer. The main class of phyto-oestrogens in the Western diet is lignans. Literature reports on the effect of lignans in breast cancer risk have been conflicting. We performed three separate meta-analyses to examine the relationships between (i) plant lignan intake, (ii) enterolignan exposure and (iii) blood enterolactone levels and breast cancer risk. Medline, BIOSIS and EMBASE databases were searched for publications up to 30 September 2008, and 23 studies were included in the random effects meta-analyses. Overall, there was little association between high plant lignan intake and breast cancer risk (11 studies, combined odds ratio (OR): 0.93, 95% confidence interval (95% CI): 0.83–1.03, P=0.15), but this association was subjected to marked heterogeneity (I2=44%). Restricting the analysis to post-menopausal women, high levels of plant lignan intake were associated with reduced breast cancer risk (7 studies, combined OR: 0.85, 95% CI: 0.78, 0.93, P<0.001) and heterogeneity was markedly reduced (I2=0%). High enterolignan exposure was also associated with breast cancer (5 studies, combined OR: 0.73, 95% CI: 0.57, 0.92, P=0.009) but, again, there was marked heterogeneity (I2=63%). No association was found with blood enterolactone levels (combined OR: 0.82, 95% CI: 0.59–1.14, P=0.24). In conclusion, plant lignans may be associated with a small reduction in post-menopausal breast cancer risk, but further studies are required to confirm these results.

High levels of endogenous circulating oestrogens (Hankinson and Eliassen, 2007) and use of exogenous oestrogens (Beral, 2003) have both been associated with increased breast cancer risk. Isoflavones and lignans are plant compounds structurally similar to 17b-oestrodiol known as phyto-oestrogens, capable of oestrogen receptor binding (Kuiper et al, 1998;Mueller et al, 2004). Isoflavones are mostly found in soybean products, which are a staple of the Asian diet, whereas lignans are the principal group of phyto-oestrogens in Western diets. Lignans are more widespread in foods than isoflavones and are present in grain cereals, vegetables, seeds, tea and coffee (Mazur, 1998a;Mazur et al, 1998b). Microflora in the colon (Setchell et al, 1981) convert plant lignans into enterolignans, which are detectable in blood and urine. Their levels have been correlated with the amount of plant lignans ingested (Nesbitt et al, 1999).
In a recent meta-analysis, an inverse dose-response relationship was shown between breast cancer risk and soy-food intake in Asian, but not in Western women (Wu et al, 2008). Lignans have been shown to exhibit anti-carcinogenic properties (Wang et al, 1994;Prasad 2000;Bergman Jungeström et al, 2007), and it is hypothesised that exposure to high levels may be associated with a reduction in breast cancer risk. However, results from a number of studies in Western populations have been variable. The aim of our systematic review was to establish whether an association exists between lignan exposure and breast cancer risk, and to quantify the association through meta-analyses to inform evidence-based dietary guidelines.

MATERIALS AND METHODS
A systematic search of Ovid Medline (US National Library of Medicine, Bethesda, MD, USA), BIOSIS (Thompson Reuters, NY, USA) and EMBASE (Reed Elsevier PLC, Amsterdam, The Netherlands) databases for relevant studies published up to and including the date, 30 September 2008 was carried out. Relevant studies included at least one keyword or Medical Subject Heading from each of the following; (i) plant lignans (matairesinol, secoisolarisiresinol, pinoresinol and lariciresinol), (ii) enterolignans (enterolactone and enterodiol) and (iii) breast cancer. The search strategy excluded reviews, animal and cell culture studies but did not impose any language restrictions.
Abstracts and full texts, where required, were independently screened by two investigators to establish the suitability for inclusion. Studies had to be of case -control or cohort design, evaluating the risk of invasive breast cancer in relation to lignan exposure and reporting odds ratios (ORs) or relative risks, as well as 95% confidence intervals (95% CIs). Cited references were also reviewed for any studies that may have been missed in the database searches.
Eligible publications were then assessed independently by three reviewers. A structured form was used to extract information about the study, subjects' characteristics including menopausal status, confounding factors and results. Wherever multiple publications of the same study were available, the paper with the most complete set of data was chosen.
Studies were then categorised as those: (i) assessing total plant lignan intake or intake of individual plant lignans if the total was not measured; (ii) investigating exposure to enterolignans (enterolactone and enterodiol) by using values produced from food by in vitro fermentation models; and (iii) examining enterolactone levels in the blood (either plasma or serum). The blood levels of enterodiol were measured in a small number of studies (Piller et al, 2006b;Verheus et al, 2007;Ward et al, 2008) and were, therefore, not considered for analysis.
