Abstract

To clarify the risk of gastric cancer associated with glutathione S-transferase T1 (GSTT1) status, a meta-analysis of published studies was performed. Eligible studies included all reports investigating an association between GSTT1 status and gastric cancer published before October 31, 2005. A qualitative scoring of papers was applied to evaluate the quality of the published data. The principal outcome measure was the odds ratio (OR) for the risk of gastric cancer associated with GSTT1 deletion status using a random effects model. Eighteen case–control studies detailing a possible association between the GSTT1 null genotype and gastric cancer were selected. Combining data from these studies, totalling 2508 cases and 4634 controls, a non-statistically significant OR for gastric cancer risk associated with GSTT1 deficiency emerged [OR = 1.09; 95% confidence interval (CI): 0.97–1.21; I2 = 0%]. When only high-quality scored studies were considered, a statistically significant increased risk appeared (OR = 1.23; 95% CI: 1.04–1.45; I2 = 0%), as well as considering only Caucasians (OR = 1.23; 95% CI: 1.03–1.56; I2 = 0%). By pooling data from seven studies (319 cases and 656 controls) that considered combinations of GSTT1 and GSTM1 genotypes, a statistically significant increased risk for gastric cancer (OR = 1.95, 95% CI: 1.42–2.67; I2 = 0%) was detected for individuals with deletion mutations in both genes compared with wild-types. In conclusion, this meta-analysis suggests that the GSTT1 null genotype may slightly increase the risk of gastric cancer and that interaction between unfavourable GST genotypes may exist. Greater attention should, therefore, be paid to the design of future studies; the investigation of interactions among multiple genotypes and environmental exposures are justified to clarify GSTT1 null status influence on gastric cancer risk.

Introduction

Sequence variations in genes coding for phase II enzymes, such as the glutathione S-transferase (GST) family, may potentially alter individual susceptibility to cancer (1). GST are a family of genes with a critical function in protecting against electrophiles and the products of oxidative stress. The GST enzymes are involved in the detoxification of many xenobiotics, including several environmental carcinogens and endogenously derived reactive oxygen species (2). Four major GST families are widely expressed in mammalian tissue: GSTA (α), GSTM (μ), GSTT (𝛉) and GSTP (π). Certain genes within the GSTM and GSTT (GSTM1 and GSTT1) subfamilies exhibit homozygous deletion (null genotype) polymorphisms that are considered important modifiers of individual risk for environmentally induced cancers (1). Individuals who have the homozygous deletion in one of these genes have no GSTM1 and GSTT1 enzyme activity, and thus are more susceptible to carcinogens such as benzo[α]pyrene-7,8-diol epoxide, the activated form of benzo[α]pyrene, and smaller reactive hydrocarbons, such as ethylene oxide and diepoxybutane (2,3). The prevalence of GSTM1 and GSTT1 null genotypes was found to vary among ethnic groups. In human populations, GSTM1 and GSTT1 are absent in 10–60 and 13–55% of individuals, respectively (4).

The common expression of GSTα, GSTT1-1 and GSTP1-1 in many cell types along the human gastrointestinal tract suggests an important role in the protection against carcinogens and other xenobiotics (5). The deletion mutations in the GSTT1 and GSTM1 genes and their association with gastric cancer have been investigated in a large number of studies. In our recent meta-analysis, which identified 15 studies investigating an association between GSTM1 null genotype and the risk of gastric cancer, we reported a slight increase in gastric cancer risk associated with GSTM1 deficiency [odds ratio (OR) of 1.24; 95% confidence interval (CI): 1.00–1.54] (6). A recent review of genetic susceptibility and gastric cancer risk reported that the results of case–control studies detailing associations between the GSTT1 gene and gastric cancer risk are inconclusive (7). Since Deakin et al. (8) first investigated the relationship between GSTT1 deficiency and gastric cancer in 1996, 17 studies have appeared in the literature, and most of them have refuted an association between GSTT1 deficiency and gastric cancer risk (3,924). One of the major problems with the published studies is that most of them were based on small numbers of cases and controls. Furthermore, because the GSTT1 genotype is presumed to affect gastric cancer risk by influencing detoxification of activated environmental carcinogens and by interaction with other unfavourable GST polymorphisms, the potential modifying effect of GSTT1 status on the relationship between tobacco smoking, other GST and gastric cancer is of particular interest, even though not often investigated.

To clarify the effect of GSTT1 status on the risk of developing gastric cancer, we carried out a quantitative meta-analysis of research published through October 31, 2005. In addition, we combined data available from published papers to explore the possible effects of the interactions between the GSTT1 genotype and smoking habits and between GSTT1 and GSTM1 genotypes with respect to gastric cancer risk.

Materials and methods

Identification of relevant studies

The digital medical databases used for the search were MEDLINE and EMBASE. The key words used for the research were glutathione S-transferase T1 or GSTT1, gastric or stomach, cancer, without restriction on language. The time period includes research articles published up to October 31, 2005.

For the meta-analysis, the following inclusion criteria were considered: We compared the results of our literature search to the review articles found using the previously mentioned databases. Furthermore, when data from one paper was republished by the same author in a larger investigation or written in English, only the most recent article was considered.

  1. Presence of a quantitative assessment of the relationship between GSTT1 status and gastric cancer.

  2. An appropriate description of GSTT1 status in cases and controls.

  3. Results expressed as relative risk (RR) or OR.

  4. Studies with a 95% CI for RR or OR, or with the possibility to calculate these measures if standard deviation (SD) values were present.

Quality assessment and data extraction

Each article was blinded with respect to the authors, institutions and journals. The articles were read and scored for quality by two independent researchers using a system that incorporates elements of the methods developed by Angelillo et al. (25) and Chalmers et al. (26). We also followed suggestions useful to evaluate a molecular epidemiological study from Thakkinstian et al. (27) and Bogardus et al. (28). The criteria employed are shown in the results section below. A quality score was then calculated for each paper as the percentage of applicable criteria that were met in each study. Items concerned with efforts to minimize potential bias (nine points, items A–I, results section) were given twice the weight of those evaluating data analysis (nine points, items J–R). Lastly, even though reported in the quality score scale, the item related to Hardy–Weinberg equilibrium (item R) was not applicable because it could not have been checked in the included studies because of the analytical method used for GSTT1 genotyping, which does not provide the frequency of heterozygous individuals. Therefore, high-quality scored studies were considered as the ones with at least 70% of the total score (19 out of 26).

