Intake of Specific Nonfermented Soy Foods May Be Inversely Associated with Risk of Distal Gastric Cancer in a Chinese Population^,,

Abstract

Because the association between soy consumption and gastric cancer is inconsistent, we evaluated the putative preventive effect of soy food on gastric cancer risk in the Shanghai Women’s and Men’s Health Studies, comprising a total of 128,687 participants. Intake of nonfermented soy foods was estimated using 2 validated food-frequency questionnaires. HRs were calculated with 95% CIs for intake amounts of total nonfermented soy food intake, soy protein, and isoflavones as well as individual soy food groups using Cox proportional hazards regression. A total of 493 distal gastric cancer cases were identified by 2010. Although all risk estimates for summary measures of soy food intake above the lowest quartile (quartile 1) were suggestive of a protective effect, no statistically significant associations with risk of distal gastric cancer were found. Among the separate soy food groups, significant reductions in risk of distal gastric cancer by increasing intake of tofu were found in men in quartile 2 (HR: 0.59; 95% CI: 0.40, 0.86), quartile 3 (HR: 0.62; 95% CI: 0.44, 0.88), and quartile 4 (HR: 0.64; 95% CI: 0.42, 0.99), resulting in a significant trend (P-trend = 0.02). Dry bean intake was also inversely associated with decreased risk of gastric cancer, but in postmenopausal women only [quartile 2 (HR: 0.54; 95% CI: 0.30, 0.96); quartile 3 (HR: 0.90; 95% CI: 0.64, 1.27); and quartile 4 (HR: 0.63; 95% CI: 0.43, 0.91)], resulting in a significant trend (P-trend = 0.03). Overall, our study found no statistically significant association between nonfermented soy food intake and distal gastric cancer risk, though the data supported the hypothesis that tofu may protect against distal gastric cancer in men and dry bean consumption may decrease the risk of gastric cancer in postmenopausal women.

Introduction

Gastric cancer, the fourth most common cancer globally with about 1 million new cases annually (1), is the most frequent cancer in Japan and Korea, and the second most common in China, in both men and women (2). Furthermore, it has been ranked the second cause of cancer-related death in the world; ∼738,000 people died from gastric cancer in 2008 (1). However, gastric cancer incidence is relatively low in most Western countries, including the United States and the United Kingdom. This geographic variation in incidence has generally been explained by 3 interpopulation differences, including differences in dietary patterns (3, 4), subtype prevalence of Helicobactor pylori infection (5, 6), and genetic susceptibility (7), and by their interactions (8, 9).

Dietary habits have been considered an important factor in carcinogenesis and have been particularly emphasized in the development of gastric cancer (10, 11). The high incidence of gastric cancer in Japan, Korea, and China has often been explained by their distinct dietary habits, including eating high-salt food and nutritional sources of nitrosamine alongside a high prevalence of H. pylori (12, 13). In the case of soy foods, Japanese and Korean populations are generally known to consume more fermented soy foods, such as fermented soy paste and fermented soybeans, rather than fresh nonfermented soy foods (14–16). Although the cancer-preventive effect of soy intake has been widely studied in hormone-sensitive cancers, information is inconsistent about the role of soy foods for cancer development in organs not considered to be hormone related, such as the stomach (17, 18). It has been suggested that estrogen plays a protective role in the development of gastric cancer (19). For example, exposure to tamoxifen, an antagonist of estrogen receptor, after breast cancer or endometrial cancer treatment has been suggested as a risk factor of gastric cancer (20, 21). Although the protective mechanism of soy foods for gastric cancer is more frequently explained by anti-inflammatory effects or an antioxidative mechanism rather than the hormone-related effects (22), the potential protective pathway in gastric tissue remains to be clarified (23–25). In addition, the inconsistencies found between soy food intake and risk of gastric cancer in previous results may also be explained by confounders, or effect modifiers, such as ethnicity, sex, menopausal status or history of hormone replacement therapy in women, fermented status of soy foods, and possibly H. pylori status or genetic susceptibility (25, 26).

