Background: Studies investigating the association between genetic polymorphisms of glutathione S-transferases (GST) and risk of adult brain tumors have reported conflicting results. The rationale of this meta-analysis was to determine whether GST variants increase the susceptibility of adult brain tumors by pooling data.

Methods: Two investigators independently searched the HuGENet database, MEDLINE, EMBASE, conference articles, and manually reviewed bibliographies of retrieved articles. Papers were included if they were observational studies investigating the influence of GSTM1, GSTT1, GSTP1 I105V, or GSTP1 A114V on the development of adult brain cancers. Potential sources of heterogeneity between studies were explored in a meta-regression.

Results: We identified eight eligible studies, which included 1,630 cases of glioma, 245 cases of meningioma, and 7,151 controls. Using the random effects model, there was no association between any of the GST variants and the risk of glioma [overall odds ratio (OR), 1.08; 95% confidence interval (95% CI), 0.95-1.22]. Subgroup analyses also showed no relationship between GST variants and histopathologic groups; the overall ORs were 1.13 (95% CI, 0.88-1.43) for high-grade glioma and 1.08 (95% CI, 0.76-1.55) for low-grade glioma. A random effects meta-regression suggested that the use of in-hospital controls produced larger effect estimates in glioma than the use of population controls (overall OR, 1.30; 95% CI, 1.03-1.65). The T1 null genotype was significantly associated with a risk of meningioma (OR, 1.95; 95% CI, 1.02-3.76), but the M1 variant was not.

Conclusion: This study did not suggest any relationship between GST variants and risks of glioma; the T1 null genotype may influence the susceptibility of meningioma, but larger studies are needed to substantiate this relationship.

Except ionizing radiation, no environmental carcinogens have been firmly associated with the etiology of brain tumors. Studies in rats showed that brain tumors can be induced by various carcinogenic substances (1, 2), but observational studies in human showed no definitive association between occupational or environmental exposures and brain tumor incidence (3, 4). These negative findings may reflect the difficulty in characterizing exposure, particularly as a dose-response relationship. In addition, many individuals could be exposed to multiple rather than single toxic agents over time. Therefore, some investigators have turned to study genetic associations, such as genes involved in metabolizing chemicals [i.e., glutathione S-transferases (GST)], because genotype measurements are less prone to recall bias.

The GSTs are involved in phase II detoxification that protects cells from attack by reactive electrophiles (5). They catalyze the conjugation of glutathione to electrophilic species (such as chemical carcinogens and cytotoxic chemotherapeutic agents), which is the first step that leads to the elimination of toxic compounds. Although polymorphisms have been described in several of GST gene families, most attention has focused on allelism in GSTμ (GSTM1), GST𝛉 (GSTT1), and GSTπ (GSTP1; refs. 5, 6). GSTM1 and GSTT1 homozygotes (null genotype) have no enzymatic activities. GSTP1 has two polymorphisms: I105V and A114V, and evidence suggests that individuals with the I105V Val/Val allele may have lower affinity for electrophilic substrates and heat stability compared with the wild type (7, 8).

Because genetic variants of GSTs may reduce the cell's ability to metabolize toxins, their associations with cancers have been investigated extensively in epidemiologic studies (9-15). Likewise, there have been a number of reports on the relationship between GST variants and risk of brain cancers, but the results were conflicting. Brain tumors are uncommon cancers in adults, and recruiting sufficient subjects into case-control studies takes a lengthy period of time. As a first step to resolve these inconsistent findings, we did a meta-analysis. By pooling studies together, we also hope to increase the power of observing a small to moderate association.

Study Selection and Quality Assessment

We searched the HuGENet database (last search January 2005), MEDLINE, and EMBASE (January 1980 to January 2005) using these terms: glutathione S-transferase, glioma, brain tumors, gene, and allele. We also did manual searches of bibliographies of retrieved articles and conference abstracts in the past 3 years (American Association of Cancer Research and Society of Neuro-oncology). Two investigators (R.L. and L.C.) independently searched and reviewed abstracts in duplicate to determine if they met inclusion and exclusion criteria; any discrepancies were resolved through discussion. We considered all languages. When there were several publications for the same population, we used the most updated article. Articles were included if they fulfilled the following inclusion criteria: (a) observational studies that investigated the associations of GST variants and risk of adult brain tumors, focusing on polymorphisms in GSTM1, GSTT1, and GSTP1; (b) presentation of data necessary for the calculation of odds ratios (OR). The exclusion criteria were (a) studies that used GST polymorphisms to predict survival in brain tumors and (b) investigations of GST variants as markers for response to therapy.

