Genetic susceptibility and gastric cancer risk



The aim of the present paper is to review and evaluate, in a comprehensive manner, the most recent published evidence on the contribution of genetic susceptibility to gastric cancer risk in humans. We have identified all studies available in MEDLINE published up to October 2001. Only studies carried out in humans and comparing gastric cancer cases with at least 1 standard control group were included in the analysis. We were able to find 31 articles based on 25 case-control studies carried out in Caucasian, Asian and African populations. Most of the studies assess the effect of genes involved in detoxifying pathways (n = 12) and inflammatory responses (n = 7). The most widely studied is the GSTM1 null polymorphism. Only a very few studies have evaluated the risk of gastric cancer associated with genes acting on mucosa protection, oxidative damage and DNA repair. The most consistent results are the increased gastric cancer risk associated with IL1B and NAT1 variants, which may account for up to 48% of attributable risk of gastric cancer. Only polymorphisms at HLA-DQ, TNF and CYP2E genes may confer some protective effect against gastric cancer. The most important limitations that preclude definitive conclusions are (i) the lack of appropriate control of potential sources of bias (only 5 population-based studies have been published so far); (ii) the low number of cases analyzed (14 studies included fewer than 99 cases); and (iii) the low number of studies (n = 3) offering concomitant analysis of genetic susceptibility and exposure to relevant cofactors (Helicobacter pylori infection, diet and smoking). We conclude that the scientific data on the role of genetic factors in gastric cancer risk are promising. The lack of association reported so far should be considered with caution due to significant limitations in study design. Cohort studies taking into account simultaneously the different genetic and environmental factors potentially involved in gastric tumorigenesis are needed to ascertain not only the relative contribution of these factors to tumor development but also the contribution of their putative interactions. © 2002 Wiley-Liss, Inc.

The incidence and mortality of gastric cancer (GC) are decreasing in most European countries and also throughout the world.1 However, GC still represents the second most frequent cancer in the world and the fourth in Europe. The 5-year survival rate of GC is low, and identification and a better control of risk factors seem to be the most effective means of prevention.

Helicobacter pylori (Hp) infection is an established risk factor of GC2 that could act through different mechanisms, including inducing hyperproliferation of gastric cells, interfering with antioxidant functions and increasing the amount of reactive oxygen species and nitric oxide, which may be responsible for oxidative DNA damage. Nevertheless, it has been suggested2 that Hp may be involved with GC in only a very small proportion of people infected with the organism. The cofactors involved in producing gastric mucosa malignant transformation are currently unknown. Hereditary GC predisposition syndromes, usually associated with E-cadherin germline mutations, contribute very little to the overall load of new GC cases.3 Diet4 and smoking5 are known to play an important role. A multifactorial model of human gastric carcinogenesis is currently accepted, according to which different dietary and nondietary factors, including genetic susceptibility, are involved at different stages in the cancer process.

Individual variations in cancer risk have been associated with specific variant alleles of different genes (polymorphisms) that are present in a significant proportion of the normal population. Polymorphisms in a wide variety of genes may modify the effect of environmental exposures,6 and these gene-environmental interactions could explain the high variation in the GC incidence observed around the world.7 However, the interaction between environmental factors and genetic susceptibility, at least regarding the metabolic genes, has not yet been adequately addressed.8

Recent results suggest that single nucleotide polymorphism studies may shed light on GC tumorigenesis.9 In fact, individual genetic susceptibility may be critical in a variety of processes relevant to gastric carcinogenesis including (i) mucosal protection in the face of Hp infection and other carcinogens; (ii) the inflammatory response, which conditions the maintenance, severity and outcome of the Hp infection; (iii) the functioning of carcinogen detoxification and antioxidant protection; (iv) the intrinsic variability of DNA repair processes; and (v) cell proliferation ability. However, the mechanism of human gastric tumorigenesis is still relatively unknown, and other relevant processes could be added to this preliminary list in the coming years. Reviews published some time ago10, 11, 12, 13 revealed few and inconsistent results concerning the role of genotype. The aim of our article is to review and evaluate, in a comprehensive manner, the most recent published evidence on the relative contribution of genetic susceptibility to GC risk in humans.


We have identified all studies published up to October 2001 available in MEDLINE (National Library of Medicine) that investigated the role of genetic susceptibility in GC risk. Only studies carried out in humans and comparing GC cases with at least 1 standard control group were considered for analysis. Studies performed in cell lines or in case series without controls, or based only on serologic or histochemical assays, were excluded. We also excluded those studies that evaluated only the role of genetic factors as prognostic markers, as well as those describing somatic mutations in tumor tissue. We were able to find 31 articles based on 25 case-control studies, which are summarized in Tables I–III. We evaluated the design and size of each study, the selection and features of cases and controls, the availability and use of information on environmental factors (mainly Hp infection, diet and tobacco smoking), the available information on histologic type and anatomic subsite, the reported results and the available information on gene-environmental and gene-gene interations. For each study, a short evaluation of potential biases and confounding factors is included.

Genes have been named according to the HUGO Gene Nomenclature Committee (HGNC; Whenever possible, polymorphisms have been termed according to the proposed nomenclature of Antonarakis et al.14 In short, polymorphism designation starting with a number refers to a nucleotide position, and subsequent letters refer to the nucleotide change. Polymorphism designation starting with a letter (or 2 letters separated by a slash) refers to an amino acid substitution (single-letter amino acid code), and the number following it is the codon position. Metabolic gene allele nomenclature is according to that recommended by Garte et al.15 (


Protection of Gastric Mucosa Against Damaging Agents

Important factors in gastric mucosa protection include mucins (MUC1, MUC2, MUC5AC and MUC6) and trefoil peptides (TFF1 [originally called pS2], TFF2 [formerly spasmolytic peptide SP] and TFF3 [also called intestinal trefoil factor, ITF]).

Mucins (MUC1 and MUC6).

Mucins are glycoproteins with a high carbohydrate content that are thought to play a major role in gastric mucosa protection. However, at the same time, the gastric mucus layer is the major reservoir of the Hp. Normal gastric mucosa expresses the so-called gastric mucins MUC1 and MUC5AC in the superficial epithelium and MUC6 in the deep glands.16 These mucin core proteins are known to be highly polymorphic in their length due to the presence of variable number tandem repeats (VNTRs) in the coding region of their genes. It has been speculated that variant alleles of yet unknown functional roles may be relevant to gastric tumorigenesis.