Separate meta-analyses were performed for each group of studies described in the Methods section using adjusted ORs or relative risks for the highest vs the lowest categories of exposure. If different levels of adjustment had been carried out, the results from the most fully adjusted model were used.
Random effects models were used to calculate pooled estimates, as we anticipated heterogeneity between observational studies (DerSimonian and Laird, 1986). Study-specific weights in the random effects model were calculated and scaled to percentages. The I 2 -statistic was used to test for heterogeneity (Higgins et al, 2003). Publication or selection bias was investigated by checking for asymmetry in funnel plots (Egger et al, 1997).
Analysis was repeated and sub-divided by menopausal status (pre-and post-menopausal). Statistical analyses were performed using the STATA version 9.2 software (Stata Corporation 2005, College Station, TX, USA).

RESULTS
Following screening of abstracts and full texts and grouping into categories, 27 of the 33 articles identified were selected for data extraction. Multiple publications were identified for a number of studies. Four articles (Grace et al, 2004;McCann et al, 2006;Thanos et al, 2006;Piller et al, 2006b) were excluded, as they were based on smaller subgroup analysis of their respective larger studies. The format of certain results prevented their use, but were provided by the authors in a suitable form and therefore included in this study. Overall, 23 publications were used, providing data for 6 cohort, 6 nested case -control and 10 case -control studies. Each article contributed data to one or more meta-analyses resulting in 12 articles on plant lignan intake (see Table 1), 5 on enterolignan exposure (see Table 2) and 9 on blood enterolactone levels (see Table 3). Details of the adjustments made in each study (the most fully adjusted model was used in the meta-analysis) are shown in Tables 1 -3. There was no association between plant lignan intake and risk when 11 studies were combined, although there was a slight protective effect. The risk in the highest intake group was 0.93 times (95% CI: 0.83 -1.03, P ¼ 0.15) that of the lowest intake group (see Figure 1). When studies were analysed by menopausal status, a statistically significant reduction in risk was seen with the highest intake category of plant lignans vs the lowest intake in postmenopausal women (7 studies, combined OR: 0.85, 95% CI: 0.78, 0.93, Po0.001), with little sign of between-study heterogeneity  (I 2 ¼ 0%, 95% CI: 0, 71, P ¼ 0.46) (see Figure 2). The same effect was not observed in pre-menopausal women (7 studies, combined OR: 0.97, 95% CI: 0.82, 1.15, P ¼ 0.73). The funnel plot of studies examining plant lignan intake and overall breast cancer risk showed symmetry, suggesting a lack of publication bias. There was a statistically significant inverse association between enterolignan exposure and overall risk (combined OR: 0.73, 95% CI: 0.57, 0.92, P ¼ 0.009) (Figure 3), although there was marked heterogeneity (I 2 ¼ 63%, 95% CI: 0.0, 88, P ¼ 0.04), but there was no association between exposure and risk by menopausal status (pre-menopausal breast cancer risk: 3 studies, combined OR: 0.67, 95% CI: 0.44 -1.02, P ¼ 0.06; post-menopausal: 2 studies, combined OR: 0.85, 95% CI: 0.72 -1.01, P ¼ 0.06).

DISCUSSION
This is the first systematic review and meta-analysis of exposure to lignans and breast cancer risk based on studies using dietary assessments and serum measurements. Although exposure can be assessed by urine analysis, few studies have used this methodology and therefore, these were not included (Ingram et al, 1997;den Tonkelaar et al, 2001;Dai et al, 2002). The results show that there was no association between plant lignan intake and overall risk, and this association was subjected to marked heterogeneity. Fink (2007) McCann (2004) Linseisen (2004) Dos Santos Silva (2004) Horn-Ross (2001  However in post-menopausal women, there is a small but significant reduction in risk and a reduction in heterogeneity. A significantly decreased risk with increasing enterolignan exposure was also found. However, there was significant heterogeneity between studies making it difficult to draw clear conclusions, and the effect did not persist when analyses were stratified by menopausal status, although the number of studies included in these stratified analyses was very small. Finally, there was no association between enterolactone concentrations in blood and overall risk, or when analysis was stratified by menopausal status. The protective action of plant lignans against breast cancer in post-menopausal, but not in pre-menopausal women, would suggest that lignan activity has a physiologic effect only at low oestradiol levels. One of the mechanisms of action may be greater sex hormone-binding globulin production and binding of free oestradiol (Adlercreutz et al, 1989(Adlercreutz et al, , 1992Zeleniuch-Jacquotte et al, 2004;Low et al, 2007). Binding of type II nuclear oestrogen receptor (Adlercreutz et al, 1992;Adlercreutz, 2007) and altering oestrogen synthesis within the breast cells and extragonadal sites, such as the adipose tissue, are other possible mechanisms (Adlercreutz et al, 1993;Saarinen et al, 2007). Enterolactone has been shown to decrease local oestrogen production by inhibiting 17-hydroxysteroid dehydrogenase type I and aromatase (Wang et al, 1994;Brooks and Thompson, 2005).