Two investigators (R.L. and L.C.) independently assessed methodologic quality by using a set of published criteria for observational studies and abstracted data into standardized data collection forms (4). Any disagreements were resolved by consensus and reference to the articles. Papers were rated according to four areas: (a) quality of reporting, (b) confounding, (c) bias, and (d) external validity. We did not apply weights in the analysis based on rating scores but excluded studies in the sensitivity analysis based on methodologic weakness identified during the assessment.

For each included study, the following information was recorded: the year of publication, the country of origin, types of brain tumor, the number of cases and controls for each tumor type, matching variables, sources of the control population, the number of cases and controls with the variant allele and the wild type, histopathologic subgroups, techniques of genotyping, and the testing of gene-gene and gene-environment interactions.

Meta-analysis

We calculated the pooled ORs and the 95% confidence intervals (95% CI) separately for GSTM1, GSTT1, GSTP I105V, and GSTP A114V. We did not pool the adjusted ORs because studies either did not adjust for confounders, or the adjustments were not comparable among them. We did a test of homogeneity for each GST variant and set the critical value of P at 0.2 to avoid underestimating the presence of heterogeneity. Because there are greater potentials for bias and confounding in case control studies, we chose the random effects model (DerSimonian and Laird) to pool data (16).

Important sources of heterogeneity were further investigated in subgroups defined a priori. Some studies suggested that GST variants may preferentially influence the development of malignant glioma (17); therefore, we evaluated subgroups based on histopathology. We used three classifications: glioblastoma multiforme versus other histologies, high-grade glioma (WHO grade 3 and 4) versus low-grade glioma (WHO grade 1 and 2), and astrocytic (anaplastic astrocytoma or low-grade astrocytoma) versus oligodendroglial tumors (anaplastic oligodendroglioma, low-grade oligodendroglioma or mixed oligoastrocytoma). Other sources of heterogeneity were the type of control population and the study size. Hospital controls, in contrast to population controls, may give different estimates of the genotype-disease association, because the prevalence of their alleles may differ from that of the general population (18). In addition, smaller case-control studies tend to produce larger effect size (19). Therefore, we did a random effect, multivariable meta-regression using the control source and the study size as predictors of heterogeneity. In this analysis, hospital controls were either patients or healthy blood donors/visitors recruited within a hospital setting, whereas population controls were selected through population-based sampling methods (0 = population, 1 = hospital). We coded studies with fewer than 100 cases as small and >100 as large (0 = small, 1 = large).

Publication bias was evaluated by the Egger's and Begg's funnel plot asymmetry tests (20, 21). All statistical analyses were done using Stata statistical software, version 8.2.

Study Characteristics

A total of 95 abstracts were retrieved through MEDLINE, EMBASE, the HuGENet database, conference abstracts, and bibliographies of retrieved articles. Eighty-four of them were excluded based on inclusion and exclusion criteria. The articles of 11 of them were further reviewed (22-32). Three of the 11 were based on the same patient population (25, 29, 30), and the most updated series was used. We further excluded one article because it was a duplicate of another (26, 27). The final number of included publications was eight, with 1,630 cases of glioma, 245 cases of meningioma, and 7,151 controls. One of the eight was an abstract judged to have sufficient details for data pooling (32). Study characteristics were summarized in Table 1. All studies were published after 1990. Five were conducted in the United States; seven were published in English and one in Russian (translated into English for assessment). All articles investigated the association between GST polymorphisms and the risk of glioma; three studied this relationship in meningioma. It was not possible to do meta-analyses for acoustic neuroma and neuroepithelial tumors because there was only one study for each of them. Five studies used frequency matching, whereas the other three did not match their cases and controls. Age and sex were the common matching factors. Five studies recruited hospital controls, with three of the five from in-hospital patients. All eight studies genotyped the variant allele GSTM1; seven examined GSTT1 and the five most recent series also investigated GSTP1. All studies used PCR-based method for genotyping. Five of the eight tested for gene-gene interactions, whereas only two examined gene-environment interactions.

Table 1.

