Dietary isoflavone and the risk of colorectal adenoma: a case–control study in Japan

We conducted a case–control study in a Japanese population to investigate the association between dietary isoflavone intake and the risk of colorectal adenoma. Participants who underwent magnifying colonoscopy with dye spreading as part of a cancer screening programme responded to a self-administered questionnaire, which included lifestyle information and intake of 145 food items, before the colonoscopy. A total of 721 case and 697 control subjects were enrolled. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using unconditional logistic regression models. We found a significant inverse association between dietary isoflavone intake and the risk of colorectal adenoma in men and women combined. However, the inverse association was not linear; rather, all quartiles above the first showed a similar decrease in risk, with multivariable-adjusted ORs and 95% CIs compared with the lowest quartile of 0.77 (0.57–1.04), 0.76 (0.56–1.02) and 0.70 (0.51–0.96) in the second, third and highest quartiles, respectively (P for trend=0.03). Of interest, the observed association was more prominent in women than in men. The observed ceiling effect associated with higher isoflavone intake suggests that a lower intake of dietary isoflavone might be associated with an increased risk of colorectal adenoma in Japanese populations.

The incidence of colorectal cancer, a common malignancy in developed countries, is higher in men than in women (Parkin et al, 2002), a difference which has been related to sex hormones, in particular oestrogen. Although the function of oestrogen in the aetiology of colorectal cancer has not been established, an association is supported by several studies (MacLennan et al, 1995;Hébert-Croteau, 1998;Grodstein et al, 1999;Chlebowski et al, 2004;Linsalata et al, 2005).
Isoflavones such as genistein and daidzein are classified as phyto-oestrogens. Dietary phyto-oestrogens are plant-derived, non-steroidal compounds with weak oestrogen-like activity, occurring predominantly in soya beans and soya products. A better understanding of the association between dietary soya foods or isoflavone intake and colorectal cancer risk would be useful.
To date, seven studies have investigated the association between soya product consumption and the risk of colorectal cancer or adenoma (Hoshiyama et al, 1993;Witte et al, 1996;Marchand et al, 1997;Ho et al, 2006;Wakai et al, 2006;Oba et al, 2007;Akhter et al, 2008), of which five reported an inverse association. In contrast, only three studies have assessed the association of isoflavone intake with colorectal cancer risk (Cotterchio et al, 2006;Oba et al, 2007;Akhter et al, 2008), and we are unaware of any study of colorectal adenoma.
We have investigated the possible association between dietary isoflavone intake and the risk of colorectal adenoma, in a case -control study in a Japanese population, who were likely to consume relatively large amounts of isoflavones with substantial variation.