The apparent protective effect of dietary plant lignans in postmenopausal women is not supported by the findings from the meta-analysis of studies that measured the enterolactone levels in their blood. It would be expected that women consuming larger amounts of plant lignans would have a higher circulating concentration of enterolactone. There are a number of possible reasons for this disparity. Dietary intake of plant lignans was assessed on the basis of the subjects' self-reported dietary intake ranging from 6 months before study entry (Hedelin et al, 2008) to 3 years before breast cancer diagnosis, (dos Santos Silva et al, 2004) and thus, it reflects long-term intake. Enterolactone concentration that is measured in a single blood sample may be more indicative of recent dietary habits. There may also be a significant intra-individual variation in serum response to dietary * Test for heterogeneity Q = 8.2, df = 3, P = 0.043; I 2 =63% (95% CI; 0 -88%)  (2004) Olsen (2004) Kilkkinen (2004) Hulten ( Figure 4 Forest plot of highest vs lowest enterolactone levels in blood and breast cancer risk. intake of plant lignans (Hausner et al, 2004). For example, blood levels of enterolactone can be modulated by age, smoking, frequency of defecation, weight -obesity -body mass index and regular alcohol intake (Kilkkinen et al, 2001(Kilkkinen et al, , 2002Horner et al, 2002;Milder et al, 2007), and these factors could potentially differ by menopausal status (in particular, age and body mass index). As bacterial enzymes are involved in lignan metabolism, the use of antibiotics has also been shown to affect enterolactone serum concentration (Kilkkinen et al, 2002); antibiotic use was generally not controlled for in these studies.
It is also possible that the protective effect is caused directly by the plant lignans or chemicals within the metabolic pathway other than enterolactone, or even by a synergistic effect between plant lignans and enterolignans. However, other food constituents found to be associated with plant lignans may exert the effect. For example, a-linoleic acid, which is also thought to have anti-cancer effects (Thompson, 2003;Bougnoux and Chajes, 2003, p. 232), is found in very high levels in flaxseed, the richest source of plant lignans (Thompson et al, 1991).
Determining plant lignan intake has various limitations, which could lead to an over-or under-estimation of food content. Some food composition databases are incomplete in terms of not containing values for the more recently discovered plant lignans (e.g., medioresinol) or for the whole range of foods consumed by the study population. In addition, there are various analytical methods for determining food values ; hence, databases compiled from published values determined by different methodologies may contain inherent errors. It has also been shown that the amount of lignans in food can differ according to crop variety, location, year of harvest and processing (Thompson et al, 1997;Kuijsten et al, 2005). Dietary measurement error associated with FFQs (food frequency questionnaires) is also possible. FFQs that were used varied in length, ranging from 67 to 208 items. Only one study validated its FFQ specifically for plant lignan assessment (Torres-Sanchez et al, 2008), although a UK study used the combination of an FFQ and 24-h recalls to group participants into quartiles of intake (dos Santos Silva et al, 2004). In addition, the possibility of residual confounding cannot be ruled out.
Consumption of soy food, rich in isoflavones, has been shown to reduce breast cancer risk in Asian women but not in Western women (Wu et al, 2008), suggesting that ethnicity may play a role in this effect. It is not known whether there are differential physiologic effects of lignans in people of different races, although there is some evidence of variation in the urinary excretion of lignans between white, African American and Latino women (Horn-Ross et al, 1997). Of the 23 articles used for the metaanalyses, only 3 American studies provided complete data with regard to ethnicity (Horn-Ross et al, 2001McCann et al, 2004); hence, it was impossible perform sub-analyses for examining this.
In summary, the meta-analyses presented in this study, indicate that plant lignans and enterolignans are unlikely to significantly protect all women against breast cancer development. However, our results suggest that high plant lignan intake is associated with a 15% decreased risk in post-menopausal women, which is a small reduction that could be due to residual confounding. If real, the reason for the selective effect is not clear. Additional studies of the effect of lignan exposure on post-menopausal breast cancer risk are needed to confirm these findings before reassessing the current dietary guidelines.