The misclassification bias caused by the low validity of exposure measurement for dietary soy intake and the insufficient statistical power by small sample size might also bias the study results toward the null association. A large-scale prospective study in a population with a high incidence of gastric cancer could minimize the impact of exposure misclassification and provide greater statistical power to control for related factors. Therefore, we analyzed the data from 2 large-scale, population-based, prospective studies, the Shanghai Women’s Health Study (SWHS) and the Shanghai Men’s Health Study (SMHS), to investigate whether nonfermented soy food intake is associated with a reduced risk of gastric cancer in an urban Chinese population.

Participants and Methods

Study design and participants.

The baseline survey of the SWHS was administered from 1996 to 2000, and the SMHS from 2002 to 2006, both in urban Shanghai, China. A total of 74,941 women aged 40–70 y were recruited from 80,891 eligible women (participation rate, 92.6%), and a total of 61,482 men aged 40–74 y were recruited from 83,125 eligible men (participation rate, 74.0%) during the baseline examination. The SWHS and SMHS used a similar standard protocol, including in-person interviews, to obtain information on demographic characteristics, dietary habits, and other health-related factors. The details on the study design of the SWHS (27) and SMHS (28) have been described previously. The studies were approved by the Institutional Review Boards of Vanderbilt University, the Shanghai Cancer Institute, and the National Cancer Institute, and written informed consent was obtained from all participants.

Dietary assessment.

The validation of the comprehensive FFQs in both cohorts has been described in detail elsewhere (29, 30). Briefly, the SMHS FFQ (81 items) was modified from the SWHS FFQ (77 items), with the addition of the items of garland chrysanthemum, shepherd’s purse, clover, amaranth, pea shoots, and pig’s ham hock and the removal of sweet potatoes. The SWHS FFQ combined fresh soy beans, fresh peas, and fresh broad beans into 1 item, but the SMHS FFQ listed these items in separate questions. Nonfresh soy bean curd, including fried tofu, vegetarian chicken, and bean curd cake, was combined into 1 item in both cohorts, labeled “other soyfood products.” For each food item or group of foods, we collected data for the frequency of consumption (i.e., daily, weekly, monthly, annually, and never) and the amount consumed in liang (Chinese unit, 1 liang = 50 g) per unit of time. The daily intakes of total energy, soy protein, and isoflavones were calculated by multiplying the amount of each food consumed by the nutrient content per gram of the food as obtained from the Chinese Food Consumption Tables (31). The daily intake of the dried weight of 6 individual items [soy milk, fresh soybean, dried soybean, dried or pressed bean curd (tofu), bean sprouts, and other soyfood products (i.e., cooked tofu, which includes fried tofu, vegetarian chicken, and bean curd cake)] were calculated by multiplying the amount of each soy food consumed by [1 − (percentage of water content of the food item/100)] (32) and were combined to compute the daily nonfermented soy food intake amount (g/d) of each individual. Water content of each soy food was determined based on the Chinese Food Consumption Tables (soy milk, fresh soybean, dried soybean, tofu, bean sprout, and other soyfood products were 96.4%, 29.4%, 10.2%, 85.6%, 88.8%, and 60.1% water, respectively). For the present analyses, we averaged the amounts of these nonfermented food items derived from the baseline FFQ with those from the first follow-up FFQ, which was conducted 2–3 y after the baseline survey (follow-up rate was 91.3% and 81.6% in the SWHS and SMHS, respectively).

Cancer ascertainment.

We initially identified newly diagnosed gastric cancer cases by data linkage with the Shanghai Cancer Registry, a population-based cancer registry established in 1963 and by in-person follow-up surveys every 2–3 years. Completeness of case ascertainment by the Shanghai Cancer Registry is very high because of a low out-migration rate. In the present study, new possible cancer cases extracted by data linkage with the registry database were rechecked to eliminate any false-positive matches by reviewing medical records from the diagnostic hospitals and home visiting. Death cases were ascertained by data linkage to death certificate registries and confirmation through home visits. Among the 678 gastric cancer cases (SWHS, n = 354; SMHS, n = 324) identified until the end of 2010, 615 cases (SWHS, n = 326; SMHS, n = 289) were distal gastric cancer cases, including 87 cases of unknown site, defined as noncardia cancer, which was coded as 151.1–151.9, based on the WHO International Classification of Diseases (9th version).

Statistical analysis.