Two studies (Table I) have addressed this issue for the MUC1 gene.In a case-control study in a Portuguese population (which has the highest GC risk in Europe), small MUC1 alleles were more frequently detected in GC patients.17 This led to the hypothesis that MUC1 genotypes displaying smaller glycoprotein products lead to a less efficient protection of the mucosa. However, a later study18 showed that small MUC1 alleles were more frequent in the Danish population (which has the lowest GC risk in Europe) than in the Portuguese population, which would lead us to disregard the presence of small MUC1 alleles as a significant risk factor for GC.

Table I. Case-Control Studies on Gastric Cancer Risk and Polymorphisms in Genes Related to Mucosa Protection and Inflammatory Responses1
Gene and variantAuthor and countryStudy subjectsCovariatesResults2 (OR and 95% CI)Histologic typesAnatomic sitesInteractionObservations
No. of casesNo. of controlsStudy base
  • 1

    NA, not available; SS, statistically significant; OR, odds ratio; CI, confidence interval; Hp, Helicobacter pylori; GC, gastric cancer.

  • 2

    Allele and genotype nomenclatures according to authors.

Mucosa protection
 MUC1 VNTRCarvalho et al.17 1997 Portugal159324Blood donorsAge, sexSmall vs large = 4.3 (1.8–10.5)NANANADifferent age and sex distribution between cases and control
 MUC6 VNTRGarcia et al.19 1997 Portugal157376Blood donorsAge, sexNA (small alleles higher frequency in GC, p < 0.05)No differencesNo differencesNASame study of reference (Carvalho et al.17)
Lack of information on smoking, diet, Hp infection
 TFF2 VNTRdos Santos Silva et al.23 1999 Portugal5449HospitalNoneNA (no difference in allele frequency between GC and controls)No differencesNANASmall size; control and case groups not described
Inflammatory responseLack of information on smoking, diet, Hp infection
 IL1B-31C>TEl-Omar et al.27 2000 Scotland/PolandScotland: 103100PopulationHp+, acid secretion, age, sex, tobacco, alcohol, ABO, family history HpLow vs. normal acid:No differencesNo differencesNAComplete information on confounding
 IL1RN*2 (IVS2 86 bp VNTR)Poland: 103430Population IL1B-31T+ = 7.5 (1.8–31)
 IL1RN*2/*2 = 2.1 (0.7–6.3)
GC vs controls:
 IL1B-31T+ = 1.6 (1.1–2.2)
 IL1RN*2/*2 = 2.9 (1.9–4.4)
 IL1B-31C>TMachado et al.28 2001 Portugal152220Blood donorsAge, sexIL1B-511T intestinal = 2.7 (1.5–4.9)Higher risk for intestinal than for diffuseNANADifferent age distribution between cases and controls
IL1B-511T all types = 1.7 (1.1–2.7)
IL1RN*2 intestinal = 3.1 (1.5–6.5)Lack of information on smoking, diet, Hp infection
IL1RN*2 all types = 1.7 (0.9–3.3)
IL1B-511T + IL1RN*2 = 9.0 (3.5–23.0)
PAR = 46%
 IL1RN*2 (IVS2 86 bp VNTR)
 HLA-DQB1*0301Lee et al.29 1996 USA52200HospitalSex, race, Hp (exclusion of non Caucasian)HLA-DQB1*0301 = 3.2 (NA)No differencesNo differencesNASmall size; case and control groups not described; lack of information on smoking, diet
 HLA-DRB1No differences regarding HLA-DQA1, HLA-DRB1, and HLA-TAP2 alleles
 HLA-TAP2 alleles
 HLA class I geneOhmori et al.30 1997 Japan88525HealthyNoneNo differences SS for any alelles after Bonferroni correctionNo differencesNANASmall size; control group not well described; lack of information on diet, smoking, Hp infection
 HLA class II gene
 HLA-DQA1*0102Azuma et al.31 1998 Japan82167Health testingAge, sex, Hp, serum pepsinogensNA (DQA1 *0102 SS lower risk for intestinal type and Hp+)DQA1 *0102 lower risk for intestinal typeNANASmall size; lack of information on diet, smoking
 HLA-DQA1Magnusson et al.32 2001 Sweden130263PopulationHp, Cag A, birth year, sexDQA1*0102 = 0.26 (0.09–0.75)DRB1*1601 association stronger in diffuse typeNANALack of information on diet, smoking.
 HLA-DRB1DRB1*0701 = 0.24 (0.08–0.74)
DRB1*1501 = 0.48 (0.25–0.94)
No SS after correction for multiple test
DRB1*1601 all types = 8.7 (2.7–28.0)
 TNF-308G>AJang et al.35 2001 South Korea5292HealthyNoneTNF-308A = 1.58 (0.5–5.0) TNF-238A = 0.27 (0.06–1.14)NANANASmall size; control and case groups not well described
 TNF-238G>ALack of information on smoking, diet, Hp infection

In another case-control study in the Portuguese population,19 smaller alleles of MUC6 were also more frequently detected in GC cases than in controls, suggesting that the MUC6 VNTR polymorphism is involved in a predisposition to GC development. No differences were observed in allelic distribution according to localization and histologic type, in spite of the observed differences in expression of mucins and trefoil peptides, between intestinal and diffuse GC.20

Trefoil Peptides.

Trefoil peptides (TFF1, TFF2 and TFF3) are a group of small secretory proteins that play an important role in gastric mucosa maintenance and repair.21 Strong expression of TFF1-pS2 has been associated with the diffuse type, whereas the intestinal type does not express it.22 A diallelic polymorphism in intron 2 of the TFF2 gene (spasmoysin) has been described.23 This polymorphism involves a 25 bp sequence that is tandemly repeated 48 or 53 times; its functional relevance is unknown. In a Portuguese case-control study, no significant differences in allele distribution of this TFF2 diallelic VNTR were observed between cases and controls.23

Individual Differences in Inflammatory Responses (immunogenetics)

Individual differences in the intensity of the inflammatory response (which affects the maintenance, severity and outcome of Hp infection) may contribute to gastric mucosa transformation. Moreover, the impact of gene polymorphisms on the activity of key inflammatory molecules is relatively well known.