The same two researchers extracted the data from each article using a structured sheet and entered into a database. The following data were considered: year and location of study; ethnicity, source, sex ratio and mean age (or range whenever possible) of cases and controls; the number of cases and controls with GSTT1 deficiency; OR values with their 95% CI and covariates investigated in the study.

Statistical analysis

The OR for gastric cancer associated with GSTT1 deficiency was estimated for each study. In carrying out the meta-analysis, the random effect model was used, taking into account the possibility of heterogeneity between studies, which was tested with I2 test and a standard χ2-test (29). The resulting P-value of the χ2-test for heterogeneity is reported in the result section after the result of the I2 test. The Mantel–Haenszel method (fixed effect model) was also used to assess the effect of the model's assumptions on our conclusions (30,31). Statistical analysis was undertaken using the RevMan program, release 4.2 (32). In order to detect potential publication bias, ORs and 95% CI were plotted against standard errors in each study. We also computed the power of the selected studies, in order to assess the probability of detecting an association between GSTT1 deficiency and gastric cancer at the 0.05 level of significance, assuming a genotypic risk of 2.0 and 1.5, using the method described by Schlesselmann (33). On the basis of previous literature findings, we planned to perform several subgroup meta-analyses based on the ethnicity, study design, power of the study and quality score of the papers. Lastly, we performed two additional sensitivity analyses in an attempt to evaluate whether the interaction between GSTT1 and cigarette smoking, as well as between GSTT1 and GSTM1 null genotypes, can modify the risk of gastric cancer. The z statistic was used to formally assess if any statistically significant difference among the results of each subgroup meta-analyses exists, as reported by Deeks et al. (34). In order to collect the most complete data concerning those interactions, when not extensively published in the results section of a paper, the corresponding author of the individual studies were contacted by e-mail and fax. We thus invited those authors to provide data useful for us in performing the two subgroup meta-analyses.

Results

Identification of relevant studies

Eighteen articles were retrieved by our bibliographic search, with three papers written in non-English language (10,17,20). In order to collect data from the last three papers, the corresponding authors were contacted by e-mail and fax and invited to fill in an empty table with the results of their studies. All the authors answered and their data were included in the analysis. Eighteen case–control studies were therefore considered for the present meta-analysis (3,824). In Table I the ORs are reported with their corresponding 95% CI, including all the data that were extracted from each study. In each report, GSTT1 status was determined by analysis of the gene via polymerase chain reaction. Ten case–control studies were population-based (C-C pb), with six of them enrolling sex- and age-matched controls, and eight were hospital-based case–control studies (C-C hb), of which only three had sex- and age-matched controls (Table I). History of smoking was verified for cases and controls in 12 of 18 studies, with 4 reporting results of the interaction between GSTT1 status and gastric cancer risk in relation to cigarette smoking habits (11,12,15,21). Lastly, though 17 of 18 studies collected data on GSTM1 status, only 5 reported data concerning the combination of those genotypes with respect to gastric cancer risk in a form suitable for a subgroup meta-analysis (3,12,19,21,22). Authors were contacted to obtain data on the interactions between GSTT1 status and smoking habits, and GSTT1 status and GSTM1 status; two authors answered (10,20) and their data were included in the two subgroup meta-analyses.

Table I.

Summary of studies of gastric cancer and GSTT1 status

Investigator, year/place of study
 
Casesa
 
GSTT1 null (%)
 
Controls
 
GSTT1 null (%)
 
OR (95% CI)b,c for null genotype
 
Covariates
 
Power (%) (RR > 1.5; α = 0.05)
 
Power (%) (RR > 2.0; α = 0.05)
 
