SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Soy food is known to contribute greatly to a reduction in the risk of gastric cancer (GC). However, both Japanese and Korean populations have high incidence rates of GC despite the consumption of a wide variety of soy foods. One primary reason is that they consume fermented rather than non-fermented soy foods. In order to assess the varying effects of fermented and non-fermented soy intake on GC risk in these populations, we conducted a meta-analysis of published reports. Twenty studies assessing the effect of the consumption of fermented soy food on GC risk were included, and 17 studies assessing the effect of the consumption of non-fermented soy food on GC risk were included. We found that a high intake of fermented soy foods was significantly associated with an increased risk of GC (odds ratio [OR] = 1.22, 95% confidence interval [CI] = 1.02–1.44, I2 = 71.48), whereas an increased intake of non-fermented soy foods was significantly associated with a decreased risk of GC (overall summary OR = 0.64, 95% CI = 0.54–0.77, I2 = 64.27). These findings show that a high level of consumption of non-fermented soy foods, rather than fermented soy foods, is important in reducing GC risk. (Cancer Sci 2011; 102: 231–244)

Gastric cancer (GC) is the most common cancer in Japan and Korea and the second leading cause of death from cancer globally, although the incidence and mortality have been declining over the years.(1) Dietary factors are known to play an important role in the development of GC.(2) Among dietary factors, soy has been of considerable interest in the etiology of GC.(3) It is widely known that soy foods may help reduce the risk of developing GC. This might be due to the fact that soy foods are good sources of isoflavones, which are antioxidants known to reduce the risk of GC.(3) However, despite the fact that Japanese and Korean populations generally have a high intake of soy foods, they also have a higher risk of GC than other populations, including those in the USA and Europe.

This might be explained by the fact that Japanese and Korean populations consume more fermented soy foods than non-fermented ones. Common fermented soy foods, which generally contain a high content of salt, include soy paste and fermented soybeans.(3) The most common non-fermented soy foods include soymilk, tofu, soybeans and soy nuts. There is a contradictory relationship between the intake of soy food and GC; GC risk increases with the intake of fermented soy foods and decreases with the intake of non-fermented soy foods.(4) Fermented soy may offer health benefits due to the fermentation process. However, it may have adverse effects on GC risk due to high levels of nitrate or nitrite, large amounts of salt, and the loss of key nutrients under acidic and oxygenic conditions.

An extensive meta-analysis of the relationships between fermented soy foods and GC risk has not been conducted. Thus, we carried out a meta-analysis of the relationships between the consumption of fermented and non-fermented soy foods and GC risk in Japanese and Korean populations.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Selection of studies for the meta-analysis.  We searched the reference lists of publications concerning diet and GC conducted in Japanese and Korean populations. The search engines used for this study included PubMed, KoreaMed and Ichushi. We used the following keywords: “gastric cancer” or “stomach cancer”, “soy” or “fermented soy”, and “Japan” or “Korea”. We also searched the references cited in the articles and included published works written in Japanese, Korean and English.

Inclusion/exclusion criteria.  The inclusion/exclusion criteria for this meta-analysis were as follows:

  • 1
     Only results that specified the food item studied was “fermented soy”, “non-fermented soy” or were included. The term “fermented soy” included miso (soup) or soybean paste (soup or stew), while “non-fermented soy” included tofu, bean curds or non-fermented soy (bean) products.
  • 2
     Only subjects who were Japanese or Korean, including migrants, were included.
  • 3
     Cohort or case–control studies were included. Reviews and meta-analyses were excluded.
  • 4
     We excluded case–control studies that presented mortality rather than GC incidence.
  • 5
     The studies that included adjusted 95% confidence intervals (CI) and either a relative risk (RR) or odds ratio (OR) were included for meta-analysis. We excluded studies that did not present an adjusted 95% CI or those that showed regression coefficients.
  • 6
     When multiple studies were published on the same subject population, we included only the most recent study.

Data abstraction.  Two reviewers independently examined the studies for inclusion in the meta-analysis.

Disagreements between the reviewers were resolved by consensus. We collected the following information from each study: study design, author, publication year, nation, study period, study subjects (type and sources, definition and numbers of subjects), category of food intake from the lowest to the highest, RR/OR and 95% CI, P for trend, and confounding variables.

Statistical analysis.  In order to adjust for the confounding factors and to include those studies with missing values (cross-tabulation) in the tables, we used an unconditional logistic regression analysis to compute the RR or OR with a 95% CI. We assessed statistical heterogeneity across the studies by calculating the variation between studies (τ2) from the Q statistic. Based on these results for heterogeneity, we used either a fixed-effect or random-effect model to compute the summary OR and 95% CI. For assessing publication bias, asymmetry was tested using Begg’s funnel plot. A P-value <0.05 was considered statistically significant. We carried out all analyses using STATA 10 software (STATA, College Station, TX, USA).