Characteristics of studies on GST and risk of brain tumors

Study (first author)Year of publicationCountryTypes of brain tumorCases (n)*Controls (n)*MatchingMatching variablesSource of controlGST M1GST T1GSTP I105VGSTP A104VGlioma histopathologic subgroups studiedGene- gene interactionGene- environment interaction
Pinarbasi 2005 Turkey Glioma and meningioma Glioma, 31; meningioma, 23 153 Frequency Age, gender Hospital  None None M1, T1 and P1 I105V with smoking 
Wrensch 2004 U.S.A. Glioma 458 428 Frequency Age, race, gender Population (a) GBM versus others; (b) GBM, A and AA, ODG and OA, others P53 + M1, T1, P1 None 
Butler 2003 U.S.A. Glioma 325 579 Frequency Age, gender Population  GBM, A, ODG None GST variants with living on farm status, smoking and pesticides used 
De Roos 2003 U.S.A. Glioma, meningioma, and acoustic neuroma Glioma, 422; meningioma, 172; acoustic neuroma, 79 604 Frequency Hospital, age, gender, race, distance from home to hospital Hospital GBM, AA Other A, ODG, Mixed OA I105V + A114V; I105 V + CYP2E1; I105V + T1; T1 + CYP2E1 None 
Ezer 2002 U.S.A. Glioma and neuroepithelial tumors Glioma, 141; neuroepithelial tumors, 76 653-1,709§ None None Population AA, ODG, OA None None 
Kondrateva 1999 Russia Glioma 54 103 None None Hospital-healthy    High versus low grade glioma M1 + L-MYC None 
Trizna 1998 U.S.A. Glioma 90 90 Frequency Age, race, gender Hospital-healthy   None T1 + NAT2 None 
Elexpuru-Camiruaga 1995 Great Britain Glioma and meningioma Glioma, 109; meningioma, 50 577 None None Hospital   High versus low grade glioma T1 + CYP2D6 None 
Study (first author)Year of publicationCountryTypes of brain tumorCases (n)*Controls (n)*MatchingMatching variablesSource of controlGST M1GST T1GSTP I105VGSTP A104VGlioma histopathologic subgroups studiedGene- gene interactionGene- environment interaction
Pinarbasi 2005 Turkey Glioma and meningioma Glioma, 31; meningioma, 23 153 Frequency Age, gender Hospital  None None M1, T1 and P1 I105V with smoking 
Wrensch 2004 U.S.A. Glioma 458 428 Frequency Age, race, gender Population (a) GBM versus others; (b) GBM, A and AA, ODG and OA, others P53 + M1, T1, P1 None 
Butler 2003 U.S.A. Glioma 325 579 Frequency Age, gender Population  GBM, A, ODG None GST variants with living on farm status, smoking and pesticides used 
De Roos 2003 U.S.A. Glioma, meningioma, and acoustic neuroma Glioma, 422; meningioma, 172; acoustic neuroma, 79 604 Frequency Hospital, age, gender, race, distance from home to hospital Hospital GBM, AA Other A, ODG, Mixed OA I105V + A114V; I105 V + CYP2E1; I105V + T1; T1 + CYP2E1 None 
Ezer 2002 U.S.A. Glioma and neuroepithelial tumors Glioma, 141; neuroepithelial tumors, 76 653-1,709§ None None Population AA, ODG, OA None None 
Kondrateva 1999 Russia Glioma 54 103 None None Hospital-healthy    High versus low grade glioma M1 + L-MYC None 
Trizna 1998 U.S.A. Glioma 90 90 Frequency Age, race, gender Hospital-healthy   None T1 + NAT2 None 
Elexpuru-Camiruaga 1995 Great Britain Glioma and meningioma Glioma, 109; meningioma, 50 577 None None Hospital   High versus low grade glioma T1 + CYP2D6 None 

Abbreviations: GBM, glioblastoma mutiforme; A, astrocytoma; AA, anaplastic astrocytoma; ODG, oligodendroglioma; OA, oligoastrocytoma.

*

The number of subjects with genotypes.

Significant interaction of GSTT1 and P53 in patients with glioblastoma multiforme.

Significant interactions between GSTP1 I105V and A114V in glioma, I105V and CYP2E12 in glioma and acoustic neuroma.

§

The authors used healthy controls published in the literature; GSTM1: 1473, GSTT1: 782, GSTP1 I105V: 1709, A114V:653.