IL1B and IL1RN.

Interleukin-1β (IL-1β) and the interleukin-1 receptor antagonist (IL-1ra) are potent cytokines that play a key role in modulating the inflammatory response in the gastrointestinal mucosa24 as well as in regulating gastric acid secretion.25 The IL-1ra is an anti-inflammatory cytokine that binds to the IL-1 receptors, thereby modulating the proinflammatory effects of IL-1 proteins. Interindividual variations in IL-1 and IL-1ra protein levels seem to be determined26 by functional polymorphisms in the transcription regulatory regions of their respective genes.

An important case-control study recently conducted in Scotland and Poland27 has shown that the IL-1β gene (IL1B) and the IL-1ra gene (IL1RN) variants IL1B-31C>T (T+ genotype) and IL1RN IVS2, 86 bp VNTR (*2/*2 genotype), thought to increase IL-1β production and to inhibit gastric acid secretion, were associated with (i) an increased risk of chronic hypochlorhydric response to Hp infection; and (ii) an increased GC risk. The combined attributable risk of GC for these polymorphic variants was 38%. There were no differences between histologic types and anatomic subsites. The 2 polymorphisms are in nearly complete linkage disequilibrium, and 1 (-31C>T), in the TATA-box of the IL1B gene, markedly affects DNA-protein interactions in vitro. These interactions were not affected by the IL1B-511C>T polymorphism, which was found to be in almost total linkage disequilibrium with -31C>T.

A new case-control study in Portugal28 has confirmed this result. Alleles IL1B-511T and IL1RN*2 were associated with an increased risk for GC. For the IL1RN*2 allele, the association was statistically significant only for the intestinal type. For the diffuse type, significance was not reached, but the number of cases was low. The IL1B-511T allele apparently acts in a dominant manner, whereas the IL1RN*2 allele could be recessive; an interaction between them was observed. Twenty-two percent of cases and 4.6% of controls were IL1B-511T/IL1RN*2 carriers, and the combined population attributable fraction of cases was 46%.


HLA antigens are cell surface proteins that bind to the antigen-specific T-cell receptor, forming an important signaling event for the cell inducing or suppressing immune response. There are 2 classes of HLA genes: HLA class I (HLA-A) and HLA class II (HLA-DR, HLA-DQ and HLA-DP). All these genes are highly polymorphic, with variant alleles of putative functional significance. A variety of associations among GC risk, Hp infection and specific HLA alleles have been described. In a case-control study in the United States,29 the HLA-DQB1*0301 allele was more common in GC than in controls (odds ratio [OR] = 3.2). No differences were observed among HLA-DQA1, DRB1 and TAP2 variant alleles. Paradoxically, the same HLA-DQB1*0301 allele was protective against Hp infection, leaving open the possibility of an increased risk through an Hp-independent pathway. Japanese studies have provided controversial data. One case-control study30 did not show any differences in HLA class I and class II alleles between GC and controls. Another case-control study suggested that the absence of HLA-DQA1*0102 may be a host genetic risk factor for Hp infection and the intestinal type of GC.31 Finally, a recent European study32 has confirmed absence of the HLA-DQA1*0102 allele as a risk factor for Hp infection, but the study was not able to show an association with GC risk. On the contrary, the HLA-DRB1*1601 allele was positively associated with GC risk, being stronger in the Hp(−) and diffuse type of GC. Together these results point to the existence of variant alleles in different members of the HLA system that may confer differential risk for Hp infection and GC, a fact that may help in elucidation of the molecular basis of Hp infection and human gastric tumorigenesis.


Tumor necrosis factor is another potent proinflammatory cytokine. TNF levels are increased in Hp-infected patients,33 probably because of a link to the secretion of urease.34 As reported for other inflammatory cytokines, variant alleles of functional relevance have been described for the TNF gene. In a case-control study in Korea,35 the TNF-308A allele, which is known to upregulate transcription of the TNF gene, was found not to be associated with GC risk. On the other hand, the TNF-238A allele, which is known to downregulate TNF transcription, could be protective against GC. The major limitation of this study was the small sample size, which precludes drawing more definitive conclusions.


Prostaglandin-endoperoxide synthases 1 and 2, also known as cyclooxygenases (COX1 and COX2), are key enzymes in the conversion of arachidonic acid to prostaglandins (PGs), which are important mediators of inflammation. PTGS2 is not only an inflammatory molecule but also mediates oncogenic signals. PTGS2 expression is induced by tumor promoters, growth factors and oncogenes36 and may affect cell proliferation, angiogenesis, carcinogen metabolism, production of reactive oxygen species or modulation of the immune system.

A role for PTGS2 in human gastric carcinogenesis can be envisioned. PTGS2 expression is induced by Hp infection in gastric mucosa,37 and Hp eradication diminishes PTGS expression in intestinal metaplasia.38 PTGS2 may also be required in the IL-1β inhibition of acid secretion.24 PTGS2 is overexpressed in a variety of tumors including GCs,39 being higher in the intestinal type40 and in tumors arising in the gastric body.41 Several studies have shown that consumption of anti-inflammatory drugs may protect against GC,42, 43, 44 a protective effect that is stronger in Hp-infected subjects.44 Recently, variant alleles have been described in the PTGS2 gene. However, as far as we know, no studies have yet been reported on the role of PTGS2 polymorphisms in GC risk.

Ability To Detoxify Carcinogens

Inherited polymorphisms in metabolic enzymes are supposed to account for the variability observed in the metabolism of xenobiotics and carcinogens.45 However, dietary and other environmental factors may significantly affect their activity. The metabolic system is rather nonspecific, in order to be efficient in the face of a wide spectrum of molecules. The large number of metabolic enzymes can be grouped into families with different but overlapping substrate specificities. Phase I enzymes (like the cytochrome P450 or CYP superfamily) catabolize oxidative reactions that introduce electrophilic groups in the molecules and make them more reactive, usually leading to carcinogen activation. Phase II enzymes introduce a hydrophilic group (i.e., glutathione peptides or acetyl groups) into the intermediate molecules, which usually results in detoxification of activated carcinogens. Although both phases usually occur sequentially, they may be independent. When one is analyzing carcinogen detoxification, it is of interest to analyze enzymes involved in both phases.