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

We identified a total of 69 articles in the initial computerized search of published work. After screening the articles according to title and abstract, 43 articles (two review papers, one meta-analysis, 27 experimental studies or clinical trials, two studies of populations from countries other than Japan or Korea, four studies with the same subject population, and seven studies on other foods, soy foods or non-dietary factors) were excluded. Twenty-two articles (nine cohort studies(2,5–12) and 13 case–control studies(13–25)) assessing the effect of consumption of fermented soy foods on GC risk were included in the meta-analysis. Eighteen articles (six cohort studies(2,9–11,26) and 12 case–control studies(15,16,18–20,22–25,27–29)) assessing the effect of consumption of non-fermented soy food on GC risk were included in the meta-analysis.

Table 1 lists the included cohort and case–control studies on the effect of fermented soy products on GC risk. Similarly, Table 2 lists the included cohort and case–control studies on the effect of non-fermented soy products on GC risk. Most studies adjusted for confounding factors, including age and sex. We obtained statistically significant results when testing for heterogeneity between studies of fermented soy foods (overall summary Q = 98.18 with 28 degrees of freedom (df), P < 0.001; cohort studies Q = 27.36 with 11 df, P = 0.004; case-control studies Q = 66.51 with 16 df, P < 0.001) and non-fermented soy foods (overall summary Q = 61.58 with 22 df, < 0.001; cohort studies Q = 18.07 with eight df, = 0.021; case-control studies Q = 30.08 with 13 df, = 0.005). Therefore, we selected a random-effect model to produce the summary statistics. The results of the meta-analysis of the relationships between GC risk and non-fermented soy food intake and fermented soy food intake are summarized in Figures 1 and 2, respectively. A high intake of non-fermented soy foods was significantly associated with a decreased risk of GC (overall summary OR = 0.64, 95% CI = 0.54–0.77, I2 = 64.27; cohort studies OR = 0.83, 95% CI = 0.60–1.13, I2 = 55.73; case–control studies OR = 0.57, 95% CI = 0.46–0.71, I2 = 56.78), whereas a high intake of fermented soy foods was significantly associated with an increased risk of GC (overall summary OR = 1.22, 95% CI = 1.02–1.44, I2 = 71.48; cohort studies OR = 1.12, 95% CI = 0.88–1.41, I2 = 59.79; case–control studies OR = 1.34, 95% CI = 1.04–1.73, I2 = 75.94). In order to assess for publication bias, Begg’s funnel plot(30,31) for the assessment of publication bias is presented in Figure 3 for fermented soy food and in Figure 4 for non-fermented soy food. The funnel plots did not detect a publication bias in the meta-analyses of the effect of fermented (Z = 0.11, = 0.910) or non-fermented soy foods (Z = 0.95, = 0.342) on GC risk.

Table 1.   Intake of fermented soy products and gastric cancer risk in Korean or Japanese populations
Author, year, nationStudy periodStudy subjectsCategoryRR (95% CI)P for trendConfounding variables considered
Source of subjectsNo. subjectsPrimary outcome studiedNo. incidence or deaths of GC
  1. †Compared with subjects without atrophic gastritis. §LSS (Life Span Study cohort) includes atomic bomb survivors and unexposed subjects in Hiroshima and Nagasaki. ¶Compared with the general population control. BMI, body mass index; CI, confidence interval; GC, gastric cancer; JACC, The Japan Collaborative Cohort Study for Evaluation of Cancer Risk; JPHC cohort, The Japan Public Health Center-based prospective study; NA, not available; NS, not significant; RR, relative risk.