Among the total participants of the SWHS and SMHS, we excluded the 1,669 individuals with gastrectomy history before the baseline questionnaire (SWHS, n = 346; SMHS, n = 1,323) and 1,495 female individuals and 14 male individuals who had reported a previous cancer diagnosis in the baseline questionnaire or were identified by cancer registry data. In addition, in the final analysis we also excluded 227 individuals who had an estimated extreme total energy intake of <500 or >4000 kcal/d in baseline FFQ or in follow-up FFQ 1 (SWHS, n = 75; SMHS, n = 152), 421 individuals with a follow-up period of <1 y (SWHS, n = 153; SMHS, n = 268), 3894 individuals who had reported a history of peptic ulcer medication in the baseline questionnaire (SWHS, n = 2,411; SMHS, n = 1,483), and 17 individuals with missing essential information such as status of current smoking or menopause (SWHS, n = 15; SMHS, n = 2). After excluding 185 distal gastric cancer cases and 7552 noncancer cases for the reasons just stated, a total of 128,687 individuals (SWHS, n = 70,446; SMHS, n = 58,241), including 493 newly incident distal gastric cancer cases (SWHS, n = 282; SMHS, n = 211), were included for the present analyses.

All participantss were divided into 4 quartiles according to their sex and median nonfermented soy food intake, including soy milk, tofu, other soyfood products, dry bean, fresh bean, and bean sprout (Table 1), from the baseline examination and follow-up survey 1. The lowest quartile (quartile 1) was used as the reference. To calculate HRs and 95% CIs for the association between nonfermented soy intake and incident distal gastric cancer under control of potential confounders, we used Cox proportional hazards regression with age as the time scale, with α = 0.05. In the fully adjusted model, we selected the variables that were independently associated with distal gastric cancer risk in the age-adjusted model and that we a priori believed to be associated with soy food intake as confounders. These covariates included smoking status [never or ever (smoked at least 1 cigarette/d for >6 mo)]; smoking amount (pack-years); drinking status [never or ever (drank at least 3 times/wk for >6 mo)]; BMI (as a continuous variable); physical activity (metabolic equivalent hours per week per year); presence of family history of gastric cancer; past history of chronic gastritis; born in Shanghai and house income as social indicators; median intakes of total energy (kcal/d), sodium, red meat, fruit except watermelon, and all vegetables from the 2 FFQs; and menopausal status (in women). Continuous variables are presented as means ± SDs. We also examined the association of soy food and gastric cancer risk among women in models stratified by menopausal status. We used the median soy food intake value of each quartile to test for a linear trend across the intake amount. We have applied the fractional polynomials approach to find the best fitting model (33). In our data, any polynomial model does not provide a better fit compared with a simple linear model at the 5% concentration. Other possible risk factors for gastric cancer, such as use of aspirin, education level, hormone replacement therapy in women, and dietary salt intake amount, did not alter the results after adjustment, thus they were excluded from the final model. The end of the study period was defined as the first of the following outcomes: first cancer diagnosis, any cause of death, or date of last follow-up. All analyses were carried out with SAS version 9.3 (SAS Institute).

TABLE 1.

Baseline characteristics comparing distal gastric cancer case individuals and noncase individuals as well as individuals in quartiles 1–4 of daily nonfermented soy food intake levels in the SWHS and SMHS¹