Phase I Enzymes.

The 2 isoforms exhibiting the highest catalytic activity are CYP1A1 and CYP1A2. CYP1A1, CYP1A246 and CYP3A447 are expressed in gastric mucosa with intestinal metaplasia but not in mucosa without metaplasia. Furthermore, members of the subfamilies CYP1A and CYP3A are expressed in stomach cancer but not in the normal stomach.48 The phase I enzymes are described below.

  • 1CYP1A1. The gene product of CYP1A1 is the aryl hydrocarbon hydroxylase (AHH), which metabolizes polycyclic aromatic hydrocarbons (PAHs).45 Two polymorphisms have been extensively studied: 3801T>C (also known as MspI or m1) in the 3′ noncoding region and 2455A>G (also known as m2) or I462V in exon 7.49 In spite of its expression in gastric mucosa, no studies have been identified that assess the potential influence of CYP1A1 polymorphisms in GC risk.
  • 2CYP2E. CYP2E, the only member of the CYP2E subfamily identified so far, metabolizes low molecular weight toxins and catalyzes exogenous N-nitrosoamines. Its activity is also modulated by several dietary factors. At least 4 genetic polymorphisms have been associated with cancer risk.50, 51 These are as follows: -1293G>C (PstI) and -1053C>T (RsaI) in the 5′-flanking region (variant alleles, CYP2E1*5A and *5B); 7632T>A (DraI, variant alleles, *5A and *6) and 9893C>G (TaqI, variant allele, *1B). One of them (RsaI) is associated with increased gene expression.50 In Japanese case-control studies,52, 53, 54 the RsaI variant allele was not associated with increased GC risk, of either the intestinal or the diffuse type (Table II).
  • 3CYP2A6. The protein encoded by this gene also plays a role in N-nitrosamine metabolism.45 Three alleles have been described, and the CYP2A6*2 allele has a mutation (479T>A or L160H) that leads to a defective enzyme containing histidine for leucine at position 160.56 Individuals with such a mutation would be protected because of low oxidative activity of procarcinogens, but so far no study has been published regarding GC risk.
Table II. Case-Control Studies on Gastric Cancer Risk and Polymorphisms in Genes Related to Metabolic Enzymes1
Gene and variantAuthor and countryNo. of Study subjectsCovariateResults2 (OR and 95% CI)Histologic typesAnatomic sitesInteractionObservations
No. of casesNo. of controlsStudy base
  • 1

    NA, not available; OR, odds ratio; CI, confidence internal; Hp, Helicobacter pylori; BMI, body mass index; CVD, cardiovascular disease.

  • 2

    Allele and genotype nomenclatures according to authors. However, the use of patients with gastric or duodenal ulcer as controls hampers the interpretation of these results. Interestingly, the same RsaI variant allele was shown to be protective against GC in non-Japanese Brazilians but not in Japanese-Brazilian subjects,55 with no differences regarding tumor histology.

(Metabolic enzymes)
CYP2E-1053C>T (RsaI)Kato et al.52 1995 Japanese150203HospitalNoneHeterozygote allelesNANANAControls with benign gastric disease
 (C/R vs. C/C = 1.04  (0.74–3.08)Lack of information on diet smoking, Hp infection
Homozygote rare allele
 (R/R vs. C/C = 0.57  (0.22–1.50)
CYP2E-1053C>T (RsaI)Kato et al.54 1997 Japan284284HospitalAge, sex, pepsinogen, HpC/C vs. C/R = 1.05  (0.7–1.5)No differencesNANAControls with benign gastric disease
C/C vs. R/R = 1.02  (0.5–2.2)Lack of information on diet, smoking
CYP2E-1053C>T (RsaI)Kato et al.53 1996 Japanese82151HospitalAge, sex, pepsinogen, HpCYP2E C/R R/R vs. C/CNo differencesNANAControls with benign gastric disease
GSTM1 null Intest = 0.90 (0.45–1.8)Lack of information on diet smoking
 Diffuse = 0.61  (0.3–1.3)
GSTM1 null
 Intest = 0.8 (0.4–1.8)
 Diffuse = 0.9 (0.4–2)
CYP2E-1053C>T (RsaI)Nishimoto et al.55 2000 BrazilJapanese-Brazilians:HospitalTobacco, alcohol, dietary pattern, sex, family historyRsaINo differences in BraziliansNANot foundLack of information on Hp infection
 96 193Brazilians: C/A, C/C =  0.46 (0.2–1.0)Variant genotype higher in diffuse type in Japanese cases
Non Japanese-brazilians:Japanese: C/A, C/C =  0.98 (0.5–1.9)
 236 236
GSTM1 nullStrange et al.58 1991 UK1949HospitalNoneM1 null = 2.9 (1.25–6.73)NANANAVery small size; control died from CVD
Lack of information on diet, tobacco, Hp infection
GSTM1 nullHarada et al.59 1992 Japan1984Blood donorsNoneM1 null = 3.08  (0.92–10.9)NANANAVery small size; case-control groups not described
Lack of information on diet, tobacco, Hp infection
GSTM1 null GSTT1 nullKatoh et al.60 1996 Japan139126Health testingMedical history, smoking, occupationM1 null = 1.70 (1.05–2.8)No differencesNAM1 null, higher risk in low smokingLack of information on diet, Hp infection
In low smoking group =  5.8 (1.2–28.3) but not dose-response T1 null = 1.13  (0.7–1.8)
GSTM1 nullKato et al.54 1997 Japan284284HospitalAge, sex, pepsinogen, Hp, histologic typeM1 null = 1.17 (0.8–1.7)No differencesNANASame study reference (Kato et al.54)
GSTM1 null GSTT1 nullDeakin et al.62 1996 UK136577HospitalAge, sex, smokingM1 null: no association T1 null: no associationNo differencesNAM1-T1 not foundLack of information on diet, Hp infection
GSTM1 null GSTT1 nullSetiawan et al.63 2000 China91429PopulationAge, sex, educ., smoking, fruit-salt, alcohol, BMI, HpM1 null = 0.60 (0.3–1.4) T null = 2.5 (1.01–6.2)No differencesNo differencesM1, T1 null, higher risk in smokers and low fruit intakeAll information on confounding collected
M1-T1 not found
GSTM1 null GSTT1 null GSTM3*B  (IVS6del3) GSTP1 1578A>G (l105V)Lan et al.64 2001 Poland304427PopulationAge, sex, tobacco, alcohol, fruit, family history, education, Hp, BMIM1 null = 0.92 (0.7–1.3)No differencesNo differencesT1 null, higher risk in smokersAll information on confounding collected
M3, AA = 0.99 (0.7–1.4)
P1, GG = 0.67 (0.4–1.2)
T1 null All = 1.48  (0.97–2.25)
21–49 years = 3.9  (0.9–11.1)
Current smoker = 3.1  (1.5–6.5)
GSTM1 nullSaadat et al.61 2001 Iran42131Blood donorsAge, sexM1 null = 2.3 (1.15–4.95)NANANASmall size
GSTT1 nullT1 no associationLack of information on diet, tobacco, Hp infection
M1 null + T1 null = 3.31  (1.1–9.6)
GSTP1 1578A>G (I105V)Katoh et al.65 1999 Japan140122Health testingAge, sex, smokingP1 A/G or G/G vs. A/A =  1.56 (0.9–2.73)No differencesNANot foundSame study of reference (Katoh et al.60)
GSTP1 1578A>G (I105V)Setiawan et al.66 2001 China84429PopulationAge, sex, education, BMI, Hp, tobacco, salt, fruitsP1 = 1.3 (0.1–11.2)NANANot foundSame study of reference (Setiawan et al.63)
NAT1 *10 (1088T>A 1095C>A)Boissy et al.69 2000 UK94112HospitalAge, sex, smokingNAT1 slow vs. rapid = 2.6  (1.3–5.3)NANANADifferences by age and sex between cases and controls
NAT2NAT2 slow vs. rapid = 3.8  (0.9–17)Lack of information on diet, Hp, smoking only yes/no
NAT1*10 (1088T>A 1095C>A) NAT2Katoh et al.68 2000 Japan140122HealthyAge, sex, smokingNAT1*4 functional  references allele vs.  NAT1*10 = 1.49  (0.85–2.62)NAT1 higher risk in intestinal typeNANAT1 higher risk in heavy smokersSame study of reference (Katoh et al.60)
Intestinal = 3.03 (1.1–8.5)
Diffuse = 0.98 (0.5–1.8)
Intestinal smoker = 4.24  (0.9–20.1)
NAT2, no association