Deakin et al., 1996 (8)/Staffordshire, UK 114 cases from one hospital; age, 68; 70.5% male; Caucasians 18.4 509 controls from the same hospital of cases; age, 70; 48% male; Caucasians 18.5 1.00 (0.59–1.68)b Age, gender, grading, histotype, smoking, GSTM1 status 52 100 
Katoh et al., 1996 (9)/Kitakyushu, Japan 139 consecutive cases from three hospitals in the same city; histologically confirmed; age, 62.2 (13.2); 70.5% male; Asians 47.5 126 sex- and age-matched hospital controls from the same city of cases; age, 61.9 (16.8); 57.1% male; Asians 44.4 1.13 (0.70–1.83)b Age, gender, grading, occupation,area of residence, medical history, smoking habits, GSTM1 status 33 92 
Wang et al., 1998 (10)/Taiwan 83 cases; age, 59.6 (14.1); 60.2% male; Asians 43.4 83 controls from the same hospital of cases; age, 56.7 (14.7); 60.2% male; Asians 54.2 0.65 (0.35–1.19) Age, gender, H.pylori infection, smoking habits, CYP2E1 and GSTM1 status 79 100 
Setiawan et al., 2000 (11)/Yangzhong City, China 81 incident cases; histologically confirmed diagnosis; age, NS; 71.4% male; Asians 54.0 418 healthy population-based cancer-free individuals; age, NS;49.7% male; Asians 46.0 2.50 (1.01–6.22)c Age, gender, BMI,medical history, education, occupation, family history of cancer, fruit and salt intake, alcohol and green tea consumption, H.pylori infection, smoking habits, GSTM1 status 36 100 
Saadat et al., 2001(3)/Shiraz, Iran 42 cases; histologically confirmed diagnosis; age, NS; sex ratio, NS; Caucasians 35.7 131 sex- and age-matched healthy blood donors; age, NS; sex ratio, NS; Caucasians 31.3 1.22 (0.58–2.57)b Age, gender, GSTM1 status 16 45 
Lan et al., 2001 (12)/Warsaw, Poland 293 incident cases from 22 hospitals of one city; ages, 21–79; 65.8% male; Caucasians 20.4 418 sex- and age-matched healthy population-based controls from the same city of the cases; ages, 21–79; 64.4% male; Caucasians 15.7 1.48 (0.97–2.25)c Age, gender, education, area of residence, BMI, H.pylori infection, family history, fruit consumption, smoking habits, IL-1, GSTM1, GSTM3 and GSTP1 status 100 100 
Cai et al., 2001 (13)/Changle County, China 95 incident cases; histologically confirmed diagnosis; ages, 32–78; 85.3% male; Asians 43.2 94 sex- and age-matched healthy population controls from the same county of the cases; ages, 37–79; 87.2% male; Asians 50.0 0.76 (0.1–1.4)b Age, gender, education, fish sauce consumption, alcohol, smoking habits, GSTM1 status 23 66 
Wu et al., 2002 (14)/Taipei, Taiwan 356 incident cases from one hospital; histologically confirmed diagnosis; ages, 25–87; 61.2% male; Asians 50.8 278 controls from the same hospital of cases; ages, 22–86; 56.1% male; Asians 46.7 1.18 (0.86–1.61)b Age, gender, histotype, anatomy, stage, GSTM1 and CYP2E1 status (Rsa/Pst and Dra77 100 
Gao et al., 2002 (15)/Huaian City, China 153 cases (hospital surgical cases and incident cases from a regional registry) from the same city; histologically confirmed diagnosis; ages, 40–81; 77.1% male; Asians 46.4 223 sex- and age-matched healthy population controls from the same city of the cases; ages, 35–81; 66.8% male; Asians 53.3 0.77 (0.48–1.25)c Age, gender, diet, tea and alcohol consumption, smoking habits, GSTM1 status 41 100 
Sgambato et al., 2002 (16)/Basilicata, Italy 8 consecutive cases from one hospital; age, NS; sex ratio, NS; Caucasians 0.0 100 healthy controls fromthe same hospital of cases; age, NS; sex ratio, NS; Caucasians 18.0 0.26 (0.01–4.75)b Age, gender, occupation, smoking history, GSTM1 status 11 
Ye et al., 2003 (17)/Wuhan, China 56 cases from two hospitals; histologically confirmed diagnosis; age, 57.6 (22–79); 75% male; Asians 60.7 56 healthy controls from one of the two hospitals of cases; matched for sex, age, smoking, dietary habits and family history of cancer; age, 58.0 (26–86); 69.6% male; Asians 46.4 1.78 (0.84–3.78)b Age, gender, education, occupation, living condition, family history of cancer, dietary habits, drinking, smoking and CYP2E1 status 18 42 
Choi et al., 2003 (18)/Ikasan, South Korea 80 surgical cases from one hospital; histologically confirmed diagnosis; age, NS; sex ratio, NS; Asians 53.8 177 healthy cancer-free individuals; age, NS; sex ratio: NS; Asians 53.1 0.97 (0.55–1.71)b Stage, grading, histotype, GSTM1 status 26 71 
Colombo et al., 2004 (19)/Saõ Jose do Rio Preto and Barretos, Brazil 100 incident cases from two hospitals; histologically confirmed diagnosis; ages, 28–93; 73% male; 87 Caucasians, 13 Negroids 17.0 150 sex- and age-matched healthy population-based controls; ages, 20–93; 60% male; 135 Caucasians, 15 Negroids 18.6 0.89 (0.46–1.73)b Age, gender, family history, occupation, ethnicity, histotype, H.pylori infection, alcohol consumption, smoking habits, GSTM1 and CYP2E1 status 29.4 83.3 
Torres et al., 2004 (20) / Popayan, Colombia 46 cases from one hospital; histologically confirmed diagnosis; age, 60; 50% male; Caucasians 17.4 96 cancer-free controls from the same hospital of cases; age, 58; 42.7% male; Caucasians 14.6 0.47 (0.09–2.27)c Age, gender, area of residence, family history, occupation, H.pylori infection, diet, alcohol consumption, smoking habits, GSTM1 and TNF status 17.9 41.4 
Tamer et al., 2004 (21) / Mersin and Kocaceli, Turkey 70 surgical cases from two hospitals; histologically confirmed diagnosis; age, 57.6 (9.8); 67.1% male; Caucasians 30.0 204 healthy cancer-free individuals; ages, 62.0 (7.0); 56.4% male; Caucasians 26.0 1.36 (0.93–2.94)c Age, gender, smoking habits, GSTM1 and GSTP1 status 25.2 74.2 
Palli et al., 2005 (22) / Tuscany, Italy 175 surgical cases from the main hospitals of the area; histologically confirmed diagnosis; age, 68.7 (10.4); 57.1% male; Caucasians 23.4 546 healthy population-based randomly selected controls from the same area of cases; age, 55.5 (7.0); 49.3% male; Caucasians 16.7 1.68 (1.01–2.80)c Age, gender, area of residence, family history of gastric cancer, H.pylori infection, GSTM1 status 42.7 100 
Mu et al., 2005 (23) / Taixing City, China 196 cases from Taixing Tumour Registry; age, NS; 66.99% male; Asians 47.4 393 sex- and age-matched healthy population-based randomly selected controls from the same area of cases; age, NS; 69.16% male; Asians 48.8 1.12 (0.72–2.74)c Age, gender, education, area of residence, income, BMI, very hot food eating habits, family history of stomach disease and gastric cancer, H.pylori infection, alcohol and tea habits, smokinghistory, p53, GSTP1, GSTM1 status 64.6 100 
Nan et al., 2005 (24) / Cheongju and Daejon, South Korea 421 cases from two hospitals; histologically confirmed diagnosis; age, 60.0 (11.2); 65.6% male; Asians 42.7 632 sex- and age-matched hospital-based cancer-free individuals from the same hospitals of cases; age, 59.4 (10.7); 65.5% male; Asians 40.2 1.11 (0.86–1.44)b Age, gender, diet, GSTM1 status 100 100 
Investigator, year/place of study
 
Casesa
 
GSTT1 null (%)
 
Controls
 
GSTT1 null (%)
 
OR (95% CI)b,c for null genotype
 
Covariates
 
Power (%) (RR > 1.5; α = 0.05)
 
Power (%) (RR > 2.0; α = 0.05)
 