Significant interaction.

Table 2 summarized the genotype frequencies of brain tumor cases and controls. The distribution of GST variants in the Turkish study was different from the others (31); this is most noted for the frequency of GSTM1 null genotype in the control group, which differed significantly from that of the healthy Turkish population (33). Because their results represented outliers, we excluded this article in a sensitivity analysis (Table 3). The conclusion, nevertheless, is robust for glioma even when this article was not included.

Table 2.

Distribution of GSTM1, GSTT1 GSTP1 I105V, and A114V genotypes among brain tumors cases and controls

First authorBrain tumor type% GSTM1 null
% GSTT1 null
% I105V Val/Val
% 114 Ala/Val or Val/Val
CasesControlsCasesControlsCasesControlsCasesControls
Pinarbasi Glioma 48.4 24.2 32.3 20.3 3.2 13.7 — — 
Wrensch Glioma 52.0 51.9 20.1 21.6 10.9 12.7 15.7 14.5 
Butler Glioma 50.6 50.7 17.2 14.6 11.5 12.9 — — 
De Roos Glioma 52.6 55.8 19.9 18.4 16.9 10.4 14.6 13.5 
Ezer Glioma 51.8 49.0 14.9 18.1 4.3 10.1 9.9 14.0 
Kondrateva Glioma 52.0 52.0 — — — — — — 
Trizna Glioma 52.2 43.3 27.8 30.0 — — — — 
Elexpuru- Camiruaga Glioma 59.6 54.6 32.1 18.4 — — — — 
Pinarbasi Meningioma 47.8 24.2 26.1 20.3 8.7 13.7 — — 
De Roos Meningioma 49.7 55.8 23.9 18.4 7.1 10.4 12.4 13.5 
Elexpuru-Camiruaga Meningioma 55.1 54.6 44.7 18.4 — — — — 
De Roos Acoustic neuroma 53.4 55.8 15.7 18.4 12.5 10.4 13.7 13.5 
Ezer Neuroepithelial tumor 57.9 49.0 10.7 18.1 5.3 10.1 19.7 14.0 
First authorBrain tumor type% GSTM1 null
% GSTT1 null
% I105V Val/Val
% 114 Ala/Val or Val/Val
CasesControlsCasesControlsCasesControlsCasesControls
Pinarbasi Glioma 48.4 24.2 32.3 20.3 3.2 13.7 — — 
Wrensch Glioma 52.0 51.9 20.1 21.6 10.9 12.7 15.7 14.5 
Butler Glioma 50.6 50.7 17.2 14.6 11.5 12.9 — — 
De Roos Glioma 52.6 55.8 19.9 18.4 16.9 10.4 14.6 13.5 
Ezer Glioma 51.8 49.0 14.9 18.1 4.3 10.1 9.9 14.0 
Kondrateva Glioma 52.0 52.0 — — — — — — 
Trizna Glioma 52.2 43.3 27.8 30.0 — — — — 
Elexpuru- Camiruaga Glioma 59.6 54.6 32.1 18.4 — — — — 
Pinarbasi Meningioma 47.8 24.2 26.1 20.3 8.7 13.7 — — 
De Roos Meningioma 49.7 55.8 23.9 18.4 7.1 10.4 12.4 13.5 
Elexpuru-Camiruaga Meningioma 55.1 54.6 44.7 18.4 — — — — 
De Roos Acoustic neuroma 53.4 55.8 15.7 18.4 12.5 10.4 13.7 13.5 
Ezer Neuroepithelial tumor 57.9 49.0 10.7 18.1 5.3 10.1 19.7 14.0 
Table 3.