	Case (n = 493)	Noncase (n = 128,194)	Quartile 1 (n = 32,188)	Quartile 2 (n = 32,161)	Quartile 3 (n = 32,154)	Quartile 4 (n = 32,184)
Women, n	282	70,164	17,622	17,605	17,600	17,619
Median nonfermented soy food intake,2 g/d			11.20	19.11	27.27	42.32
Age at baseline, y	57.9 ± 9.1	52.4 ± 9.0*	51.5 ± 9.1	51.7 ± 8.9	52.5 ± 8.9	53.9 ± 9.1
Menopause, %	73.8	48.3*	43.7	45.4	49.1	55.5
BMI, kg/m	24.8 ± 3.5	24.0 ± 3.4*	23.7 ± 3.4	23.9 ± 3.4	24.1 ± 3.4	24.4 ± 3.5
Ever drinker,3 %	2.84	2.27	1.87	2.28	2.24	2.69
Ever smoker,4 %	4.96	2.68*	3.04	2.39	2.55	2.76
Smoking, pack-years	11.2 ± 9.52	11.7 ± 13.7	12.0 ± 13.3	11.6 ± 13.5	11.9 ± 15.4	11.0 ± 12.6
Regular exercise,5 %	45.7	34.7*	27.8	32.6	36.7	41.9
MET, h/wk	0.9 ± 1.8	0.7 ± 1.6*	0.5 ± 1.2	0.6 ± 1.4	0.7 ± 1.6	0.9 ± 1.8
Total energy, kcal/d	1610 ± 355	1640 ± 350	1490 ± 308	1600 ± 310	1680 ± 320	1800 ± 370
Sodium, mg/d	312 ± 129	347 ± 138*	283 ± 117	328 ± 119	363 ± 127	413 ± 151
Red meat, mg/d	43.3 ± 30.4	48.0 ± 30.1*	42.2 ± 26.9	47.1 ± 28.0	49.8 ± 29.6	53.0 ± 34.3
Fruit (excluding watermelon), mg/d	115 ± 93	135 ± 96*	109 ± 86	127 ± 88	140 ± 93	162 ± 108
Vegetables, mg/d	278 ± 150	303 ± 152*	217 ± 107	269 ± 116	320 ± 132	407 ± 176
Family total cancer history, %	29.4	26.4	24.7	26.8	26.9	27.0
Family gastric cancer history, %	8.9	5.7*	5.6	5.7	5.9	5.8
Gastritis history, %	16.3	17.4	17.9	17.4	17.1	17.2
Aspirin or analgesic use,6 %	1.8	2.0	1.5	1.7	2.0	2.8
Born in urban Shanghai, %	52.5	69.2*	70.8	69.6	69.0	67.0
Men, n	211	58,030	14,566	14,556	14,554	14,565
Median nonfermented soy food intake,2 g/d			17.17	28.06	38.08	56.56
Age at baseline, y	63.3 ± 9.4	55.2 ± 9.7*	55.1 ± 10.0	54.8 ± 9.5	55.2 ± 9.6	55.7 ± 9.6
BMI, kg/m2	23.8 ± 3.0	23.78 ± 3.1	23.6 ± 3.1	23.7 ± 3.0	23.8 ± 3.0	24.0 ± 3.0
Ever drinker,3 %	37.9	33.7	29.2	31.8	34.2	39.8
Ever smoker,4 %	67.3	69.3	69.8	69.0	69.8	68.6
Smoking, pack-years	28.1 ± 17.4	24.3 ± 16.0*	24.8 ± 16.7	23.9 ± 15.5	23.9 ± 15.7	24.5 ± 16.2
Regular exercise,5 %	46.5	35.4*	31.5	33.9	37.2	39.0
MET, h/wk	1.4 ± 2.1	1.0 ± 2.0*	0.8 ± 1.8	0.9 ± 1.9	1.0 ± 2.0	1.2 ± 2.3
Total energy, kcal/d	1870 ± 400	1910 ± 420	1760 ± 400	1870 ± 380	1950 ± 390	2080 ± 430
Sodium, mg/d	379 ± 150	389 ± 145	325 ± 130	368 ± 124	402 ± 129	462 ± 159
Red meat, mg/d	60.5 ± 36.2	62.1 ± 37.5	54.9 ± 35.6	60.2 ± 34.0	64.1 ± 35.7	69.3 ± 42.6
Fruit (excluding watermelon), mg/d	67.0 ± 68.5	70.5 ± 63.7	56.5 ± 56.2	64.8 ± 56.3	73.4 ± 62.6	87.2 ± 73.9
Vegetables, mg/d	339 ± 182	361 ± 181	262 ± 141	320 ± 134	377 ± 151	485 ± 208
Family total cancer history, %	28.4	28.2	27.7	28.3	28.3	28.7
Family gastric cancer history, %	6.6	6.3	6.1	6.3	6.5	6.5
Gastritis history, %	21.3	13.9*	15.0	14.1	13.0	13.6
Aspirin or analgesic use,6 %	10.9	7.1*	7.4	6.3	7.0	7.6
Born in urban Shanghai, %	54.5	74.5*	74.2	75.2	74.8	73.5

¹Values are means ± SDs or percentages. *Different from cases, P < 0.05. MET, metabolic equivalent task; SMHS, Shanghai Men’s Health Study; SWHS, Shanghai Women’s Health Study.

²Dry weights of soy milk, tofu, fresh bean, dry bean, and soy bean sprout.