Cohort studies
 Kato I et al. 1992, Japan(5)1985–1991Population-based, Aichi Prefecture9753Death57 (35 men, 22 women)Miso soup0.845Adjusted for age and sex
 <1 cup/day1.00
 1 cup/day0.88 (0.42–1.82)
 ≥2 cups/day1.04 (0.48–2.25)
 Inoue M et al. 1996, Japan(6)1985–1995Subjects who underwent gastroscopic examination, Aichi Prefecture5373 (2552 men, 2821 women)Incidence69 (51 men, 18 women)Soybean paste soupNAAdjusted for age and sex
 Rarely1.00
 Occasionally 4.68 (0.88–24.99)
 Daily3.62 (0.79–16.70)†
 Galanis DJ et al. 1998, Japan(7)1975–1994Population-based, Japanese–American residents of Hawaii11 907 (5610 men, 6297 women)Incidence108 (64 men, 44 women)Miso soupNAAdjusted for age, years of education, Japanese place of birth and gender
 None1.0
 1 or more times/week1.2 (0.8–1.8)
 Ngoan LT et al. 2002, Japan(2)1986–1989Population-based, Fukuoka Prefecture13 250 (5917 men, 7333 women)Death116 (77 men, 39 women)Miso soup≥0.05Adjusted for age, sex, smoking, processed meat, liver, cooking or salad oil, and suimono
 2–4 times/week1.00
 Once/day0.9 (0.5–1.8)
 Twice/day or more1.7 (0.6–4.5)
 Tsugane S et al. 2004, Japan(8)1990–2001JPHC Cohort, Iwate, Akita, Nagano, Okinawa Prefectures18 684 menIncidence358 menMiso soup0.002Adjusted for age, smoking, fruit intake and non-green vegetable intake
 Not daily1.00
 1 cup/day1.76 (1.16–2.67)
 2 cups/day1.99 (1.39–2.87)
 3 or more cups/day1.75 (1.22–2.51)
20 381 womenIncidence128 womenMiso soup0.65
 Not daily1.00
 1 cup/day1.15 (0.66–2.01)
 2 cups/day0.82 (0.49–1.38)
 3 or more cups/day1.11 (0.67–1.84)
 Sauvaget C et al. 2005, Japan(9)1980–1999Participants in LSS cohort, Hiroshima and Nagasaki Prefectures38 576 (14 885 men, 23 691 women)Incidence1270 (719 men, 551 women)Miso soup>0.05Adjusted for age, sex, city, radiation dose, sex-specific smoking habit and education
 <2 times/week1.00
 2–4 times/week1.01 (0.88–1.16)
 5 times or more/week1.01 (0.88–1.16)
 Tokui N et al. 2005, Japan(10)1988–1999Participants in JACC study (45 areas)110 792Death574 menMiso soupAdjusted for age
 None1.000.36
 Several times/ month1.42 (0.79–2.56)
 Several times/ week1.50 (0.85–2.65)
 Every day1.44 (0.86–2.42)
Death285 womenMiso soup
 None1.000.19
 Several times/ month1.38 (0.69–2.74)
 Several times/ week1.18 (0.60–2.31)
 Every day1.46 (0.81–2.61)
 Nagata C et al. 2002, Japan(11)1992–1999Community-based, Takayama13 880 menDeath81 menFermented soy products0.76Adjusted for age, total energy, smoking and BMI at age approximately 21 years
 Low1.00
 Middle1.17 (0.69–2.01)
 High0.91 (0.53–1.58)
16 424 womenDeath40 womenFermented soy products0.1Adjusted for age, total energy, marital status, age at menarche, BMI at age approximately 21 years
 Low1.00
 Middle0.78 (0.38–1.61)
 High0.52 (0.24–1.13)
 Khan MMH et al. 2004, Japan(12) Population-based, Hokkaido1524 menDeath36 menMiso soup<0.05Adjusted for age and smoking
 ≤Several times/ month1.0
 ≥Several times/ week0.2 (0.1–0.8)
Case–control studies
 Kono S et al. 1988, Japan(13)1979–1982Hospital-based, Karatsu Stomach InstituteCases: newly diagnosed as having gastric cancer at the institute General population controls: random sampling from the computerized file of residents General population controls: 148 men, 130 womenMiso soup≥0.05Matched for sex, year of birth
 None or 1–3 times/month1.00
 1–3 times/week0.49 (0.19–1.29)
 ≥1 time/day0.52 (0.22–1.26)
Hospital controls: patients without gastric cancer Hospital control: 1171 men, 1474 womenMiso soup≥0.05
 None or 1–3 times/month1.00
 1–3 times/week0.62 (0.30–1.33)
 ≥1 time/day0.71 (0.39–1.42)
 Kato I et al. 1990, Japan(14)1985–1989Hospital-based Aichi Cancer Center HospitalCases: histologically confirmed cases who received gastroscopic examination 1247 menSoybean paste soupNAAdjusted for age, residence
 <1–2/month1.00
 2–3 times/week0.91 (0.50–1.65)
 Daily1.04 (0.60–1.81)
Controls: patients with normal gastric mucosa1767 womenSoybean paste soupNA
 <1–2/month1.00
 2–3 times/week0.79 (0.40–1.54)
 Daily0.83 (0.44–1.58)
 Hoshiyama Y et al. 1992, Japan(15)1984–1990Hospital-based, Saitama Cancer Center HospitalCases: newly histologically confirmed cases who received gastroscopic examination in the institute. Population controls: residents in the study area 206 men, 88 womenMiso soup<0.01Matched for sex, age, administrative division Adjusted for sex, age, administrative division, smoking status
 <1 cup/day1.00
 1 cup/day1.4 (0.8–2.5)
 ≥2 cups/day2.0 (1.2–3.5)
Hospital controls: inpatients without cancer 104 men, 98 womenMiso soup0.08Adjusted for sex, age (three categories), area (three categories), smoking status
 <1 cup/day1.0
 1 cup/day1.5 (0.8–2.9)
 ≥2 cups/day1.8 (1.0–3.4)