Sensitivity analyses of glioma and GST variants

GST genotype variantsExclusion of one study that did not recruit its own control group
Exclusion of one study with results that were outliers
No. studiesn (Ca/Co)Pooled OR (95% CI)No. studiesn (Ca/Co)Pooled OR (95% CI)
GSTM1 7* 1,479/2,609 1.08 (0.90-1.31) 7* 1,589/3,929 0.98 (0.87-1.11) 
GSTT1 6 1,424/2,424 1.22 (0.94-1.58) 6 1,534/3,053 0.90 (0.70-1.15) 
GSTP1 I105V 4 1,227/1,832 1.00 (0.60-1.65) 4 1,337/3,388 1.10 (0.66-1.82) 
GSTP1 A114V 2§ 874/1,095 1.10 (0.85-1.42) 3 1,015/1,748 1.02 (0.80-1.30) 
GST genotype variantsExclusion of one study that did not recruit its own control group
Exclusion of one study with results that were outliers
No. studiesn (Ca/Co)Pooled OR (95% CI)No. studiesn (Ca/Co)Pooled OR (95% CI)
GSTM1 7* 1,479/2,609 1.08 (0.90-1.31) 7* 1,589/3,929 0.98 (0.87-1.11) 
GSTT1 6 1,424/2,424 1.22 (0.94-1.58) 6 1,534/3,053 0.90 (0.70-1.15) 
GSTP1 I105V 4 1,227/1,832 1.00 (0.60-1.65) 4 1,337/3,388 1.10 (0.66-1.82) 
GSTP1 A114V 2§ 874/1,095 1.10 (0.85-1.42) 3 1,015/1,748 1.02 (0.80-1.30) 

Abbreviations: Ca/Co, cases/controls; n, number of subjects with particular genotypes.

*

Included (22, 23, 26, 28, 30-32).

Included (22, 23, 28, 30-32).

Included (22, 30-32).

§

Included (22, 30).

Included (22, 24, 30).

Quality Assessment of Studies

The inter-rater agreement for quality assessment was very good (Cohen's κ = 0.76). We identified several methodologic weaknesses during the assessment. In four of the eight studies, there was no demographic comparison to ascertain whether cases and controls were comparable (23, 24, 26, 28); in those that presented these data (22, 3032), there were baseline differences between cases and controls. For example, genotyped cases were on average 5 to 6 years younger than genotyped controls in two studies (30, 31), and cases had 20% fewer men but 20% more women in another study (31).

One group of investigators did not recruit their own controls but used healthy population published in the literature as the control group (24); moreover, there was no information on their comparability. Consequently, their risk estimates might have been biased, because it is likely that these cases and borrowed “controls” came from different study base, geographically or temporally. After we excluded this article in the sensitivity analysis, however, the results were unchanged (Table 3).

Three of the eight studies did not adjust for potential confounders (24, 26, 31); three adjusted for all genotypes simultaneously (22, 23, 30), but there was little confounding between them. Three studies reported quality control measures for genotyping with replicates (22, 28, 30), but only one stated the reliability of their assays (22). No study mentioned blinding of the genotyping personnel. Only one investigation assessed the prevalence of GSTP1 genotypes for departure from the Hardy-Weinberg equilibrium and showed no deviation (22). Similar calculation was impossible for GSTM1 and T1 genotypes because they were coded as wild type or null. No study stated whether subgroup analyses were planned a priori or on a post hoc exploratory basis. On the positive side, all had histologic confirmation of their cases.

Meta-analysis of GST Variants and Glioma

The results of this meta-analysis in glioma were presented in Fig. 1A-D. We found significant tests of homogeneity for GSTM1 (χ2 = 10.16, P = 0.18), GSTT1 (χ2 = 12.53, P = 0.05), and GSTP1 I105V (χ2 = 16.69, P = 0.002). Using the random effects model, none of the four GST variants showed a significant association with glioma. For histopathologic subgroup evaluations, none of the variant alleles was associated with glioblastoma multiforme versus other histologies, high-grade glioma versus low-grade glioma, and astrocytic versus oligodendroglial tumors. The results were shown in Table 4A and B.

Figure 1.

A. Meta-analysis of GSTM1 and risk of glioma. B. Meta-analysis of GSTT1 and risk of glioma. C. Meta-analysis of GSTP1 I105V and risk of glioma. D. Meta-analysis of GSTP1 A114V and risk of glioma.

Figure 1.

A. Meta-analysis of GSTM1 and risk of glioma. B. Meta-analysis of GSTT1 and risk of glioma. C. Meta-analysis of GSTP1 I105V and risk of glioma. D. Meta-analysis of GSTP1 A114V and risk of glioma.

Close modal
Table 4.