Deakin et al., 1996 (8)/Staffordshire, UK 114 cases from one hospital; age, 68; 70.5% male; Caucasians 18.4 509 controls from the same hospital of cases; age, 70; 48% male; Caucasians 18.5 1.00 (0.59–1.68)b Age, gender, grading, histotype, smoking, GSTM1 status 52 100 
Katoh et al., 1996 (9)/Kitakyushu, Japan 139 consecutive cases from three hospitals in the same city; histologically confirmed; age, 62.2 (13.2); 70.5% male; Asians 47.5 126 sex- and age-matched hospital controls from the same city of cases; age, 61.9 (16.8); 57.1% male; Asians 44.4 1.13 (0.70–1.83)b Age, gender, grading, occupation,area of residence, medical history, smoking habits, GSTM1 status 33 92 
Wang et al., 1998 (10)/Taiwan 83 cases; age, 59.6 (14.1); 60.2% male; Asians 43.4 83 controls from the same hospital of cases; age, 56.7 (14.7); 60.2% male; Asians 54.2 0.65 (0.35–1.19) Age, gender, H.pylori infection, smoking habits, CYP2E1 and GSTM1 status 79 100 
Setiawan et al., 2000 (11)/Yangzhong City, China 81 incident cases; histologically confirmed diagnosis; age, NS; 71.4% male; Asians 54.0 418 healthy population-based cancer-free individuals; age, NS;49.7% male; Asians 46.0 2.50 (1.01–6.22)c Age, gender, BMI,medical history, education, occupation, family history of cancer, fruit and salt intake, alcohol and green tea consumption, H.pylori infection, smoking habits, GSTM1 status 36 100 
Saadat et al., 2001(3)/Shiraz, Iran 42 cases; histologically confirmed diagnosis; age, NS; sex ratio, NS; Caucasians 35.7 131 sex- and age-matched healthy blood donors; age, NS; sex ratio, NS; Caucasians 31.3 1.22 (0.58–2.57)b Age, gender, GSTM1 status 16 45 
Lan et al., 2001 (12)/Warsaw, Poland 293 incident cases from 22 hospitals of one city; ages, 21–79; 65.8% male; Caucasians 20.4 418 sex- and age-matched healthy population-based controls from the same city of the cases; ages, 21–79; 64.4% male; Caucasians 15.7 1.48 (0.97–2.25)c Age, gender, education, area of residence, BMI, H.pylori infection, family history, fruit consumption, smoking habits, IL-1, GSTM1, GSTM3 and GSTP1 status 100 100 
Cai et al., 2001 (13)/Changle County, China 95 incident cases; histologically confirmed diagnosis; ages, 32–78; 85.3% male; Asians 43.2 94 sex- and age-matched healthy population controls from the same county of the cases; ages, 37–79; 87.2% male; Asians 50.0 0.76 (0.1–1.4)b Age, gender, education, fish sauce consumption, alcohol, smoking habits, GSTM1 status 23 66 
Wu et al., 2002 (14)/Taipei, Taiwan 356 incident cases from one hospital; histologically confirmed diagnosis; ages, 25–87; 61.2% male; Asians 50.8 278 controls from the same hospital of cases; ages, 22–86; 56.1% male; Asians 46.7 1.18 (0.86–1.61)b Age, gender, histotype, anatomy, stage, GSTM1 and CYP2E1 status (Rsa/Pst and Dra77 100 
Gao et al., 2002 (15)/Huaian City, China 153 cases (hospital surgical cases and incident cases from a regional registry) from the same city; histologically confirmed diagnosis; ages, 40–81; 77.1% male; Asians 46.4 223 sex- and age-matched healthy population controls from the same city of the cases; ages, 35–81; 66.8% male; Asians 53.3 0.77 (0.48–1.25)c Age, gender, diet, tea and alcohol consumption, smoking habits, GSTM1 status 41 100 
Sgambato et al., 2002 (16)/Basilicata, Italy 8 consecutive cases from one hospital; age, NS; sex ratio, NS; Caucasians 0.0 100 healthy controls fromthe same hospital of cases; age, NS; sex ratio, NS; Caucasians 18.0 0.26 (0.01–4.75)b Age, gender, occupation, smoking history, GSTM1 status 11 
Ye et al., 2003 (17)/Wuhan, China 56 cases from two hospitals; histologically confirmed diagnosis; age, 57.6 (22–79); 75% male; Asians 60.7 56 healthy controls from one of the two hospitals of cases; matched for sex, age, smoking, dietary habits and family history of cancer; age, 58.0 (26–86); 69.6% male; Asians 46.4 1.78 (0.84–3.78)b Age, gender, education, occupation, living condition, family history of cancer, dietary habits, drinking, smoking and CYP2E1 status 18 42 
Choi et al., 2003 (18)/Ikasan, South Korea 80 surgical cases from one hospital; histologically confirmed diagnosis; age, NS; sex ratio, NS; Asians 53.8 177 healthy cancer-free individuals; age, NS; sex ratio: NS; Asians 53.1 0.97 (0.55–1.71)b Stage, grading, histotype, GSTM1 status 26 71 
Colombo et al., 2004 (19)/Saõ Jose do Rio Preto and Barretos, Brazil 100 incident cases from two hospitals; histologically confirmed diagnosis; ages, 28–93; 73% male; 87 Caucasians, 13 Negroids 17.0 150 sex- and age-matched healthy population-based controls; ages, 20–93; 60% male; 135 Caucasians, 15 Negroids 18.6 0.89 (0.46–1.73)b Age, gender, family history, occupation, ethnicity, histotype, H.pylori infection, alcohol consumption, smoking habits, GSTM1 and CYP2E1 status 29.4 83.3 
Torres et al., 2004 (20) / Popayan, Colombia 46 cases from one hospital; histologically confirmed diagnosis; age, 60; 50% male; Caucasians 17.4 96 cancer-free controls from the same hospital of cases; age, 58; 42.7% male; Caucasians 14.6 0.47 (0.09–2.27)c Age, gender, area of residence, family history, occupation, H.pylori infection, diet, alcohol consumption, smoking habits, GSTM1 and TNF status 17.9 41.4 
Tamer et al., 2004 (21) / Mersin and Kocaceli, Turkey 70 surgical cases from two hospitals; histologically confirmed diagnosis; age, 57.6 (9.8); 67.1% male; Caucasians 30.0 204 healthy cancer-free individuals; ages, 62.0 (7.0); 56.4% male; Caucasians 26.0 1.36 (0.93–2.94)c Age, gender, smoking habits, GSTM1 and GSTP1 status 25.2 74.2 
Palli et al., 2005 (22) / Tuscany, Italy 175 surgical cases from the main hospitals of the area; histologically confirmed diagnosis; age, 68.7 (10.4); 57.1% male; Caucasians 23.4 546 healthy population-based randomly selected controls from the same area of cases; age, 55.5 (7.0); 49.3% male; Caucasians 16.7 1.68 (1.01–2.80)c Age, gender, area of residence, family history of gastric cancer, H.pylori infection, GSTM1 status 42.7 100 
Mu et al., 2005 (23) / Taixing City, China 196 cases from Taixing Tumour Registry; age, NS; 66.99% male; Asians 47.4 393 sex- and age-matched healthy population-based randomly selected controls from the same area of cases; age, NS; 69.16% male; Asians 48.8 1.12 (0.72–2.74)c Age, gender, education, area of residence, income, BMI, very hot food eating habits, family history of stomach disease and gastric cancer, H.pylori infection, alcohol and tea habits, smokinghistory, p53, GSTP1, GSTM1 status 64.6 100 
Nan et al., 2005 (24) / Cheongju and Daejon, South Korea 421 cases from two hospitals; histologically confirmed diagnosis; age, 60.0 (11.2); 65.6% male; Asians 42.7 632 sex- and age-matched hospital-based cancer-free individuals from the same hospitals of cases; age, 59.4 (10.7); 65.5% male; Asians 40.2 1.11 (0.86–1.44)b Age, gender, diet, GSTM1 status 100 100 
a