 Hoshiyama Y. and Sasaba T 1992, Japan(16)1984–1990Hospital-based, Saitama Cancer Center HospitalCases: single and multiple gastric cancer cases, newly diagnosed and of adenocarcinoma type in the institute Population controls: residents in the study area 483 menMiso soup0.00Matched for age, sex. Adjusted for age, smoking status
 ≤1 cup/day1.0
 2 cups/day1.3 (0.9–1.9)
 ≥3 cups/day2.3 (1.5–3.7)
Miso soup0.01Adjusted for age, smoking status
 ≤1 cup/day1.0
 2 cups/day1.6 (0.7–3.5)
 ≥3 cups/day3.4 (1.4–8.3)
 Iwasaki J et al. 1992, Japan(17)1980–1986Population-based, mountain villages in the Shizuoka Prefecture and farming-fishing villages in the Chiba PrefectureCases: gastric cancer death Controls: general population control 83 menMiso soupNAMatched for sex, age (±2 years), district and year of death (±6 years)
 Not frequently1.000
 Frequently1.235 (0.594–2.590)
 Inoue M et al. 1994, Japan (18)1988–1991Hospital-based, Aichi Cancer Center HospitalCases: histologically confirmed incident case Controls: outpatients of the same hospital 420 men, 248 womenMiso soup<0.05Matched for sex, age (±2 years), time of first hospital visit (±2 months) Adjusted for sex
 3–4 times or more/week vs less0.95 (0.75–1.20)
 Watabe K et al. 1998, Japan(19)1996–1997Hospital-based, Sapporo Medical UniversityCases: histologically confirmed Control: randomly selected from the telephone book 242 cases and 484 controlsMiso soupNAMatched for sex, age (±3 years), registered residence
 ≥3/day2.14 (1.43–3.15)
 Ito LS et al. 2003, Japan(20)1988–1998Hospital-based, Aichi Cancer Center HospitalCases: histologically confirmed cases Controls: cancer-free first visit outpatients at the center 36 490 womenSoybean paste soupNSAdjusted for age, year and season of first hospital visit, smoking habit, family history of gastric cancer
 Almost never1.00
 Occasionally0.62 (0.42–0.92)
 3–4 times/week0.69 (0.47–1.01)
 Every day0.74 (0.47–1.14)
 Machida–Montani A 2004, Japan(21)1998–2002Hospital-based, four hospitals in Nagano prefectureCases: non-cardia gastric cancer. Control: participants of a health check-up program at the same hospitals 235Miso soup0.04Matched by age (within 3 years), sex, residential area Adjusted for H. pylori infection, smoking status, JA  membership, family history of gastric cancer, total vegetable intake, total fruits intake, salt intake and total energy intake
 <3 cups/day1.0
 3 cups/day1.8 (1.0–3.3)
 ≥4 cups/day2.1 (0.9–5.1)
 Lee JK et al. 1995, Korea(22)1990–1991Hospital-based, Hanyang University Hospital, Asan Medical Center, SeoulCases: histologically confirmed gastric cancer. Hospital controls: orthopedic, urological, ophthalmological, gynecological wards, ambulatory patients 132 men, 81 womenSoybean paste stew<0.001Matched by age (within 2 years), sex. Adjusted for age, sex, education, economic status and residence
 1–3/months1.0
 ≥2–3/week5.5 (2.5–12.1)
 Park HS et al. 1998, Korea(23)1996Hospital-based, Korea Cancer Center Hospital, SeoulCases: histologically confirmed gastric cancer. Hospital controls: patients without gastric cancer 87 men, 39 womenSoybean pasteNSMatched by age (within 5 years), sex, admission date. Adjusted for age, sex, education, economic status, residence
 ≤1/week 1.0
 ≥2/week0.8 (0.5–1.2)
 Zhang YW et al. 2009, Korea(24)2000–2005Hospital-based, Chungbuk National University Hospital, Eulji University HospitalCases: histologically confirmed gastric cancer. Hospital controls: patients without gastric cancer 315 men, 156 womenSoybean paste<0.05Matched by age (within 3 years), sex. Adjusted for age, sex and energy intake
 Low1.0
 High1.63 (1.24–2.14)
 Kim HJ et al. 2002, Korea(25)1997–1998Hospital-based, Hanyang and Hallim University HospitalCases: histologically confirmed gastric cancer. Hospital controls: patients without gastric cancer 93 men, 43 womenSoybean pasteNAMatched by age (within 2 years), sex, during same period in the same hospital. Adjusted for sex, age, socioeconomic status, family history, refrigerator use
 Low1.00
 Medium0.86 (0.48–1.55)
 High0.84 (0.43–1.63)
Table 2.   Intake of non-fermented soy products and gastric cancer risk in Korean or Japanese populations
Author, year, nationStudy periodStudy subjectsCategoryOR (95% CI)P for trendConfounding variables considered
Source of subjectsNo. subjectsPrimary outcome studiedNo. incidence or deaths of GC
  1. †Compared with subjects without atrophic gastritis. §LSS (Life Span Study cohort) includes atomic bomb survivors and unexposed subjects in Hiroshima and Nagasaki. ¶Compared with the general population control.BMI, body mass index; CI, confidence interval; GC, gastric cancer; JACC, The Japan Collaborative Cohort Study for Evaluation of Cancer Risk; NA, not available; NS, not significant; OR, odds ratio.