³Drank at least 3 times/wk for >6 mo.

⁴Smoked at least 1 cigarette/d for >6 mo.

⁵Exercised at least 1 time/wk for >3 mo continuously for the past 5 y.

⁶Took analgesics or aspirin-based medicine at least 3 times/wk for >2 mo continuously.

Results

The following are the general characteristics of distal gastric cancer case individuals compared with noncase individuals at risk of gastric cancer: distal gastric cancer case individuals were more likely to be older, regular exercisers, and born outside of urban Shanghai than noncase individuals among men and women (Table 1). However, these differences between case and noncase individuals disappeared in both sexes when adjusted by age at baseline examination. Compared with individuals in the lowest quartile (quartile 1) of daily nonfermented soy food intake, individuals in the highest quartile (quartile 4) were more likely to be older and postmenopausal, have a higher BMI, have higher average levels of physical activity, be regular aspirin users, be ever drinkers, be regular exercisers, and be more likely to have been born outside of urban Shanghai, after adjusting for age among both men and women. Daily intake of total energy, sodium, red meat, fruit excluding watermelon, and all vegetables also increased by increasing quartile of soy food intake (Table 1).

Overall, although all risk estimates for each summary measure of soy food intake above quartile 1 were suggestive of a protective effect, no statistically significant associations with risk of distal gastric cancer were found. These null associations did not differ according to sex. Soy protein and isoflavones also showed similar null associations with distal gastric cancer risk in all individuals and both sexes (Supplemental Table 1).

When examined in separate soy food categories, no association between each soy food category and risk of distal gastric cancer was found overall (Table 2). When stratified by sex, men (but not women) in quartiles 2, 3, and 4 of tofu intake, compared with those in quartile 1 of tofu intake, had decreased risks of distal gastric cancer (HR: 0.59; 95% CI: 0.40, 0.86; P = 0.01; HR: 0.62; 95% CI: 0.44, 0.88; P = 0.01; and HR: 0.64; 95% CI: 0.42, 0.99, P = 0.05 respectively), resulting in a significant trend (P-trend = 0.02). In men’s quartile 1, tofu intake was <3.1 oz (87.9 g)/d of dry weight of tofu, or ∼0.62 oz (17.6 g)/d of tofu. No associations between soy milk, dry bean, fresh bean, bean sprout, or other soyfood products and risk of distal gastric cancer were found among men or women (Table 2).

TABLE 2.

HRs and 95% CIs of distal gastric cancer risk for daily intake of each group of nonfermented soy food in quartiles 1–4 in the baseline examination and follow-up survey 1 of SWHS and SMHS¹