Ages of cases and controls: mean (SD) or range given wherever possible.

b

Crude OR.

c

Adjusted OR.

Quality assessment

Table II shows the quality scoring items with the relative percentages of the studies complying with those criteria. The quality scoring procedure was performed for all the studies included in the meta-analysis, with the exception of the articles written in a non-English language. The potential for selection bias may be a concern in all of the individual studies considered. Cases and controls were chosen in most of the reports in the appropriate manner: 5 of 15 studies used cancer registries to identify cases and 7 identified randomly selected cases. Most of the studies stated that cancer diagnosis was validated by histology and all the studies specified disease criteria (Table II). Furthermore, in all of the studies the controls were drawn from the same population as the cases; 10 reports selected population controls. However, the response rates for cases and controls were frequently <75%, which may have led to a selection bias. Misclassification bias should be considered as a concern: only 2 of 15 studies clearly stated that the analyst was unaware of the clinical status of the subjects when genotyping their samples, so exposure misclassification may exist (Table II). One other issue is that almost all the studies did not report the reproducibility of the laboratory method used. Age, sex and other potential confounders/effect modifiers were considered in almost all of the studies, while demographic data were reported in only seven reports. Statistical analysis was judged appropriate in all of the studies, and P-values and/or 95% CI were always listed, while no study provided power calculations. Quality scores for the individual studies ranged from 0.46 to 0.81, and 5 of 15 resulted as high-quality scored studies (11,12,14,22,23).

Table II.

Items used in the quality scoring of observational studies in epidemiology


 
Quality scoring items
 
% of studies complyinga
 
Case–control studies   
A1 Cases either randomly selected or selected to include all cases in a specific population 80 
B1 Response rate for identified cases >75% 40 
C1 Controls drawn from the same population of cases 100 
D1 Population-based controls 67 
E1 Response rate for identified controls >75% 53 
Cohort studies   
A2 Initial response rate >75% NAb 
B2 Comparison of persons who did and did not participate NA 
C2 Follow-up rate >75% NA 
D2 Comparison of who were and were not lost to follow-up NA 
E2 Exposed/non-exposed subjects identified without knowledge of disease status NA 
All studies   
Disease validated by histology or other gold standard 69 
Specific disease/not disease or exposure/not exposure criteria given 100 
Exposure/disease assessment made blindly with respect to the case–control/exposure status of subjects (analyst unaware of the clinical status/exposure of the sample)c 13 
Reproducibility of laboratory tests mentionedc 13 
Age and gender considered as potential confounders 93 
Collection of data on other potential confounders/effect modifiers 100 
Demographic data listed 47 
Place and time of the study reported 93 
Power calculations performed 
Precise P-values or confidence interval given 100 
Statistic test specified 100 
Appropriate statistical analysis 100 
Hardy–Weinberg equilibrium assessedc NA 

 
Quality scoring items
 
% of studies complyinga
 
Case–control studies   
A1 Cases either randomly selected or selected to include all cases in a specific population 80 
B1 Response rate for identified cases >75% 40 
C1 Controls drawn from the same population of cases 100 
D1 Population-based controls 67 
E1 Response rate for identified controls >75% 53 
Cohort studies   
A2 Initial response rate >75% NAb 
B2 Comparison of persons who did and did not participate NA 
C2 Follow-up rate >75% NA 
D2 Comparison of who were and were not lost to follow-up NA 
E2 Exposed/non-exposed subjects identified without knowledge of disease status NA 
All studies   
Disease validated by histology or other gold standard 69 
Specific disease/not disease or exposure/not exposure criteria given 100 
Exposure/disease assessment made blindly with respect to the case–control/exposure status of subjects (analyst unaware of the clinical status/exposure of the sample)c 13 
Reproducibility of laboratory tests mentionedc 13 
Age and gender considered as potential confounders 93 
Collection of data on other potential confounders/effect modifiers 100 
Demographic data listed 47 
Place and time of the study reported 93 
Power calculations performed 
Precise P-values or confidence interval given 100 
Statistic test specified 100 
Appropriate statistical analysis 100 
Hardy–Weinberg equilibrium assessedc NA 
a

If compliance was not specifically indicated in the text, non-compliance was assumed.

b

NA: not applicable in the present meta-analysis; see Materials and methods and Results section.

c

Specific for molecular-epidemiological studies.