Cohort studies
 Ngoan LT et al. 2002, Japan(2)1986–1989Population-based Fukuoka Prefecture13 250 (5917 men, 7333 women)Death116 (77 men, 37 women)Tofu≥0.05Adjusted for age, sex, smoking, processed meat, liver, cooking or salad oil, and suimono
 ≤2–4 times/ month1.0
 2–4 times/week0.6 (0.3–1.1)
 Once/day or more0.4 (0.2–0.9)
 Sauvaget C et al. 2005, Japan(9)1980–1999Participants in LSS cohort§, Hiroshima and Nagasaki Prefectures38 576 (14 885 men, 23 691 women)Incidence1270 (719 men, 551 women)Tofu0.492Adjusted for age, sex, city, radiation dose, sex-specific smoking habit and education
 <2 times/week1.00
 2–4 times/week0.94 (0.83–1.06)
 5 times or more/week1.01 (0.85–1.20)
 Tokui N et al. 2005, Japan(10)1988–1999Participants in JACC study (45 areas)110 792Death574 menBean curds0.97Adjusted for age
 1–2/month or less1.0
 1–2/week1.12 (0.76–1.65)
 3–4/week1.09 (0.74–1.60)
 1+/day1.07 (0.73–1.58)
Death285 womenBean curds0.25
 1–2/month or less1.00
 1–2/week1.24 (0.65–2.36)
 3–4/week1.38 (0.74–2.59)
 1+/day1.41 (0.75–2.64)
 Nagata C et al. 2002, Japan(11)1992–1999Community-based, Takayama study13 880 menDeath81 menNon-fermented soy products0.0006Adjusted for age, total energy, smoking and BMI at age approximately 21 years
 Low1.00
 Middle 0.73 (0.44–1.23)
 High0.46 (0.26–0.81)
16 424 womenDeath40 womenNon-fermented soy products0.09Adjusted for age, total energy, marital status, age at menarche, BMI at age approximately 21 years
 Low1.00
 Middle 0.97 (0.48–1.96)
 High0.48 (0.21–1.10)
 Kurosawa M et al. 2006, Japan(26)1989–1999Population-based, three municipalities of Higashi-Yamanashi Country8035 (3652 men, 4383 women)Death76 (54 men, 22 women)Bean and bean products0.66Multivariate, adjusted by backward elimination method
 Low (2–7/week)1.00
 Intermediate (8/week)1.79 (0.81–3.96)
 High (9–10/week)0.88 (0.31–2.56)
 Khan MMH et al. 2004, Japan(12) Population-based, Hokkaido1524 menDeath36 menTofu>0.05Adjusted for age and smoking
 ≤Several times/month1.0
 ≥Several times/week3.6 (0.5–26.0)
1634 womenDeath15 womenTofu>0.05Adjusted for age, health status, health education, health screening and smoking
 ≤Several times/month1.0
 ≥Several times/week1.1 (0.1–8.5)
Case–control studies
 Hoshiyama Y et al. 1992, Japan(15)1984–1990Hospital-based, Saitama Cancer Center HospitalCases: newly histologically confirmed cases who received gastroscopic examination in the institute. Population controls: residents in the study area 206 men, 88 womenSoybean products (except miso soup)<0.01Matched for sex, age, administrative division. Adjusted for sex, age, administrative division, smoking status
 ≤4/week1.0
 5–7/week 0.5 (0.4–0.8)
 ≥8/week0.6 (0.4–1.0)
Hospital controls: inpatients without cancer 104 men, 98 womenSoybean products (except miso soup)0.36Adjusted for sex, age (three categories), area (three categories), smoking status
 ≤4/week1.0
 5–7/week0.9 (0.6–1.3)
 ≥8/week0.8 (0.4–1.4)
 Hoshiyama Y. and Sasaba T 1992, Japan(16)1984–1990Hospital-based, Saitama Cancer Center hospitalCases: single and multiple gastric cancer cases newly diagnosed and of adenocarcinoma type in the institute. Population controls: residents in the study area 483 menSoybean products (except miso soup)0.00Match for age, sex. Adjusted for age, smoking status
 ≤4/week1.0
 5–7/week0.6 (0.4–0.8)
 ≥8/week0.6 (0.3–0.9)
Soybean products (except miso soup)0.28Adjusted for age, smoking status
 ≤4/week1.0
 5–7/week0.8 (0.4–1.7)
 ≥8/week0.5 (0.2–1.7)
 Inoue M et al. 1994, Japan(18)1988–1991Hospital-based, Aichi Cancer Center HospitalCases: histologically confirmed incident case. Controls: outpatients of the same hospital 420 men, 248 womenBean curdsNAMatched for sex, age (±2 years), time of first hospital visit (±2 months). Adjusted for sex