Phase II Enzymes.

The main components of these enzymes are glutathione S-transferases (GSTs) and N-acetyl transferases (NATs). GST enzymes bind glutathione (GSH), a nucleophilic tripeptide, to carcinogens, facilitating detoxification of several compounds. Four major GST families are expressed in humans: GSTA (α), GSTM (μ), GSTT (θ) and GSTP (π); GSTTT1, GSTM1 and GSTM3 are the most studied genes because of the presence of null alleles (homozygously deleted) in the population. The phase II enzymes are described below.

  • 1GSTM1. This protein is expressed in the liver, brain and stomach. Absence of GSTM1 expression (found in 10–60% of individuals),57 due to an inherited deletion of the paternal and maternal alleles of the GSTM1 gene, should confer an increased cancer risk because of a low ability to detoxify carcinogens. Several studies have suggested that the null allele may be a risk factor for GC. Out of 9 case-control studies, 458, 59, 60, 61 found an increased risk of GC (OR range = 1.7-3.1). The remaining 5 showed no association.53, 54, 62–64 Interactions between null genotype and smoking60, 63 and low consumption of fruit63 have been observed; no association with Hp infection was found.54, 63 It is noteworthy that no differences were observed by histologic type and anatomic subsite.
  • 2GSTT1. Similar to what occurs for GSTM1, a variable proportion of individuals (13–55%)57 express no GSTT1 because of a null genotype. Comparisons of GSTT1 null genotype and GC risk are also inconsistent. Most studies did not find an association60–62, 64 although 1 study64 found a positive statistically significant association in current smokers of young age. On the other hand, a population-based Chinese study63 showed an increased GC risk (OR = 2.5) adjusted for smoking, dietary factors and Hp infection. An interaction between GSTM1 and GSTT1 was also explored; the findings were contradictory. No interaction was observed in some studies,62, 63 whereas an OR of 3.3 was reported for the simultaneous presence of both null genotypes in another study.61
  • 3GSTP1. The gene of this enzyme (GSTP1) also presents functional polymorphisms in humans. An A to G substitution in exon 5 (1578A>G) leads to an Ile to Val replacement at amino acid 105 of the GSTP1 protein (I105V), reducing its enzymatic activity. Two studies65, 66 have investigated the association between GC and the GSTP1 I105V genotype. Although a trend was observed toward an increased risk for the variant allele in the Japanese study,65 both results were negative. No interaction between smoking and genotype was observed.
  • 4NAT1/NAT2. In humans, 2 genes (NAT1 and NAT2) are responsible for N-acetyltransferase activity. N-acetyltransferases catalyze N-acetylation (mainly deactivation) and O-acetylation (mainly activation) of aromatic and heterocyclic amine carcinogens. Polymorphisms in their genes differentiate subjects into rapid and slow acetylators, influencing the metabolism of environmental arylamines, and modifying the individual susceptibility to cancer.45 There are ethnic differences in the allele frequencies as well as tissue-specific variations in the expression and function of NAT1 and NAT2.67 Two studies have observed a significant increase of GC risk (OR from 1.5 to 2.6) associated with the the NAT1*10 allele (1088T>A and 1095C>A in the 3′ untranslated region), which has also been associated with higher levels of NAT activity.68, 69 In the Japanese study,68 the risk of GC was particularly high in the intestinal type of cancer and in heavy smokers, suggesting that NAT1 may be involved in smoking-induced gastric tumors. In contrast, no significant association with NAT2 rapid acetylator genotypes was observed in these studies. In the English study,69 an increased risk associated with the NAT2*4 allele (wild type) was observed that did not reach statistical significance. Although it may be hypothesized that slow acetylators may confer some protection against GC, this finding has not been described so far.