Statistical analysis

Figure 1 shows the results of the combined data, depicting a plot of ORs (95% CI) for the risk of developing gastric cancer associated with GSTT1 deficiency in the 18 case–control studies, involving a total of 7142 subjects (2508 cases and 4634 controls). The meta-analysis resulted in a statistically non-significant association between GSTT1 deficiency and gastric cancer risk (OR = 1.09; 95% CI: 0.97–1.21; I2 = 0%, P for heterogeneity = 0.48). This analysis is based on combining data from studies based on a number of different ethnic groups. Separate meta-analyses were therefore conducted stratified by ethnicity. For Asians (10 studies) we found an OR = 1.02 (95% CI: 0.89–1.18; I2 = 12.1%, P for heterogeneity = 0.33), whereas for Caucasians (8 studies) the OR value was 1.27 (95% CI: 1.03–1.56; I2 = 0%, P for heterogeneity = 0.87; P-value of z statistic among the two groups = 0.08) (Fig. 1). By considering separately C-C hb and C-C pb studies, similar OR values were found [C-C hb: OR of 1.08 (95% CI : 0.92–1.27; I2 = 0%, P for heterogeneity = 0.56); C-C pb: OR of 1.09 (95% CI: 0.92–1.29; I2 = 16.7%, P for heterogeneity = 0.29)]. When combining data from the studies with a power of at least 80% for a RR = 2.0 (Table I), we found an OR of 1.08 (95% CI: 0.95–1.24; I2 = 17.9%, P for heterogeneity = 0.27). In addition, when only high-quality papers were considered, a statistically significant increase in risk for gastric cancer appeared (OR = 1.23, 95% CI: 1.04–1.45; I2 = 0%, P for heterogeneity = 0.41), while an OR of 0.99 (95% CI: 0.85–1.15; I2 = 0%, P for heterogeneity = 0.75; P-value of z statistic among the two groups = 0.07) resulted by pooling data from low-quality papers (Fig. 1).

Fig. 1.

Forest plot of developing gastric cancer associated with GSTT1 null status and subgroup analyses.

Fig. 1.

Forest plot of developing gastric cancer associated with GSTT1 null status and subgroup analyses.

The funnel plot (Figure 2) shows no evidence of publication bias, with the exception of one outlier (16). As far as the interaction between GSTT1 status and smoking habits is concerned, in relation to gastric cancer risk, we obtained data from 6 of the 12 studies in a form suitable for a subgroup meta-analysis. After stratification by smoking status, an overall OR of 1.54 (95% CI: 0.95–2.48; I2 = 59.2%, P for heterogeneity = 0.03) for the risk of gastric cancer appears in ever-smokers with GSTT1 deficiency compared with ever-smokers with a normal GSTT1 genotype. On the other hand, an overall estimate of OR = 0.86 (95% CI: 0.59–1.27; I2 = 26.3%, P for heterogeneity = 0.24; P-value of z statistic among the two groups = 0.006) appears when comparing the effect of GSTT1 status (null versus normal genotype) in never-smokers with respect to gastric cancer risk (Fig. 1). Furthermore, pooling data concerning the combination of GSTT1 and GSTM1 genotypes from seven studies demonstrated that individuals with combined deletion mutations in those genes have an OR of 1.95 for gastric cancer (95% CI: 1.42–2.67; I2 = 0%, P for heterogeneity = 0.58) in comparison with individuals with wild-type genotypes.

Fig. 2.

Funnel plot of the published studies considered in the meta-analysis.

Fig. 2.

Funnel plot of the published studies considered in the meta-analysis.

Discussion

Most of the cancer susceptibility genes identified to date are rare and highly penetrant. While the individuals with rare alterations of these genes (e.g. tumor suppressor genes) have a dramatically higher risk of cancer, more common differences in low penetrant susceptibility genes (e.g. xenobiotic metabolism genes) could be responsible for a relatively small, but observable increase of gastric cancer risk at the population level (35). Although gastric cancer is one of the most common malignancies worldwide, its pathogenesis and the molecular genetic events that contribute to its development are poorly understood (7). One of the most widely studied metabolic polymorphisms examined as a susceptibility factor for gastric cancer is the GSTT1 null allele; its frequency in the Caucasian population is ∼13–26% (4). Since Deakin et al. (8) first investigated a possible relationship between GSTT1 deficiency and gastric cancer risk, a further 17 reports have been published examining this hypothesis (3,924), with conflicting results.

This led us to undertake the present meta-analysis, which aims to derive an estimate of the gastric cancer risk associated with GSTT1 status. The main finding of this meta-analysis of 18 case–control studies involving 7142 subjects is that the GSTT1 null status seems to be unrelated to gastric cancer risk. Absence of heterogeneity between studies emerged from the statistical analysis; however, as Blettner et al. (36) pointed out, the tests formally used to assess heterogeneity have low statistical power to detect it. Therefore, we decided a priori to perform several subgroup meta-analyses according to ethnicity, study design, quality and power of the studies. Considering separately Caucasian and Asian studies, the association between GSTT1 null status and gastric cancer risk reaches a slightly statistically significant level among Caucasians. In human populations, the frequency of GSTT1 deficiency is 13–26% and 36–52% in Caucasian and Asian individuals, respectively (4). However, the Caucasian reports in the subgroup analysis include a mixture of populations from very distant countries, so the result must be interpreted with caution. Unexpectedly, when only high-quality scored studies were considered, a statistically significant increased risk of gastric cancer for GSTT1 null individuals was detected.

Since GSTT1 is presumed to confer susceptibility to gastric cancer via an interaction with carcinogens, it is interesting to note that no data was collected from the cases or the controls on tobacco usage, alcohol intake, food consumption or Helicobacter pylori infection in many of the studies. By combining the data available from six studies, which investigated the possible interaction between GSTT1 status and cigarette smoking with respect to gastric cancer risk, a slightly increased risk appears for ever-smokers with a GSTT1 null genotype compared with individuals with a GSTT1 normal genotype. Even though the result is not statistically significant, from the z test a statistically significant difference among the estimates computed in ever- and never- smokers appears, so it would be interesting to pool all the missing data to confirm this result. If genetic susceptibility to gastric cancer is, in part, mediated through polymorphic variation, it is probable that the risk associated with any one locus will be small because an interaction is likely to operate in these circumstances. Hence, combinations of certain genotypes may be more discriminating as risk factors than a single locus genotype. Unfortunately, only a few reports investigated this aspect, even though 17 of 18 selected studies collected data on GSTM1 status. By pooling the data from seven available studies investigating a possible interaction between GSTT1 and GSTM1 status and gastric cancer risk, a 95% statistically significant increased risk of gastric cancer appeared for individuals with combined deletion mutations in GSTT1 and GSTM1 genes in comparison with individuals with both homozygous wild genotypes.

The main limitations of the study have to be considered in interpreting the results. First, only published studies were included in the meta-analysis; therefore a publication bias may have occurred. It is known that positive results usually have a greater probability of being published, and even though unpublished studies are generally of lesser quality than published ones (37), if they are not included an overestimation of the GSTT1 null effect may appear.