 3–4 times or more/week vs less0.89 (0.71–1.11)
 Watabe K et al. 1998, Japan(19)1996–1997Hospital-based, Sapporo Medical UniversityCases: histologically confirmed. Controls: randomly selected from the telephone book 242 cases and 484 controlsTofuNAMatched for sex, age (±3 years), registered residence
 ≥1/week0.69 (0.45–1.07)
 Ito LS et al. 2003, Japan(20)1988–1998Hospital-based, Aichi Cancer Center HospitalCases: histologically confirmed cases Controls: cancer-free first visit outpatients at the center 36 490 womenTofuNSAdjusted for age, year and season of first hospital visit, smoking habit, family history of gastric cancer
 <1 times/week1.00
 1–2 times/week0.84 (0.63–1.11)
 3–4 times/week0.81 (0.61–1.09)
 ≥5 times/week0.86 (0.63–1.17)
 Hamada GS et al. 2002, Japan(27)1991–1994Hospital-based, Japanese–Brazilians, Sao PauloCases: histologically confirmed cases Controls: preferably identified from among the inpatients of the same hospital 120 men, 72 womenSoybean products0.33Adjusted for country of birth
 <1 day/week 1.0
 1–2 day/week0.6 (0.3–1.2)
 3–4 day/week0.7 (0.3–1.5)
 Daily0.4 (0.1–1.1)
 Lee SA et al. 2002, Korea(28)2000Hospital-based, Asan Medical Center, SeoulCases: newly diagnosed as having gastric cancer at the institute Hospital controls: patients without gastric cancer 116 men, 83 womenSoybean curds<0.001Adjusted for age, sex, and H. pylori infection
 <1/month1.0
 ≥1/month0.3 (0.1–0.5)
 Lee JK et al. 1995, Korea(22)1990–1991Hospital-based, Hanyang University Hospital, Asan Medical Center, SeoulCases: histologically confirmed gastric cancer Hospital controls: orthopedic, urological, ophthalmological, gynecological wards, ambulatory patients 132 men, 81 womenSoybean curds<0.01Matched by age (within 2 years), sex. Adjusted for age, sex, education, economic status and residence
 None or 4–5/year1.0
 1–3/months0.5 (0.1–1.5)
 ≥2–3/week0.2 (0.1–0.8)
 Park HS et al. 1998, Korea(23)1996Hospital-based, Korea Cancer Center Hospital, SeoulCases: histologically confirmed gastric cancer Hospital controls: patients without gastric cancer 87 men, 39 womenSoybean curds<0.05Matched by age (within 5 years), sex, admission date Adjusted for age, sex, education, economic status, residence
 ≤2–3/month1.0
 1–2/week0.7 (0.4–1.2)
 ≥3/week0.3 (0.2–0.7)
 Zhang YW et al. 2009, Korea(24)2000–2005Hospital-based, Chungbuk National University Hospital, Eulji University HospitalCases: histologically confirmed gastric cancer Hospital controls: patients without gastric cancer 315 men, 156 womenNon-fermented soybean foods<0.05Matched by age (within 3 years), sex. Adjusted for age, sex, and energy intake
 Low1.0
 High0.57 (0.43–0.75)
 Kim HJ et al. 2002, Korea(25)1997–1998Hospital-based, Hanyang and Hallim University HospitalCases: histologically confirmed gastric cancer. Hospital controls: patients without gastric cancer 93 men, 43 womenSoybean productsNAMatched by age (within 2 years), sex, during same period in the same hospital Adjusted for sex, age, socioeconomic status, family history, refrigerator use
 Low1.00
 Medium0.81 (0.46–1.42)
 High0.35 (0.16–0.75)
 Ko KP et al. 2009, Korea(29)1993–2004Community-based, Korean multi-center cancer cohortCases: Linkages to the Korean cancer registry and National death certificate databases Community control 236 men, 100 womenSoybeanNAMatched by age (within 5 years), sex, residence area, year of recruitment; 1 case : 4 control
 ≥3–4 times/week1.00
 <3–4 times/ month0.54 (0.26–1.12)
image