A: Subgroup analyses: GBM versus other histologies and oligodendroglial versus astrocytic tumors
Genotype variantsn* (cases/controls)Pooled OR (95% CI)n (cases/controls)Pooled OR (95% CI)
GSTM1 GBM, 410/1,107; others, 460/1,107 GBM, 1.05 (0.83-1.32); others, 0.86 (0.58-1.28) Astro, 237/2,580; Oligo, 209/2,580 Astro, 0.98 (0.75-1.28); Oligo, 0.93 (0.52-1.68) 
GSTT1 GBM, 404/1,108; others, 465/1,108 GBM, 0.96 (0.71-1.29); others, 1.05 (0.80-1.38) Astro, 231/1,890; Oligo, 206/1,890 Astro, 0.84 (0.51-1.40); Oligo, 1.02 (0.63-1.65) 
GSTP1 I105V GBM, 445/1,100; others, 426/1,100 GBM, 1.15 (0.57-2.34); others, 1.28 (0.62-2.68) Astro, 236/2,809; Oligo, 208/2,809 Astro, 0.86 (0.34-2.20); Oligo, 0.89 (0.49-1.63) 
GSTP1 A114V GBM, 415/1,095; others, 459/1,095 GBM, 1.10 (0.80-1.52); others, 1.10 (0.80-1.50) Astro, 237/1,993; Oligo, 184/1,993 Astro, 1.08 (0.64-1.82); Oligo, 1.29 (0.70-2.39) 

 
    
B: Subgroup analyses: high-grade versus low-grade glioma
 
   
Genotype variants No. studies n (cases/controls) Pooled OR (95% CI) 
GSTM1 4 HGG, 470/2,757; LGG, 176/2,757 HGG, 1.07 (0.87-1.31); LGG, 1.17 (0.58-2.39) 
GSTT1 3§ HGG, 429/1,880; LGG, 149/1,880 HGG, 1.34 (0.78-2.33); LGG, 1.07 (0.61-1.87) 
GSTP1 I105V 2 HGG, 343/2,313; LGG, 136/2,313 HGG, 1.07 (0.25-4.67); LGG, 0.60 (0.14-2.61) 
GSTP1 A114V 2 HGG, 346/1,257; LGG, 112/1,257 HGG, 0.91 (0.51-1.62); LGG, 1.39 (0.57-3.39) 
A: Subgroup analyses: GBM versus other histologies and oligodendroglial versus astrocytic tumors
Genotype variantsn* (cases/controls)Pooled OR (95% CI)n (cases/controls)Pooled OR (95% CI)
GSTM1 GBM, 410/1,107; others, 460/1,107 GBM, 1.05 (0.83-1.32); others, 0.86 (0.58-1.28) Astro, 237/2,580; Oligo, 209/2,580 Astro, 0.98 (0.75-1.28); Oligo, 0.93 (0.52-1.68) 
GSTT1 GBM, 404/1,108; others, 465/1,108 GBM, 0.96 (0.71-1.29); others, 1.05 (0.80-1.38) Astro, 231/1,890; Oligo, 206/1,890 Astro, 0.84 (0.51-1.40); Oligo, 1.02 (0.63-1.65) 
GSTP1 I105V GBM, 445/1,100; others, 426/1,100 GBM, 1.15 (0.57-2.34); others, 1.28 (0.62-2.68) Astro, 236/2,809; Oligo, 208/2,809 Astro, 0.86 (0.34-2.20); Oligo, 0.89 (0.49-1.63) 
GSTP1 A114V GBM, 415/1,095; others, 459/1,095 GBM, 1.10 (0.80-1.52); others, 1.10 (0.80-1.50) Astro, 237/1,993; Oligo, 184/1,993 Astro, 1.08 (0.64-1.82); Oligo, 1.29 (0.70-2.39) 

 
    
B: Subgroup analyses: high-grade versus low-grade glioma
 
   
Genotype variants No. studies n (cases/controls) Pooled OR (95% CI) 
GSTM1 4 HGG, 470/2,757; LGG, 176/2,757 HGG, 1.07 (0.87-1.31); LGG, 1.17 (0.58-2.39) 
GSTT1 3§ HGG, 429/1,880; LGG, 149/1,880 HGG, 1.34 (0.78-2.33); LGG, 1.07 (0.61-1.87) 
GSTP1 I105V 2 HGG, 343/2,313; LGG, 136/2,313 HGG, 1.07 (0.25-4.67); LGG, 0.60 (0.14-2.61) 
GSTP1 A114V 2 HGG, 346/1,257; LGG, 112/1,257 HGG, 0.91 (0.51-1.62); LGG, 1.39 (0.57-3.39) 

Abbreviations: n, number of subjects with the particular genotype; GBM, glioblastoma multiforme; Astro, astrocytic tumors; Oligo, oligodendroglial tumors; HGG, high-grade gliomas; LGG, low-grade glioma.