Protection Against Oxidative Damage and Other Inductors of DNA Damage

Reactive oxygen species (ROS) may be involved in several steps of the carcinogenic process including carcinogen activation, oxidative DNA damage, tumor promotion and inhibition of DNA repair activities. In vivo investigations have shown that Hp induces the synthesis of ROS in gastric epithelial cells70 and that bacterial eradication attenuates oxidative stress in human gastric mucosa.71 Protection against this oxidative damage may come from the protective action of enzymes, such as the mitochondrial manganase superoxide dismutase (SOD2), which is involved in removal of this damaging ROS. Although high SOD levels have been found in precancerous gastric lesions in humans,72 no studies have addressed the putative role of individual variations in SOD2 activity in GC risk.


5,10-Methylenetetrahydrofolate reductase (MTHFR) catalyzes the reduction of methylene −THF to methyl THF, the predominant circulating form of folate for the remethylation of homocysteine to methionine. Concomitant deficiency of folic acid or vitamin B12 leads to massive uracil incorporation into DNA and chromosome breaks by deficient methylation of DNA. Low dietary folic acid has been associated with increased GC risk.73 The individual variability of key enzymes in folate metabolism, such as MTHFR, may modify the impact of dietary intake in cancer risk.

Two main polymorphisms, 677C>T and 1298A>C, have been described in the MTHFR gene. Homozygosity for both variant alleles is associated with decreased enzyme activity and lower blood folate levels, whereas heterozygotes (677CT) have about 65% of normal enzyme activity. In a case-control study in China,74 the 677TT genotype was associated with increased risk of GC (OR = 1.9) compared with the 677CC genotype, an effect that was stronger (OR = 2.5) for gastric cardia cancer (Table III). No association was observed with the 1298A>C polymorphisms. It may be hypothesized that 677TT homozygotes (12% of Caucasian populations) and 1298CC homozygotes (10%), as well as double heterozygotes (13% of the population), might be protected when folate intake is adequate but may have an increased risk of gastrointestinal cancer when folate intake is low.75 Unfortunately, no information on vegetable, fruit or folate intake was available from the Chinese study.

Table III. Case-Control Studies on Gastric Cancer Risk and Polymorphisms in Genes Related to Oxidative Damage, DNA Repair and Oncogenes
Gene and variantAuthor and countryStudy subjectsCovariatesResults2 (OR and 95% CI)Histologic typesAnatomic sitesInteractionObservations
No. of casesNo. of controlsStudy base
  • NA, not available; SS, statistically significant; OR, odds ratio; CI, confidence interval.

  • 2

    Allele and genotype nomenclatures according to authors.

Oxidative damage:
 MTHFR 677C>T  1298A>CShen et al.74 China187166PopulationAge, sex, residential area, smoking, alcohol, Hp, family history677TT vs. 677CC = 1.87 (1.0–3.5) 1298A>C No associationNA677TT higher risk in cardia cancerSuggestion of higher risk for older, smokers, male and family history of cancerPrevalence of Hp infection and tobacco higher in controls
DNA repair:Lack of information on diet and folate
 XRCC1 26304C>T (R194W)  28152G>A (R399Q)Shen et al.77 China188166PopulationAge, sex, residence, family history, smoking, alcohol, Hp26304CC = 1.45 (0.93–2.25)NAXRCC1 26304CC higher risk in cardia cancerSuggestion of higher risk in female, nonsmokers, without family history of cancerSame study of reference (Shen, et al.74)
28152GA/AA = 1.53 (0.98–2.39)
Both = 1.73(1.12– 2.69)
OGG1 S326CShinmura et al.79 Japanese3542HealthyNoneNA (Ser326 vs. Cys326, no differences)NANANASmall size; case-control group not described
Lack of information on tobacco, diet, Hp infection
OGG1 S326CHanakoa et al.80 BrazilJapanese-Brazilians:HospitalAge, sex, dietary habits, smoking, alcohol, HpCys alleleNo differencesNANot foundSame study of reference (Nishimoto et al.55)
9696JBs = 1.01 (0.5–1.9)
Non Japanese-Brazilians:
236236NJBs = 0.85 (0.6–1.3)
 MYCL1 (L-myc) 2886 EcoRIKato et al.53 Japanese82151HospitalAge, sex, pepsinogen, HpL/L vs. S/S + L/SNo differencesNANASame study of reference (Kato et al.53)
Intest = 1.43 (0.5–4.0)
Diffuse = 1.23 (0.4–3.4)
 MYCL1 (L-myc) 2886 EcoRIKato et al.54Japan284284HospitalAge, sex, pepsinogen, HpL/L vs. L/S = 1.55 (1.03–2.34)No differencesNANASame study of reference (Kato et al.54)
L/L vs. S/S = 1.2 (0.75–1.93)
 MYCL1 (L-myc) 2886 EcoRIShibuta et al.84 Japan61107HealthyAge, sex, race, smokingLS + SS vs. LL = 3.09 (1.33–7.21)No differencesNANot foundSmall size; cases and control groups not described
Lack of information on diet, Hp infection
 MYCL1 (L-myc) 2886 EcoRIIshizaki et al.83 Japan60100HealthyNoneNA (Differences not SS, no association)NANANASmall size; cases and control groups not described
Lack of information on diet, smoking, Hp infection

Repair of DNA Damage

In recent years much attention has been paid to the potential role of variations in DNA repair capability and cancer risk. An essential role for polymorphisms in DNA repair enzymes in the susceptibility to cancer has been postulated.76 DNA repair is a complex process involving a variety of mechanisms (i.e., nucleotide excision repair, double-strand break repair, mismatch repair, removal of modified bases such as methyl groups) that may contribute in a distinct manner to the genome homeostasis. A low number of studies have addressed the role in GC risk of individual variability in DNA repair due to polymorphisms in DNA repair genes.