Figure 1.  Summary estimate of the relationships between fermented soy food intake and gastric cancer risk in Japanese and Korean populations. CI, confidence interval; OR, odds ratio; RR, relative risk; shaded box, point estimate of each study; horizontal line, 95% CI of each study; diamond, summary point estimate and its 95% CI of the studies. The size of the shaded box indicates the relative weight of each study in the meta-analysis.

Download figure to PowerPoint

image

Figure 2.  Summary estimate of the relationships between non-fermented soy food intake and gastric cancer risk in Japanese and Korean populations. CI, confidence interval; OR, odds ratio; RR, relative risk, shaded box, point estimate of each study; horizontal line, 95% CI of each study; diamond, summary point estimate and its 95% CI of the studies. The size of the shaded box indicates the relative weight of each study in the meta-analysis.

Download figure to PowerPoint

image

Figure 3.  Begg’s funnel plot for fermented soy foods in case–control studies and cohort studies.

Download figure to PowerPoint

image

Figure 4.  Begg’s funnel plot for non-fermented soy foods in case–control studies and cohort studies.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Soy food has been a major plant source of dietary protein for Asians, especially Japanese and Koreans, and many epidemiological studies suggest that soy intake may be a strong protective factor against cancer in humans.(32) Soy food consumption might help reduce the risk of various cancers, including breast cancer, colon cancer, colorectal cancer and gynecological cancer.(3,32–39) However, although soy foods were among the specific food items assessed in the studies included in this meta-analysis, few of the investigators discussed their findings on soy consumption in connection with GC risk.

In the present study, a high intake of non-fermented soy food decreased GC risk (overall summary OR = 0.64, 95% CI = 0.54–0.77; cohort studies OR = 0.83, 95% CI = 0.60–1.13; case–control studies OR = 0.57, 95% CI = 0.46–0.71). Soy has a protective effect against GC.(40) Compared with other plant food sources,(34) non-fermented soy foods are an abundant source of isoflavones, including genistein and daidzein,(41) which are primary anticancer elements in soy. They may block the intragastric formation of carcinogenic N-nitroso compounds.(42,43) Genistein has been shown to inhibit the growth of GC cells.(44) In addition, genistein lessens gastric carcinogenesis by increasing apoptosis and decreasing cell proliferation and the angiogenesis of antral mucosa and GC.(45) Other components in soy foods may also be important anticancer agents.(11)

It has been noted that the form and sources of isoflavones might change the association between soy isoflavones and cancer risk.(46) The dietary habits of Korean and Japanese people are characterized by a high intake of salty foods and carbohydrates, and a higher intake of cooked rather than fresh vegetables.(25) Increases in the risk of GC associated with a high intake of fermented soy foods have been reported in epidemiological studies among Koreans,(22,47) Japanese,(15,48) Taiwanese(49) and Chinese(50) populations. The GC risk increases with the consumption of fermented soy foods and decreases with the consumption of non-fermented soy foods.(4) In other words, the effect of soy foods on the risk of GC differs depending on the preparation of the soy food.

In our study, a high intake of fermented soy food, such as soybean paste and stews, was significantly associated with an increased GC risk (overall summary OR = 1.22, 95% CI = 1.02–1.44; cohort studies OR = 1.12, 95% CI = 0.88–1.41; case–control studies OR = 1.34, 95% CI = 1.04–1.73). Fermented soy foods generally contain a considerable amount of salt added during preparation or fermentation. Many studies already show that a high salt diet is a risk factor for GC.(22,25,51,52) For example, Koreans have the highest rates of 24-h urinary sodium excretion(47) and one of the highest rates of mortality from GC.(53,54) Salt is not a carcinogen, but consumption of salt and foods preserved in salt might increase (atrophic) gastritis through damage to the gastric mucosa, which then induces DNA synthesis and cell proliferation, leading to GC.(55) Moreover, foods high in salt improve Helicobacter pylori colonization in the stomach.(56)H. pylori infection is associated with an increase in the endogenous synthesis of nitrate in the stomach. It decreases gastric vitamin C concentrations(57) and increases endogenous N-nitroso compound formation. Therefore, a high intake of salt and foods preserved in salt has been considered a plausible cause of GC in many studies.(8,22,26,28,58–60) Antioxidant loss in fermented soy foods as a result of processing and storage under acid and oxygen might explain the beneficial effects of non-fermented soy foods on GC risk.(60–62) Another possible explanation is that fermented soy foods may contain nitroso compounds, thereby inducing gastric carcinogenesis.(63,64) Indeed, fermented soybean pastes have a high level of nitrate.(65) After we absorb dietary nitrate, 25% of it is secreted into the saliva and oral bacteria reduce approximately 5% to nitrite.(53) Since most saliva is swallowed, 80% of gastric nitrite in the normal acidic stomach comes from the reduction of ingested or endogenous nitrate.(66) Gastric nitrosation, which is carried out by nitrite, might produce unstable N-nitroso compounds that do not reach extra-gastric sites and act in the stomach to initiate GC, or may produce stable N-nitroso compounds that induce cancer at other sites.(66)