*

The number of cases and controls was based on two studies (22, 30).

The number of cases and controls was based on three studies (22, 24, 30).

Included (22-24, 26).

§

Included (22-24).

Included (22, 24).

Included (22, 24).

In the random effects meta-regression, the use of hospital controls produced a significantly stronger association than the use of population controls in the P1 I105V variant (Table 5). For the M1 variant, smaller studies found significantly larger effect estimates; in contrast, larger studies seemed to detect a bigger effect in the P1 I105V variant, but the CI was wide because there were few data points in that category. Overall, the control source but not the study size was a predictor of the between study heterogeneity when all genotypes were combined. The between study variance τ2 was reduced from 0.039 to 0.020 after accounting for both predictors. Two of the five studies' hospital controls were healthy subjects (26, 28). When we compared only diseased hospital controls with population controls in our regression model, the control source was still a significant predictor (overall OR, 1.30; 95% CI, 1.03-1.64) but study size was not (overall OR, 0.63; 95% CI, 0.34-1.17). Likewise, we reached the same conclusion when the two studies with healthy hospital controls were analyzed as population controls (data not shown).

Table 5.

Results of the meta-regression

PredictorsTypes of brain tumorsGSTM1, OR (95% CI)GSTT1, OR (95% CI)GSTP1 I105V, OR (95% CI)GSTP1 A114V, OR (95% CI)All genotypes combined, OR (95% CI)
Control source Glioma 0.95 (0.72-1.24) 1.51 (0.58-2.63) 2.22 (1.40-3.53) 1.18 (0.61-2.26) 1.30 (1.03-1.65) 
Study size Glioma 0.64 (0.41-0.99) 1.22 (0.58-2.58) 8.32 (1.04-16.52) —* 0.93 (0.64-1.36) 
 Meningioma 0.52 (0.23-1.35) 0.57 (0.17-1.97) — — 0.56 (0.23-1.35) 
PredictorsTypes of brain tumorsGSTM1, OR (95% CI)GSTT1, OR (95% CI)GSTP1 I105V, OR (95% CI)GSTP1 A114V, OR (95% CI)All genotypes combined, OR (95% CI)
Control source Glioma 0.95 (0.72-1.24) 1.51 (0.58-2.63) 2.22 (1.40-3.53) 1.18 (0.61-2.26) 1.30 (1.03-1.65) 
Study size Glioma 0.64 (0.41-0.99) 1.22 (0.58-2.58) 8.32 (1.04-16.52) —* 0.93 (0.64-1.36) 
 Meningioma 0.52 (0.23-1.35) 0.57 (0.17-1.97) — — 0.56 (0.23-1.35) 
*

All studies have >100 cases.

Use of either the Begg's (rank correlation test) or Egger's funnel plot asymmetry test (regression method) did not reveal any significant publication bias. Begg's test score = −1 (SD = 5.32, P = 0.85); the Egger's test β coefficient = 0.28 (95% CI, −0.23 to 0.80; P = 0.20).

Meta-analysis of GST Variants and Meningioma

Three studies investigated the association among GSTM1, GSTT1, and the risk of meningioma (Table 6). The test of homogeneity was significant for both the M1 and T1 variants (χ2 = 6.31, P = 0.043 and χ2 = 6.15, P = 0.046, respectively). Using the random effects model, the result for GSTM1 was not significant, but there was a significant increase in risk associated with GSTT1. When we excluded the Turkish study with outlier data in a sensitivity analysis, there was only a trend of association between GSTT1 and meningioma (OR, 2.20; 95% CI, 0.89-5.39; P = 0.08). In the meta-regression analyses, study size was not associated with estimates of effect in either variant allele. We were unable to investigate the control source as a predictor because all three studies used hospital controls.

Table 6.

Meta-analysis of GST variants and meningioma

GST genotype variantsNo. studiesn (cases/controls)Pooled OR (95% CI)P
GSTM1 3* 244/1,334 1.20 (0.66-2.16) 0.56 
GSTT1 3 242/1,251 1.95 (1.02-3.76) 0.046 
GST genotype variantsNo. studiesn (cases/controls)Pooled OR (95% CI)P
GSTM1 3* 244/1,334 1.20 (0.66-2.16) 0.56 
GSTT1 3 242/1,251 1.95 (1.02-3.76) 0.046 
*

Included (22, 23, 31).