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P-trend
Soy milk2
Men, g/d	0 (0.0)	0.30 (0.1–0.9)	1.93 (1.0–3.9)	5.79 (>3.9)
Women, g/d	0 (0.0)	0.59 (0.1–1.3)	2.57 (1.4–4.5)	6.43 (>4.5)
HR (95% CI)3
All	1.00	0.84 (0.65, 1.08)	0.83 (0.66, 1.06)	0.87 (0.69, 1.11)	0.22
Men	1.00	1.07 (0.73, 1.57)	0.94 (0.65, 1.38)	1.05 (0.73, 1.50)	0.95
Women	1.00	0.69 (0.50, 0.97)	0.75 (0.55, 1.02)	0.75 (0.54, 1.04)	0.08
Tofu2
Men, g/d	1.93 (<3.1)	4.10 (3.2–5.3)	6.28 (5.4–8.4)	10.46 (>8.4)
Women, g/d	1.53 (<2.5)	3.14 (2.5–4.1)	5.63 (4.2–6.8)	9.42 (>6.8)
HR (95% CI)3
All	1.00	0.71 (0.55, 0.91)	0.89 (0.70, 1.12)	0.72 (0.55, 0.95)	0.08
Men	1.00	0.59 (0.40, 0.86)	0.62 (0.44, 0.88)	0.64 (0.42, 0.99)	0.02
Women	1.00	0.82 (0.58, 1.15)	1.18 (0.86, 1.63)	0.82 (0.57, 1.17)	0.73
Dry bean2
Men, g/d	0 (0.0)	0.31 (0.1–0.5)	0.61 (0.6–1.2)	2.15 (>1.2)
Women, g/d	0 (0.0)	0.24 (0.1–0.3)	0.61 (0.4–0.9)	1.72 (>0.9)
HR (95% CI)3
All	1.00	0.82 (0.61, 1.11)	0.90 (0.72, 1.14)	0.82 (0.65, 1.03)	0.10
Men	1.00	1.13 (0.76, 1.69)	0.89 (0.61, 1.29)	0.93 (0.65, 1.32)	0.53
Women	1.00	0.57 (0.36, 0.92)	0.91 (0.68, 1.22)	0.74 (0.54, 1.01)	0.10
Fresh bean2
Men, g/d	4.72 (<7.9)	10.62 (7.9–13.4)	16.71(13.4–21.3)	29.45 (>21.3)
Women, g/d	1.86 (<3.2)	4.96 (3.3–6.7)	8.97 (6.8–11.9)	17.73 (>11.9)
HR (95% CI)3
All	1.00	0.90 (0.70, 1.15)	0.81 (0.62, 1.05)	1.03 (0.79, 1.34)	0.93
Men	1.00	0.70 (0.47, 1.04)	0.80 (0.54, 1.18)	1.04 (0.70, 1.53)	0.79
Women	1.00	1.07 (0.78, 1.48)	0.83(0.58, 1.18)	1.04 (0.73, 1.48)	0.78
Bean sprout2
Men, g/d	0 (0.0)	0.08 (0.1–0.21)	0.38 (0.22–0.58)	1.00 (>0.58)
Women, g/d	0 (<0.1)	0.08 (0.1–0.18)	0.30 (0.19–0.55)	1.01 (>0.55)
HR (95% CI)3
All	1.00	1.00 (0.79, 1.27)	0.99 (0.77, 1.26)	0.98 (0.76, 1.27)	0.87
Men	1.00	1.05 (0.72, 1.53)	1.10 (0.76, 1.61)	1.08 (0.73, 1.60)	0.64
Women	1.00	0.96 (0.70, 1.31)	0.91 (0.65, 1.26)	0.91 (0.65, 1.29)	0.54
Other soyfood products2
Men, g/d	2.85 (>4.8)	6.13 (4.8–8.5)	9.98 (8.6–12.8)	15.90 (>12.8)
Women, g/d	1.64 (>3.0)	4.28 (3.0–5.7)	7.79 (5.7–10.0)	14.26 (>10.0)
HR (95% CI)3
All	1.00	0.85 (0.67, 1.08)	0.68 (0.53, 0.89)	0.93 (0.71, 1.21)	0.21
Men	1.00	0.85 (0.60, 1.21)	0.69 (0.46, 1.03)	0.77 (0.51, 1.16)	0.11
Women	1.00	0.85 (0.62, 1.17)	0.69 (0.48, 0.97)	1.06 (0.75, 1.52)	0.77

¹Values are medians (ranges) of intake per quartile in terms of dry weight of the soy product(s). SMHS, Shanghai Men’s Health Study; SWHS, Shanghai Women’s Health Study.

²Median intake of each quartile used as continuous variables in a Cox proportional hazards model.

³Data are adjusted for age, BMI, metabolic equivalents hours per week per year, chronic gastritis history, family gastric cancer history, born in urban Shanghai, family income, ever drink, ever smoke, and smoking amounts at baseline examinations as well as for median intakes of total calories, red meat, vegetables, sodium, fruit (excluding watermelon), and sex (for the models including both sexes, only). Quartile 1 value 1.00 is the reference.

When we stratified by menopausal status of women, there was a suggestion of a decreased risk of gastric cancer for postmenopausal women only whose dry bean intake was greater than that of quartile 1, that is, women in quartiles 2, 3, and 4 (HR: 0.54; 95% CI: 0.30, 0.96; HR: 0.90; 95% CI: 0.64, 1.27; and HR: 0.63; 95% CI: 0.43, 0.91, respectively), resulting in a significant trend (P-trend = 0.03). Otherwise, no differences were found from the models stratified by menopausal status (data not shown).

Discussion

Overall, we found no statistically significant association between total nonfermented soy food intake and the risk of distal cancer in the SWHS and SMHS cohorts. However, among the separate soy food groups, 2 statistically significant associations were found. Increasing tofu intake was associated with decreasing risk of gastric cancer in men, and increasing dry bean intake was associated with decreasing risk of gastric cancer in postmenopausal women. Our study findings were consistent with results from the previous epidemiological reports on the relation between soy consumption and gastric cancer risk.