MATERIALS AND METHODS
The Colorectal Adenoma Study in Tokyo (CAST) consisted of examinees who underwent magnifying colonoscopy with dye spreading as a part of a cancer screening programme provided by the Research Center for Cancer Prevention andScreening, National Cancer Center, between February 2004 andFebruary 2005. Details of this programme have been provided elsewhere (Otani et al, 2006;Yamaji et al, 2009). Eligible examinees were men aged 50 -79 years or women aged 40 -79 years who underwent total colonoscopy from the anus to the caecum. Those with a history of any of the following diseases and conditions were considered ineligible: colorectal adenoma, malignant neoplasm, ulcerative colitis, Crohn's disease, familial adenomatous polyposis, carcinoid tumour or colectomy. All subjects gave written informed consent, and the study protocol was approved by the institutional review board of the National Cancer Center, Tokyo, Japan.
Among a consecutive series of 3212 examinees, 2234 met the eligibility criteria above. According to pit-pattern classification (Sano et al, 2005), 526 men and 256 women were determined to have at least one adenoma. Of the remaining 1452 examinees, 482 men and 721 women were also free from other benign lesions (e.g. hyperplastic polyps, inflammatory polyps and diverticulum) and were identified as potential controls. The potential control of 256 women was frequency matched to the female case in five age categories (40 -49, 50 -54, 55 -59, 60 -64 and X65 years old) and screening periods (first and second half). Because there were fewer potential male controls than male cases, all men of the potential controls were included in the study. Finally, the CAST consisted of 782 cases and 738 controls.
We conducted a self-administered questionnaire survey among participants before their screening examination. The questionnaire assessed demographic characteristics, personal and family medical history, use of drugs and supplements, anthropometric measurements, physical activity, history of smoking and alcohol drinking, reproductive factors (women only) and other lifestyle-related factors.
The self-administered questionnaire included a food frequency questionnaire (FFQ) that asked about 145 food and beverage items in terms of standard portion sizes and 9 frequency categories. The standard portion sizes of respective food items were specified in the FFQ and the amounts were determined into three categories of less than half, the same as, and more than 1.5 times the reference portion size. The frequency response choices for food items were less than once per month, 1 -3 times per month, 1 -2 times per week, 3 -4 times per week, 5 -6 times per week, once per day, 2 -3 times per day, 4 -6 times per day and X7 times per day. Soya food items included miso (fermented soya bean paste) soup, tofu (soya bean curd) in miso soup, tofu (boiled or cold) in other dishes, atsuage (fried tofu), koyadofu (freeze-dried tofu), aburaage (deepfried tofu), natto (fermented soya bean) and tonyu (soya milk made from soya bean as major ingredient). The frequency of miso soup consumption was divided into six categories of almost never, 1 -3 days per month, 1 -2 days per week, 3 -4 days per week, 5 -6 days per week and daily. Portion size was specified with one bowl of miso soup equalling 150 ml into the nine categories of o1 bowl per day, 1 bowl per day, 2 bowls per day, 3 bowls per day, 4 bowls per day, 5 bowls per day, 6 bowls per day, 7 -9 bowls per day and X10 bowls per day. Daily intake of genistein and daidzein was calculated using values in a specially developed food composition table for isoflavone in Japanese foods (Kimira et al, 1998;Arai et al, 2000). This FFQ was modified from an FFQ used in a previous population-based prospective study that had additional food items, the validity and reproducibility of which were evaluated previously (Tsubono et al, 1996;Tsugane et al, 2003). Validity was assessed among subsamples of that prospective study using 14-or 28-day dietary records. Spearman's correlation coefficients between the energy-adjusted intake of isoflavone from the questionnaire and dietary records were 0.60 (daidzein) and 0.59 (genistein). Moreover, Spearman's correlation coefficients for daidzein and genistein between energy-adjusted intake from the FFQ and those from serum concentrations were 0.26 and 0.22, respectively, and with those from creatinine-adjusted urinary excretion were 0.40 and 0.30, respectively (Yamamoto et al, 2001).

Statistical analysis
After the exclusion of 102 subjects who reported extreme total energy intake (upper and lower 2.5%) or submitted incomplete answers regarding dietary isoflavone intake and other confounding variables, 721 cases (492 men and 229 women) and 697 controls (460 men and 237 women) were entered into the analysis. Because dietary genistein and daidzein intakes were highly correlated (Spearman r ¼ 0.99), the sum of genistein and daidzein is provided here as representative for total isoflavones. Isoflavone intake was then adjusted for total energy intake using the residual method in the statistical model after log transformation (Kipnis et al, 1993;Brown et al, 1994;Willett et al, 1997). Participants were divided into quartiles of isoflavone intake based on the distribution of all control men and women. An unconditional logistic regression model was used to calculate odds ratios (ORs) and 95% confidence intervals (95% CIs) of colorectal adenoma for isoflavone intake using the lowest quartile of intake as the referent category. In addition, ORs and 95% CIs were calculated in subgroups stratified by adenoma site (proximal colon, distal colon and rectum), size (o5 and X5 mm) and number (1 and 41). The first statistical model was adjusted for age, sex and screening period, whereas the second model was further adjusted for family history of colorectal cancer (yes or no), cigarette smoking (never, ever o20, ever 21 -40, ever 440 pack-years), alcohol drinking (never, ever o150, ever 150 -299, ever X300 g per week ethanol), body mass index (o21, 21 -22, 23 -24, X25 kg m À2 ), physical activity (quartiles based on controls), supplement use (yes or no), non-steroidal antiinflammatory drug use (yes or no). Age at menarche (o12, 12 -13, X14 years), menopausal status (pre or post) and current use of female hormones (user or non-user) were also adjusted in women. Statistical significance of the interaction terms was evaluated by the likelihood ratio test. Tests for linear trend were estimated using the median value of isoflavone intake and treating this variable as continuous. All statistical tests were two tailed, and P-values o0.05 were considered statistically significant. All analyses were performed using Stata Statistics/Data Analysis Special Edition version 9.2 (2007; StataCorp LP, College Station, TX, USA).