Two polymorphisms (XRCC1 26304C>T or R194W and XRCC1 28152G>A or R399Q) of as yet unknown functional roles have been described. One Chinese study has analyzed their association with GC.77 In that Chinese population, the frequency of the 26304 T allele was 34.6%, and that of the 28152 A allele was 25.6%. Only individuals homozygous for both putative high-risk variants showed a statistically significant increased risk of GC (OR = 1.73), which was higher for gastric cardia (OR = 2.18). Both genotypes were associated with high risk in nonsmokers, in non-alcohol drinkers, in individuals not infected with Hp and in those without a family history of GC.


O6-methylguanine-DNA methyltransferase (MGMT) is a DNA repair protein that removes mutagenic and cytotoxic adducts from O6-guanine, an important step in the formation of mutations–mainly G:C to A:T transitions—after exposure to methylating agents such as nitrosamines. Recently, a polymorphism at exon 5, codon 160 of the MGMT gene (G160R), with an as yet unknown functional role, has been described. The MGMT variant allele has been associated with an earlier age of diagnosis.78


8-Oxoguanine (8-oxoG) is a base modification, usually induced by ROS, causing G:C to T:A transitions. At least 2 types of 8-oxoguanine-DNA glycosylases (OGG), capable of 8-oxoG repair, are expressed, with OGG1 being the major enzyme class. A functional S326C polymorphism of OGG1 has been identified. In Japan, the variant allele was higher in GC cases than in healthy individuals,79 although the association did not reach statistical significance. No differences regarding GC risk were observed in Japanese-Brazilian and non-Japanese Brazilian populations.80

Other Repair Enzymes.

A number of variants of other DNA repair enzymes may influence cancer risk. However no studies have been performed so far. Microsatellite instability (MSI), secondary to deficient function of the DNA mismatch repair system, is present in 27–40% of GC cases. MSI phenotype conditions gastric tumorigenesis81 and seems to be associated with specific diet patterns.82 No studies have thus far addressed the role in GC risk of described variant alleles in DNA mismatch repair system genes. ADP-ribosyltransferase (ADPRT, also known as PARP) is involved in DNA repair, and its downregulation results in genetic instability. A polymorphism of unknown functional relevance has raised interest as a putative risk factor for cancer, but no studies have yet been performed regarding GC risk.

Oncogenes and Tumor Suppressor Genes


Also known as L-myc, MYCL1 is a proto-oncogene involved in the control of cell proliferation. An EcoRI restriction site at nucleotide position 2886 in intron 2 allows the recognition of 2 variant alleles: S (short) and L (long), with 3 genotypes (LL/LS/SS). The presence of the short allele (S/S or S/L) has been correlated with an increased aggressiveness of the tumor, including poor prognosis and metastasis of GC.83, 84 A Japanese case-control study failed to show an association between short alleles and GC risk.83 On the contrary, in another case-control study in Japan, the presence of the S/S or S/L genotype was associated with a moderate increase in GC risk (OR = 1.55)53, 54 that was stronger for the intestinal type. Another Japanese study84 also showed a high increased risk (OR = 3.09) for the short alleles, with no differences according to histologic type.


The tumor protein 53 gene (TP53), a known tumor suppressor gene, plays a key role in DNA damage repair. The TP53 gene and protein aberrations are present in a high proportion of GC cases,85 being an early event in the intestinal type86 and associated with tumor progession in the diffuse type. Several TP53 polymorphisms have been identified, although no study has so far addressed the role of TP53 allelic variants in genetic susceptibility to GC.

Cadherin 1 (E-cadherin).

Cadherin 1 is an important mediator of homophilic recognition signals leading to cell-cell contact inhibition. Loss of cadherin 1 function is a key event in human gastric tumorigenesis. Germline mutations in the cadherin 1 gene (CDH1) have been described in hereditary GC; these confer a highly penetrant susceptibility to the diffuse type of GC, which is inherited as an autosomal dominant trait.3 In sporadic GCs, somatic mutations (mainly missense mutations and intragenic in-frame deletions) are very common in the diffuse and mixed types.87 Decreased expression without mutation is also seen in a proportion of intestinal cancers. Although several variant alleles have been described, no study has been published regarding their potential contribution to GC risk.


The role of genetic polymorphisms in GC risk has been of increasing interest in the last several years, probably because of advances in DNA analysis technologies and human genome knowledge. Of the 31 articles identified in our review, 14 were published during the years 2000 and 2001. Most of these studies (Tables I–III addressed the effect of genetic variants of metabolic enzymes (15 articles) and inflammation mediators (7 articles). This is probably because these variants have been characterized better, and also an increased cancer risk associated with their functions is plausible. The most widely studied polymorphism is the GSTM1 null allele. However, only a few studies have evaluated the risk of GC associated with genes that act in the areas of mucosa protection, oxidative damage and DNA repair.

The most consistent results (Table IV) are the increased GC risk associated with the IL1B and NAT1 variants.The finding that IL-1 gene cluster haplotypes (IL1B-31T+ and IL1RN*2/*2), which are thought to increase IL-1β production and inhibition of gastric acid secretion, are associated with increased risk of chronic hypochlorhydria and also GC risk27 supports the hypothesis that intensity and maintenance of inflammatory response to Hp infection may be relevant to early stages of gastric tumorigenesis.9 The approximately 40% attributable risk of GC for the polymorphism emphasizes the importance of the finding. Although penetrance of common variant alleles is usually low, their high prevalence makes it likely that they account for a significant fraction of the observed population risk. The potential protective role of variants at other genes involved in inflammatory responses (HLA-DQA1*0102 and TNF-238A) reinforces the contribution of immunogenetics to GC development.

Table IV. Genetic Polymorphisms found to be Associated with Gastric Cancer Risk1
Genes related toIncreased riskProtective effect
All2Only in IntestinalAll2Only in Intestinal
  • 1

    Data in parentheses are numbers of studies with positive results in relation to the total number of published studies.

  • 2

    In all gastric cancer cases.