Secondly, the two subgroup meta-analyses considering interactions between GSTT1 null genotype and cigarette smoking, as well as between GSTT1 null and GSTM1 null genotypes, were performed on the basis of a fraction of all of the possible data to be pooled, so selection bias may have occurred and our results may be over inflated. In this context, it is well known that an important issue in performing a meta-analysis is that literature-based meta-analysis rather than individual data-based meta-analysis could be a potential source of bias. Meta-analyses may be inadequate to calculate a pooled estimate since published estimates are based on heterogeneous populations, different study designs and different statistical models. More reliable results can be expected if individual data are available for a pooled analysis, so that confounding factors can be considered, although this approach requires the authors of all of the published studies to share their data.

Thirdly, the quality score of the individual studies included in our meta-analysis was assessed on the basis of efforts to minimize the potential for selection bias, misclassification related to exposure, collection of data on potential confounders and method of statistical analysis. It is known that any quality assessment system has not yet been validated (38) and it is evident that our quality scale has a subjective component. Despite these limitations, assessment of the quality of individual studies used in our meta-analysis allowed us to draw two conclusions: the possibility of a selection bias and even more misclassification related to exposure cannot be ruled out; most importantly, pooling high-quality scored studies resulted in higher risk estimates than did pooling low-quality scored studies. If high-quality scored studies are more likely to yield valid information than low-quality studies, we can conclude that, on the basis of the currently available data, an additional slight risk of gastric cancer for GSTT1 null individuals may exist.

Despite these remarks, some interesting conclusions have emerged. From the results of this quantitative meta-analysis that combined the data from 7142 people (2508 cases and 4634 controls), it appears that GSTT1 null status has a very small effect on the risk of gastric cancer per se but it may modulate the tobacco-related carcinogenesis of gastric cancer, and that the combination of unfavourable genotypes may result in an additional risk of gastric cancer. A clearer picture of the interaction between different polymorphisms and environmental factors on gastric cancer risk will be adequately addressed only by large and well-designed epidemiological studies.

This work was supported by a grant from Catholic University of Sacred Heart (D1 projects 2005). We are indebted to Dr Hsieh Ling-Ling, Dr Ye Mei and Dr Torres María M. for their collaboration and their availability to share their data for the present meta-analysis. We are grateful to Aaron Isaacs for the linguistic revision of the final manuscript.

Conflict of Interest Statement: None declared.