Specifically, 2 recent meta-analyses investigated the preventive effects of soy consumption and risk of gastric cancer. In the most recent meta-analysis, higher intake of nonfermented soy foods was associated with a statistically significant decreased risk of gastric cancer (OR: 0.64; 95% CI: 0.54, 0.77), whereas higher intake of fermented soy foods was associated with a statistically significant increased risk (OR: 1.22; 95% CI: 1.02, 1.44) (17). This meta-analysis included 18 articles, comprising 6 cohort studies and 12 case-control studies, but all were from studies of Korean or Japanese populations (22). A Chinese meta-analysis, including 10 cohort studies, 16 case-control studies, and 2 cross-sectional studies, also reported an inverse association between the highest consumption of soybean products and risk of gastric cancer (OR: 0.58; 95% CI: 0.52, 0.65) (34). However, although these meta-analyses and individual case-control studies have reported associations between soy intake and gastric cancer, few large prospective studies have been restricted to more homogenous categories for a more precise analysis of these associations, as has the present study, through the exclusion of fermented soy foods and a focus on distal gastric cancer only.

Recently a Japanese prospective cohort study, which included a total of 1249 gastric cancer cases, found no evidence for a gastric cancer–preventive effect of isoflavones or soy food in either sex. This study however did report an increasing trend of gastric cancer risk with increasing amounts of isoflavones among women who had used hormonal therapy (18). This finding was consistent with the results from a previous study that the phytoestrogenic effect may be altered by an individual’s estrogen environment (35). In our study, hormonal therapy was not a factor because only 2% of the women in our study used hormonal therapy. Furthermore, we found no significant interaction between menopausal status and total nonfermented soy food intake in relation to distal gastric cancer (P-trend = 0.28), although there was a suggestion that dry bean intake reduces risk for postmenopausal women only.

Previous in vivo studies have suggested that isoflavones, which are abundant in soy food, suppress carcinogenesis in gastric cells (23, 36), but the potential mechanism remains unclear. Isoflavones have a similar structure to 17β-estradiol and are able to suppress the effects of estrogen by competitively binding to the β-estrogen receptor. However, phytoestrogens have been found to have both estrogenic and antiestrogenic effects, depending on the specific tissue and the individual’s concentration of circulating estrogens (37). Another biologically plausible pathway that is able to explain a potentially protective effective of isoflavones on gastric cancer risk is through anti-inflammatory and antioxidative effects. Studies have reported on the relation of soybean curd and H. pylori infection (38), interaction of soy bean intake and interleukin-10 genetic variants on gastric cancer risk (24), and genistein inhibition on the growth of H. pylori (39), because soy products can enhance the immune function of gastric cells (40). Genistein may induce apoptosis in cancer cell lines, arrest the gastric cancer cell proliferation cycle at the G2/M phase (40), and play a role as a tyrosine kinase inhibitor that inhibits the proliferation or angiogenesis of gastric epithelial cells (36).

In our study, tofu and dry bean intakes were inversely associated with the risk of distal gastric cancer in men and postmenopausal women, respectively. The inverse association between tofu intake and gastric cancer was supported by a previous review of in vitro and in vivo studies (41) and several epidemiological studies (15, 42, 43). A Korean case-control study, including 213 cases and an equal number of controls, also reported an inverse association between tofu intake and risk of gastric cancer (OR: 0.2; 95% CI: 0.1,0.8) (43); however, in a prospective study of 38,576 Japanese atomic-bomb survivors, no significant inverse association between tofu and gastric cancer was found (15). A Japanese cross-sectional study reported an inverse association between tofu and H. pylori infection (P-trend = 0.01), although no clear explanation was suggested (38). A Japanese prognosis study of gastric cancer reported that frequent intake of tofu decreases the HR for gastric cancer death (HR: 0.65; 95% CI: 0.42, 0.99) (42). Soy also contains a variety of bioactive components, including saponins, trypsin inhibitors, phytic acid, protease inhibitors, and isoflavones (44). The contents of the components in tofu are reduced during the processing of soy beans to tofu, though quantification studies of isoflavone isomers have reported that tofu contains similar amounts of isoflavones as soy milk and miso and represents the major source of isoflavones among soy foods (45, 46).