Mucosa protectionMUC1 (1/1)
MUC6 (1/1)
Inflamatory responsesIL-1B (2/2)IL1RN*2(1/2)TNF(1/1)HLA (1/4)
HLA (1/4)
Metabolic enzymesGSTM1 null (4/9)NAT1*10(1/2)CYP2E1 (1/4)
GSTT1 null (1/5)
Oxidative damageMTHFR (1/1)
Repair DNA damageXRCC1 (1/1)
Oncogenes, tumor suppressor genesMYCL1 (2/4)

Individual differences in xenobiotic transformation are also of importance. The observed relationship between fast acetylators (NAT1*10) and increased GC risk supports a role in procarcinogen activation in human GC. The fact that NAT1 has been associated with activation of tobacco-related carcinogens further supports a role for smoking in GC development. The less clear protective effect of the RsaI variant of the CYP2E gene leaves open the possibility that variants of other enzymes involved in these processes may significantly modulate GC risk.

Most of the studies on the role of genetic polymorphisms and GC risk have provided inconsistent results. Disparity in findings among epidemiologic studies may result from limitations in design, the presence of confounding factors and potential sources of bias that may significantly affect their conclusions.88 Most studies have been carried out with a limited number of cases. (Seventeen of the 31 articles presented results with less than 99 cases.) In these studies, failure to detect associations between genetic variation and GC risk may be due to insufficient potency.89 This may be especially true if the putative effect of the genotype is modified by other environmental exposures.

In studies on metabolic polymorphisms, an additional limitation is the variable influence of the genotype in protein function. The expression of gene products may be induced or inhibited by several environmental exposures and other conditions (smoking, physical activity, drugs, alcohol, diet, etc.) that may also modify cancer risk by other mechanisms. It is noteworthy that metabolic gene variants may play a role in a subgroup of the population exposed to environmental carcinogens, i.e., in smokers. However, in the absence of any carcinogenic substrate, even a putatively high-risk allele would have no effect on risk.8 This important statement can be illustrated in the results observed in 3 of the studies reviewed. In the Chinese study,63 the GSTM1 null variant was not associated with GC risk in all subjects, but if we considered only subjects who smoked heavily, the OR was 5 (95% CI = 1.5–17.7). In the Polish study,64 the risk of GC associated with the the GSTT1 null genotype was 1.18 (95% CI = 0.6–2.4) in nonsmokers and 3.08 (95% CI = 1.5–6.5) in current smokers. In the Japanese study,68NAT1 genetic variation was not associated with GC risk in all subjects, but in heavy smokers the OR was 2.97 (95% CI = 1.23–7.14). Therefore it may not be appropriate to draw conclusions about the relative contribution of the genetic background to the observed cancer risk, without taking other cofactors into account.

The interaction between exogenous and endogenous factors is also relevant regarding the presence of oxidative species, which is putatively associated with chronic Hp infection or other carcinogens. Several lines of evidence strongly support the hypothesis that the presence of antioxidants, either exogenous (i.e., vitamin C, folates) or endogenous (i.e., superoxide dismutase), may be critical in gastric mucosa protection. Moreover, exogenous antioxidant levels do not only depend on intake. Variant alleles of MTHFR, a key enzyme in folic metabolism, exert a protective effect in GC. Unfortunately, the lack of concomitant information on folate intake and other cofactors prevents us from drawing more definitive conclusions.

Information is scarce on GC and DNA repair gene polymorphisms, an emerging area in the cancer susceptibility field. Only the XRCC1 and OGG1 variant alleles have been studied, with promising results for XRCC1. Studies addressing this issue should take into account a number of other relevant conditions.76 In case-control studies, tumor burden is a potentially confounding factor in the measurement of DNA repair. In addition, the potential impairment of DNA repair capability by chemotherapy should be considered. Finally, similar to what occurs with xenobiotic metabolism, other factors such as age or dietary habits may significantly influence enzyme activity.

Even so, in most of the articles published (n = 19), gene-environmental interactions were not considered. In 6 studies they were not found,55, 62, 65, 66, 80, 84 whereas in another 6, some level of interaction was observed.60, 63, 64, 68, 74, 77 Only a few studies assessed the effect of gene-gene interactions. An interaction was shown between IL-1B and IL-RN genetic variants,28 but no interaction was observed between GSTM1 and GSTT1 null genotypes.62, 63

It is noteworthy that significant limitations in studies published so far came from potential selection bias. No prospective cohort studies on genotypes and GC risk have been identified, and only 5 of the 25 case-control published studies were population-based. Most of the case-control studies were hospital based. Diseases included in hospital controls may be associated with specific genotypes, leading to selection bias. Moreover, mixture of populations has not usually been taken into account in study design.90

Several remarks regarding tumor histology and anatomic location should be made. It has been postulated that genetic pathways involved in intestinal and diffuse type GC tumorigenesis may differ,13 and it has been suggested that genetic background could be more important in GC of the diffuse type.12 However, only 528, 31, 32, 55, 68 of the 19 articles providing information on tumor histology showed some difference in allele distribution according to histologic type. Altogether, the accumulated evidence seems to disregard significant differences between tumor histology according to the genetic variants so far analyzed. Studies addressing the role of variants in genes shown to be distinctly activated according to tumor histology (i.e., E-cadherin)87 will be valuable in clarifying this issue.

GC of the cardia deserves a special comment because of its increasing incidence in developed countries,91 in spite of the decrease in incidence of gastric body and antral tumors. In fact, combined analysis of nested case-control studies has shown that Hp does not increase the risk of cardia cancer,92 suggesting that both tumor types may be associated, at least in part, with different risk factors. However, few studies19, 27, 29, 63, 64, 74, 77 have addressed the effect of individual genetic variability and anatomic site. Only 2 have observed a higher risk of cardia cancer associated with genetic variation of MTHFR74 and XRCC1.77 The low number of cardia cases available in most of the studies prevents us from making more conclusions on the specific role of genetic variation in cardia cancer.

The best scientific evidence on genetic factors associated with GC risk will be obtained from large cohort studies that take into account simultaneously the different factors potentially involved in gastric carcinogenesis: genetic variability, Hp infection and environmental exposure. Success in identifying inherited factors involved in modifying the risk of environmental factors will depend on the direct exploration of interactions between genes and environment.93 Furthermore, simultaneous analysis of different polymorphic genes should also be made, since most of the studies performed so far have analyzed only 1 gene, therefore excluding the possibility of identifying gene-gene interactions. The results of these types of studies will allow us to estimate the relative contribution of individual genetic variation to overall GC risk.