References

1.
Hashibe,M., Brennan,P., Strange,R.C. et al. (
2003
) Meta- and pooled analyses of GSTM1, GSTT1, GSTP1, and CYP1A1 genotypes and risk of head and neck cancer.
Cancer Epidemiol. Biomarkers Prev.
 ,
12
,
1509
–1517.
2.
Egan,K.M., Cai,Q., Shu,X.O., Jin,F., Zhu,T.L., Dai,Q., Gao,Y.T. and Zheng,W. (
2004
) Genetic polymorphisms in GSTM1, GSTP1, and GSTT1 and the risk for breast cancer: results from the Shanghai Breast Cancer Study and meta-analysis.
Cancer Epidemiol. Biomarkers Prev.
 ,
13
,
197
–204.
3.
Saadat,I. and Saadat,M. (
2001
) Glutathione S-transferase M1 and T1 null genotypes and the risk of gastric and colorectal cancers.
Cancer Lett.
 ,
169
,
21
–26.
4.
Garte,S., Gaspari,L., Alexandrie,A.K. et al. (
2001
) Metabolic gene polymorphism frequencies in control populations.
Cancer Epidemiol. Biomarkers Prev.
 ,
10
,
1239
–1248.
5.
de Bruin,W.C., Wagenmans,M.J. and Peters,W.H. (
2000
) Expression of glutathione S-transferase alpha, P1-1 and T1-1 in the human gastrointestinal tract.
Jpn. J. Cancer Res.
 ,
91
,
310
–316.
6.
La Torre,G., Boccia,S. and Ricciardi,G. (
2005
) Glutathione S-Transferase M1 status and gastric cancer risk: a meta-analysis.
Cancer Lett.
 ,
217
,
53
–60.
7.
Gonzalez,C.A., Sala,N. and Capella,G. (
2002
) Genetic susceptibility and gastric cancer risk.
Int. J. Cancer
 ,
100
,
249
–260.
8.
Deakin,M., Elder,J., Hendrickse,C. et al. (
1996
) Glutathione S-transferase GSTT1 genotypes and susceptibility to cancer: studies of interactions with GSTM1 in lung, oral, gastric and colorectal cancers.
Carcinogenesis
 ,
17
,
881
–884.
9.
Katoh,T., Nagata,N., Kuroda,Y., Itoh,H., Kawahara,A., Kuroki,N., Ookuma,R. and Bell,D.A. (
1996
) Glutathione S-transferase M1 (GSTM1) and T1 (GSTT1) genetic polymorphism and susceptibility to gastric and colorectal adenocarcinoma.
Carcinogenesis
 ,
17
,
1855
–1859.
10.
Wang,K.Y., Lin,J.T., Chen,C.J. and Hsieh,L.L. (
1998
) The relationship between the genotypes of cancer susceptibility genes and the development of gastric cancer.
Chinese J. Pub. Health
 ,
17
,
226
–241
11.
Setiawan,V.W., Zhang,Z.F., Yu,G.P. et al. (
2000
) GSTT1 and GSTM1 null genotypes and the risk of gastric cancer: a case–control study in a Chinese population.
Cancer Epidemiol. Biomarkers Prev.
 ,
9
,
73
–80.
12.
Lan,Q., Chow,W.H., Lissowska,J., Hein,D.W., Buetow,K., Engel.L.S., Ji,B., Zatonski,W. and Rothman,N. (
2001
) Glutathione S-transferase genotypes and stomach cancer in a population-based case–control study in Warsaw, Poland.
Pharmacogenetics
 ,
11
,
655
–661.
13.
Cai,L., Yu,S.Z. and Zhang,Z.F. (
2001
) Glutathione S-transferases M1, T1 genotypes and the risk of gastric cancer: a case–control study.
World J. Gastroenterol.
 ,
7
,
506
–509.
14.
Wu,M.S., Chen,C.J., Lin,M.T., Wang,H.P., Shun,C.T., Sheu,J.C. and Lin,J.T. (
2002
) Genetic polymorphism of cytochrome P450 2E1, glutathione S-transferase M1 and T1, and susceptibility to gastric carcinoma in Taiwan.
Int. J. Colorectal Dis.
 ,
17
,
338
–343.
15.
Gao,C.M., Takezaki,T., Wu,J.Z. et al. (
2002
) Glutathione-S-transferase M1 (GSTM1) and GSTT1 genotype, smoking, consumption of alcohol and tea and risk of oesophageal and stomach cancers: a case–control study of a high-incidence area in Jiansu Province, China.
Cancer Lett.
 ,
188
,
95
–102.
16.
Sgambato,A., Campisi,B., Zupa,A., Bochicchio,A., Romano,G., Tartarone,A., Galasso,R., Traficante,A. and Cittadini,A. (
2002
) Glutathione S-transferase (GST) polymorphisms as risk factors for cancer in a highly homogeneous population from southern Italy.
Anticancer Res.
 ,
22
,
3647
–3652.
17.
Ye,M., Liu,J.Y. and Deng,C.S. (
2003
) Relationship between xenobiotic-metabolizing enzyme gene polymorphisms and genetic susceptibility of gastric cancer.
World Chinese J. Digest.
 ,
11
,
1314
–1317.
18.
Choi,S.C., Yun,K.J., Kim,T.H. et al. (
2003
) Prognostic potential of glutathione S-transferase M1 and T1 null genotypes for gastric cancer progression.
Cancer Lett.
 ,
195
,
169
–175.
19.
Colombo,J., Rossit,A.R., Caetano,A., Borim,A.A., Wornrath,D. and Silva,A.E. (
2004
) GSTT1, GSTM1 and CYP2E1 genetic polymorphisms in gastric cancer and chronic gastritis in a Brazilian population.
World J. Gastroenterol.
 ,
10
,
1240
–1245.
20.
Torres,M.M., Acosta,C.P., Sicard,D.M. and Groot de Restrepo,H. (
2004
) Genetic susceptibility and risk of gastric cancer in a human population of Cauca, Colombia.
Biomedica
 ,
24
,
153
–162.
21.
Tamer,L., Ates,N.A., Ates,C., Ercan,B., Elipek,T., Yildirim,H., Camdeviren,H., Atik,U. and Aydin,S. (
2005
) Glutathione S-transferase M1, T1 and P1 genetic polymorphisms, cigarette smoking and gastric cancer risk.
Cell Biochem. Funct.
 ,
23
,
267
–272.
22.
Palli,D., Saieva,C., Gemma,S. et al. (
2005
) GSTT1 and GSTM1 gene polymorphisms and gastric cancer in a high-risk Italian population.
Int. J. Cancer
 ,
115
,
284
–289.
23.
Mu,L.N., Lu,Q.Y., Yu,S.Z. et al. (
2005
) Green tea drinking and multigenetic index on the risk of stomach cancer in a Chinese population.
Int. J. Cancer
 ,
116
,
972
–983.
24.
Nan,H.M., Park,J.W., Song,Y.J., Yun,H.Y., Park,J.S., Hyun,T., Youn,S.J., Kim,Y.D., Kang,J.W. and Kim,H. (
2005
) Kimchi and soybean pastes are risk factors of gastric cancer.
World J. Gastroenterol.
 ,
11
,
3175
–3181.
25.
Angelillo,I.F. and Villari,P. (
1999
) Residential exposure to electromagnetic fields and childhood leukaemia: a meta-analysis.
Bull World Health Organ.
 ,
77
,
906
–915.
26.
Chalmers,T.C., Smith,H.Jr, Blackburn,B., Silverman,B., Schroeder,B., Reitman,D. and Ambroz.A. (
1981
) A method for assessing the quality of a randomized control trial.
Control Clin. Trials
 ,
2
,
31
–49.
27.
Thakkinstian,A., McElduff,P., D'Este,C., Duffy,D. and Attia,J. (
2005
) A method for meta-analysis of molecular association studies.
Stat. Med.
 ,
24
,
1291
–1306.
28.
Bogardus,S.T. Jr, Concato,J. and Feinstein,A.R. (
1999
) Clinical epidemiological quality in molecular genetic research: the need for methodological standards.
JAMA
 ,
281
,
1919
–1926.
29.
DerSimonian,R. and Laird,N. (
1986
) Meta-analysis in clinical trials.
Control Clin. Trials
 ,
7
,
177
–188.
30.
Mantel,N. and Haenszel,W. (
1959
) Statistical aspects of the analysis of data from retrospective studies of disease.
J. Natl Cancer Inst.
 ,
22
,
719
–748.
31.
Villari,P., Manzoli,L. and Boccia,A. (
2004
) Methodological quality of studies and patient age as major sources of variation in efficacy estimates of influenza vaccination in healthy adults: a meta-analysis.
Vaccine
 ,
22
,
3475
–3486.
32.
Review Manager (RevMan) [Computer program]. Version 4.2 for Windows (
2002
) The Cochrane Collaboration, Oxford.
33.
Schlesselmann,J.J. (
1982
) Case–control studies. Design, Conduct, Analysis. Oxford University Press, Oxford.
34.
Deeks,J.J., Altman,D.G. and Bradburn,M.J. (
2005
) Statistical methods for examining heterogeneity and combining results from several studies in meta-analysis. In Egger,M., Davey Smith,G., Altman,D.G. (eds), Systematic Reviews in Health Care. Meta-analysis in Context. BMJ Books, London, pp.
285
–312.
35.
Taningher,M., Malacarne,D., Izzotti,A., Ugolini,D. and Parodi,S. (
1999
) Drug metabolism polymorphisms as modulators of cancer susceptibility.
Mutat. Res.
 ,
436
,
227
–261.
36.
Blettner,M., Sauerbrei,W., Schlehofer,B., Scheuchenpflug,T. and Friedenreich,C. (
1999
) Traditional reviews, meta-analyses and pooled analyses in epidemiology.
Int. J. Epidemiol.
 ,
28
,
1
–9.
37.
Jekel,J.F. and Elmore,J.G. (
1996
) Epidemiology, Biostatistics and Preventive Medicine. W.B. Saunders Company, Philadelphia.
38.
Stroup,D.F., Berlin,J.A., Morton,S.C., Olkin,I., Williamson,G.D., Renniem,D., Moher,D., Becker,B.J., Sipe,T.A. and Thacker,S.B. (
2000
) Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group.
JAMA
 
19
,
2008
–2012.