RESULTS
The demographic characteristics and lifestyle factors according to case -control and sex had mentioned previously (Otani et al, 2006;Yamaji et al, 2009). Table 1 shows these factors of controls according to quartile categories of isoflavone intake. Median (interquartile range) isoflavone intake (mg per day) among the control group was 39.73 (24.77 -62.41). The proportion of women and supplement users, and mean intake of total isoflavone were higher in the highest quartile group than in the other groups. In contrast, the proportion of those using non-steroidal antiinflammatory drugs, ever smokers and mean alcohol consumption was higher in the lowest quartile group. Mean body mass index, total energy intake and age at menarche did not differ among the highest to lowest quartiles of isoflavone consumption. Table 2 shows the age, sex and screening-period-adjusted and multivariable-adjusted ORs and 95% CIs of colorectal adenoma for quartile categories of dietary isoflavone intake. Dietary isoflavone intake was associated with a reduced risk of colorectal adenoma in men and women combined. This inverse association was not linear; however, rather, a similar decrease was seen across all quartiles other than the first, with multivariable-adjusted ORs and 95% CIs in the second, third and highest compared with the lowest quartile of 0.77 (0.57 -1.04), 0.76 (0.56 -1.02) and 0.70 (0.51 -0.96), respectively (P for trend ¼ 0.03). Stratified analysis by sex showed a significant inverse association between isoflavone consumption and colorectal adenoma in women, and a non-significant inverse association in men. The multivariable-adjusted ORs (95% CIs) for women in increasing quartile intake categories were 0.53 (0.28 -0.98), 0.44 (0.24 -0.80) and 0.49 (0.27-0.90), respectively (P for trend ¼ 0.03).
We repeated the above analyses after excluding those who had changed their dietary habits in the past 5 years (n ¼ 393 subjects), primarily on the basis that the change might have been related to the presence of adenoma. Results were essentially the same as those above (data not shown). We also investigated the association between colorectal adenoma for quartile categories of dietary isoflavone intake after further adjustment of fibre, folate, calcium and vitamin D in the multivariate model and similar decreased risk of association was observed (data not shown). We further categorised the subjects into octiles. As in the quartile analysis, however, the inverse association was not linear in the octile groups; rather, the third to highest octiles showed a decrease in risk, with the pattern more apparent in women, in whom multivariable-adjusted ORs (95% CIs) in increasing octile categories   (40 -49, 50 -54, 55 -59, 60 -64, X65 years), sex and screening period (first or second). b Adjusted for age (40 -49, 50 -54, 55 -59, 60 -64, X65 years), sex, screening period (first or second), family history of colorectal cancer (yes or no), cigarette smoking (never, ever p20, ever 21 -40, ever 440 pack-years), alcohol consumption (never, ever o150, ever 150 -299, ever X300 g per week ethanol), body mass index (o21, 21 -22, 23 -24, X25 kg m À2 ), physical activity (o31.50, 31.50 to o34.55, X34.55 MET h per day based on controls), supplement use (yes or no) and non-steroidal anti-inflammatory drug use (yes or no). c Adjusted for age (50 -54, 55 -59, 60 -64, X65 years) and screening period (first or second). d Adjusted for age (50 -54, 55 -59, 60 -64, X65 years), screening period (first or second), family history of colorectal cancer (yes or no), cigarette smoking (never, ever p20, ever 21 -40, ever 440 pack-years), alcohol consumption (never, ever o150, ever 150 -299, ever X300 g per week ethanol), body mass index (o21, 21 -22, 23 -24, X25 kg m À2 ), physical activity (o31.50, 31.50 to o34.55, X34.55 MET h per day based on controls), supplement use (yes or no) and nonsteroidal anti-inflammatory drug use (yes or no). e Adjusted for age (40 -49, 50 -54, 55 -59, 60 -64, X65 years) and screening period (first or second). f Adjusted for age (40 -49, 50 -54, 55 -59, 60 -64, X65 years), screening period (first or second), family history of colorectal cancer (yes or no), cigarette smoking (never, ever p20, ever 21 -40, ever 440 pack-years), alcohol consumption (never, ever o150, ever 150 -299, ever X300 g per week ethanol), body mass index (o21, 21 -22, 23 -24, X25 kg m À2 ), physical activity (o31.50, 31.50 to o34.55, X34.55 MET h per day based on controls), supplement use (yes or no), non-steroidal anti-inflammatory drug use (yes or no), age at menarche (o12, 12 -13, X14 years), menopausal status (pre or post) and current use of female hormones (user or non-user). Stratified analysis according to site, size and number of colorectal adenomas showed no difference in the association with isoflavone intake by adenoma characteristics in overall men and women (Table 3). We further tested for effect modification between isoflavone intake and other covariates (age, sex, cigarette smoking, alcohol drinking and body mass index) in colorectal adenoma through the addition of multiplicative interaction terms into the model, but found no statistically significant P-value for interaction (data not shown).