When we stratified by menopausal status in women, dry bean intake was also inversely associated with the risk of distal gastric cancer. Dry beans are a major dietary source of fiber and flavonols, which may play a protective role against cancers, including gastric cancer (47, 48). Another possible mechanism is that dry bean consumption was associated with dose-response reduction in plasma glucose concentration, and the low glycemic load may reduce the risk of gastric cancer (47, 49, 50). Dietary dry beans are also rich in other bioactive constituents that may have anticarcinogenic properties, including saponins, protease inhibitor, inositol hexaphosphate, γ-tocoperol, and phytosterols (51), and appeared to modify cell proliferation through apoptosis (52).To further explore our study results, we have done sensitivity analyses to compare the main results with subanalyses excluding individuals diagnosed within 1, 2, and 3 y of the baseline survey, and major results were consistent with the primary analyses (data not shown).

Several characteristics in study design distinguished this study from previous investigations. We chose to exclude the fermented soy foods elements from our calculation of soy food intake because, in general, common fermented soy foods, such as soup or stew, contain a large amount of salt and lose their beneficial effects during fermentation or storage (17). Preserved or fermented soy foods, such as miso soup in Japan and soybean paste in Korea, contain high concentrations of N-nitroso compounds (25), which are potential carcinogens for many primary sites. Moreover, to reduce the misclassification of soy intake group, we estimated the exposure level using the mean of 2 FFQs (53). The correlation coefficient between nonfermented soy food intakes derived from the 2 FFQs was 0.35, and the agreement rate of the quartile group of nonfermented soy intake between the 2 FFQs was 38.0%. Because the water content of soy foods is relatively high and varies widely (from 96.4% in soy milk to 10.2% in dry beans), we also calculated dry weights of each soy food to ascertain the specific effects of each soy component. In addition, we restricted final outcomes to gastric cancer developed in the noncardia region. Distal gastric cancer is localized in the noncardia region and differs from cardia gastric cancer, which is localized to the gastroesophageal junction, in etiology and epidemiological characteristics. Another strength of our study was that our study population in is was well suited to the study of soy and gastric cancer because of its high incidence of gastric cancer (1) and dietary habits with frequent consumption of soy foods (54). Study participants were middle-aged urban Shanghai residents, who have been shown to have high participation rates (92.6% and 74.1% in SWHS and SMHS, respectively) and follow-up rates (91.3% and 81.6% in SWHS and SMHS, respectively).

This study had some limitations. First, H. pylori status was not available for inclusion in this study. H. pylori infection is the strongest risk factor for gastric cancer; however, it is unlikely that inclusion of H. pylori status of study participants would have affected our results critically because H. pylori prevalence is very high in this population (92% and 94% in SWHS and SMHS participants, respectively) (55, 56). Second, our data might not have been sufficient to detect associations in some subgroups because of the short follow-up period, particularly for the men (median 5.5 y), and the relatively low interindividual variability of soy consumption among the study populations. Third, possible dietary measurement errors may have caused modest dietary associations because any dietary changes occurring before the baseline examination and after the first follow-up were not accounted for. Fourth, despite adjusting for several conventional risk factors and selected dietary factors, the likelihood of residual confounding by unadjusted dietary factors that correlate with soy food consumption could not be discounted. We also were unable to report on fermented soy food intake, although consumption of these foods is less common in the Chinese population than in other Asian populations. Finally, our results did not suggest that a linear association exists between soy food intake and gastric cancer risk in this population, as we hypothesized. Instead, our results suggested a dichotomous association, whereby individuals in the lowest quartile of intake (quartile 1) may be the only ones at increased risk in this population of generally high soy food intake. Other unmeasured factors associated with being in quartile 1 may also have caused this association, resulting in residual confounding. Overall, we think these limitations would most likely apply generally to both the cases and controls in this prospective study, thus attenuating any associations investigated toward the null.

In summary, the present study found no association between nonfermented soy food intake and risk of distal gastric cancer overall, though an inverse association was found for tofu in men only and dry bean consumption may decrease the risk of gastric cancer in postmenopausal women. These findings did not support the hypothesis that general consumption of nonfermented soy food plays a protective role in gastric cancer etiology.