Soy food intake varies widely between Asian countries and Western countries(67) The amount of soy consumption is higher in Asian countries. For example, the amount of isoflavone consumption in Western countries averages <1 mg/day,(68,69) whereas it averages approximately 50 mg/day in Chinese and Japanese(38,70) populations. However, the incidence of GC is higher in Japan and Korea than in the USA and Europe. One of the important reasons for this is that Japanese and Korean populations consume more fermented soy foods than Americans and Europeans. However, it is known that soy is not the only plant food that provides isoflavones.(32) Isoflavones in Western diets stem from various other food sources.(71) Henceforth, some caution should be taken in associating isoflavone intake with soy food consumption.(32)

As seen above, the effects of fermented soy products might be confounded by salt intake, and the effects of non-fermented soy products might be confounded by vegetable and fruit intake(3,11) In almost all studies included in this meta-analysis, the possible confounding effects of salt, vegetable/fruits, and other dietary factors had not been considered in the soy product analysis(11) Moreover, the roles of salt, N-nitroso compounds, fruit/vegetable intake and other dietary factors have not been adjusted for in the majority of studies on soy food. Only age and sex were adjusted for in the analyses on soy foods.(3) We did not obtain information on infection with H. pylori, a major risk factor for GC,(11) and thus another potential confounding factor. H. pylori infection might be an intermediate factor between soy product intake and GC. These factors all need to be considered in future analyses on soy foods.

It is noted that there are some limitations regarding the interpretation of the meta-analysis. We selected a random-effect model in order to adjust for the effect of heterogeneity across studies, but this model does not discount the effects of heterogeneity. There is no change in the statistical significance of results obtained with a fixed-effect model and a random-effect model (overall summary OR = 1.18, 95% CI = 1.09–1.27, cohort studies OR = 1.09, 95% CI = 0.98–1.21, case–control studies OR = 1.28, 95% CI = 1.14–1.44 in the fixed-effect model for fermented soy food; overall summary OR = 0.77, 95% CI = 0.70–0.84, cohort studies OR = 0.94, 95% CI = 0.81–1.08, case–control studies OR = 0.66, 95% CI = 0.59–0.75 in the fixed-effect model for non-fermented soy food).

In addition, another bias occurs in this meta-analysis. Publication bias can cause researchers to reach incorrect conclusions from their meta-analyses because studies with statistically significant results tend to be published. The Begg’s test shows that there is no significant publication bias in this meta-analysis, but we could not ignore the possibility of this bias, inherent in any meta-analysis. Additionally, most studies were not planned to determine the effects of fermented or non-fermented soy foods on GC risk. Therefore, an outcome-reporting bias may have affected the validity of our meta-analysis.(72) Another limitation is the heterogeneity of the available data. Most studies were not designed to investigate soy foods, causing the quantification of soy intake and the extent to which confounding factors were adjusted to be different across studies in the meta-analysis.(32) The present study used a random-effect model for considering the heterogeneity in the calculation of summary statistics. However, this limitation might complicate the interpretation of the results.(32) Therefore, the conclusions or interpretations made from the results of this meta-analysis should be considered cautiously due to the above limitations and biases.

In conclusion, the results of this meta-analysis show that a high intake of fermented soy foods is associated with an increased GC risk, whereas a high intake of non-fermented soy foods is associated with a decreased GC risk. These results might explain why GC incidence rates in Japan and Korea are high despite a high intake of soy foods. A high intake of non-fermented soy foods rather than fermented soy foods should help to reduce GC incidence rates in Japan and Korea.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

This study was supported by a Grant from National Cancer Center, Korea (0910221) and a Grant for the Third Term Comprehensive Control Research for Cancer from the Ministry of Health, Labour and Welfare of Japan.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References