DISCUSSION
In this colonoscopy-based case -control study, we found a significant inverse association between dietary isoflavone intake and colorectal adenoma risk in a Japanese population. The inverse association was not linear; with a ceiling effect observed with higher intake, which implied conversely that a lower intake of dietary isoflavone was related to an increased risk.
To our knowledge, this is the first study of this subject. Because adenomas are much more common than cancers, we may have greater power to detect the isoflavone -adenoma association. To date, three studies have examined the association between dietary isoflavone intake and the risk of colon (Oba et al, 2007) or colorectal cancer (Cotterchio et al, 2006;Akhter et al, 2008), two in Japanese (Oba et al, 2007;Akhter et al, 2008) and one in a Caucasian population (Cotterchio et al, 2006). Although the former two showed no clear association, the latter demonstrated a linear inverse association between dietary isoflavone intake and colorectal cancer risk. Of importance, the Caucasian population consumed a relatively low level of isoflavones, with a cut-off point for the highest tertile of 1.097 vs 17.22 mg per day for the lowest octile in our study population; it therefore investigated the effect of lower intake levels, which we could not observe in detail in our study, and supported our finding that a lower intake of dietary isoflavone was related to an increase in risk.
Isoflavones may help prevent the development of colorectal neoplasm through their oestrogenic effect. In fact, several in vitro and in vivo experimental studies have suggested that phytooestrogens protect against colorectal cancer (Spector et al, 2003). Phyto-oestrogens distinctly interact with oestrogen receptors (ERs) in the normal human colon. Of the ER-a and ER-b oestrogen receptors, phyto-oestrogens bind with high affinity to ER-b, and consistently mediate signals to inhibit abnormal cellular proliferation in colon tumour cell lines by blocking tyrosine protein kinases, aromatase and DNA topoisomerases. Phyto-oestrogens also possess antioxidant activity, and inhibit cell-cycle progression and angiogenesis in endothelial cells (Lechner et al, 2005).
We found a stronger inverse association in women than in men, although effect modification by sex was not statistically significant. Nonetheless, this suggested difference by sex is of particular interest, because isoflavones may have a more important function in the prevention of colorectal adenoma in women than in men. Of the five of seven epidemiological studies of the association between soya product intake and colorectal neoplasm, which provided results for men and women separately (Marchand et al, 1997;Ho et al, 2006;Wakai et al, 2006;Oba et al, 2007;Akhter et al, 2008), three reported that the inverse association was more prominent among women (Marchand et al, 1997;Ho et al, 2006;Oba et al, 2007), whereas none reported that it was stronger among men. This sex difference in the effect of plant-derived oestrogen (isoflavones) on colorectal neoplasm warrants further investigation.
Our study has several methodological advantages over previous studies. First, we used a validated FFQ and calculated isoflavone intake from a variety of dietary soya sources. Many different foods contribute to isoflavone intake, and no single soya food category predominates. Moreover, our subjects had relatively large isoflavone intake with substantial variation, making them suitable for the investigation of this association. Second, subjects answered the questionnaire before diagnosis, likely minimising the possibility of recall bias.  (40 -49, 50 -54, 55 -59, 60 -64, X65 years), sex, screening period (first or second), family history of colorectal cancer (yes or no), cigarette smoking (never, ever p20, ever 21 -40, ever 440 pack-years), alcohol consumption (never, ever o150, ever 150 -299, ever X300 g per week ethanol), body mass index (o21, 21 -22, 23 -24, X25 kg/m 2 ), physical activity (o31.50, 31.50 to o34.55, X34.55 MET h per day based on controls), supplement use (yes or no) and non-steroidal anti-inflammatory drug use (yes or no).

Dietary isoflavone and colorectal adenoma M Akhter et al
Several limitations also warrant mention. First, although all examinees underwent total colonoscopy, colorectal adenoma cases were diagnosed according to pit-pattern classification and were not confirmed histologically. Given an overall accuracy of pitpattern diagnosis of approximately 90% vs data from our institute showing 95% accuracy, our analysis may have included a few falsepositive cases (Sano et al, 2005). Magnifying chromoendoscopy is nevertheless a feasible and efficient method of determining neoplastic lesions such as adenoma and the validity of its interand intraobserver consistency has been confirmed (Haenszel et al, 1973;Fu et al, 2004). Second, the presence of adenoma might have been associated with changes in dietary habits; however, the sensitivity analysis revealed that such reverse association was less likely. Finally, isoflavone intake and other variables were based on self-administered questionnaires; some misclassifications of subjects were inevitable.
In summary, we found an inverse association between dietary isoflavone intake and the risk of colorectal adenoma in a Japanese population, which was not linear, but rather showed a ceiling effect associated with higher isoflavone intake; it was more prominent in women than in men, possibly reflecting an oestrogenic effect.