Included (22, 23, 31).

This meta-analysis did not find any association among GSTM1, T1, P1 (I105V and A114V), and a risk of glioma; subgroup analyses also did not reveal any influence of GST polymorphisms on histopathologic subtypes of glioma. Perhaps these negative results support recent evidence that GST genotypes may not be accurate predictors of tissue-specific GST expression (34). Although GSTM1, T1, and P1 are all expressed in the brain, P1 is the most predominant enzyme and was hypothesized to play a critical role in protecting the brain from toxic compounds (35, 36). But if the P1 genotype obtained from the peripheral blood does not correspond to P1 expression in the brain, we may observe no consistent relationship between this variant and brain tumors.

Although this is the first full-length meta-analysis of GST variants and brain tumors, there has been an abstract presented recently on the same topic (37). In that meta-analysis, the authors suggested that GSTT1 null and GSTP1 105Val/Val were risk factors for glioblastoma multiforme (OR, 1.41; 95% CI, 1.06-1.87 and OR, 1.77; 95% CI, 1.21-2.59, respectively), and GSTT1 null was a risk factor for meningioma (OR, 1.98; 95% CI, 1.35-2.90). There was not enough detail in their presentation to ascertain why some of their results were different from ours. However, the authors included two earlier series along with the updated study (25, 29, 30). Because the participants were overlapping, inclusion of all three articles might have resulted in overrepresentation of the same data.

Similar to that abstract, we found a significant relationship between GSTT1 null genotype and a risk of meningioma. The pooled estimate, however, was based on only three hospital-based studies, and the association was no longer significant after a sensitivity analysis. Therefore, we still need further investigations, especially larger population-based studies, to substantiate our findings.

Our results showed that there were some demographic differences between cases and controls. Genotyped cases (ascertained by a cancer registry) were on average 5 to 6 years younger than genotyped controls in one study due to delay in blood sample collection, sometimes up to 6 months after diagnosis, and specimens could not be obtained from cases with the poorest survival (30). This problem raises the possibility of case selection bias, as there is evidence to suggest that GSTM1 null genotype is associated with time to specimen collection (from diagnosis) and longer case survival (30, 38). Another group used hospital cases shortly after diagnoses were made, thus minimizing the problem of selection bias from case survival (22). However, brain tumor patients were still on average older and more highly educated than controls. Studies that did not attempt matching may suffer potential biases induced by population stratification (24, 26, 28), as cases and controls could have different allele frequencies attributable to diversity in ethnic background but unrelated to disease status (39).

Our study has limitations. Meta-analysis of case-control studies is vulnerable to biases and confounding issues inherent in the original articles; therefore, study quality assessment and evaluation of heterogeneity are crucial. Results of the meta-regression suggested that the use of hospital controls produced stronger genotype-disease associations than the use of population controls. Perhaps, variant alleles were represented less frequently, or the wild-type alleles were found more often among in-hospital patients, and consequently, the estimates were biased away from the null. For example, GSTT1 wild types were often found in smokers with coronary artery disease and in patients with acute pancreatitis; the I105V Val/Val allele is uncommon in asthmatics (40-42).

There are other potential sources of heterogeneity, but because some factors were evaluated in only one study, we were unable to explore them further in subgroup analyses or meta-regression. For example, age is a modifying factor of genotype expression (43), but only one study reported genotype-brain tumor associations stratified by age groups (22). Furthermore, GST variants show substantial variations in prevalence based on ethnic groups (44), but there were so few non-Caucasian patients in brain tumor cases that no study explored ethnicity as subgroups or did stratified analyses.

A meta-analysis of gene-gene interactions was not possible here, because no study reported the same interaction pairs, although five of the eight tested them. Some interactions were statistically significant (Table 2 footnote), but these results could be due to chance because the comparison groups involved very few subjects.

Only two epidemiologic studies in brain tumors investigated gene and environment interactions and found nonsignificant results (31, 32). Another study is under way in the United States (45). Given the results of this meta-analysis, GST variants by themselves are unlikely to be strong determinants of the susceptibility of glioma; however, whether they may act in synergy with other genes or environmental factors is the question for future studies.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dagmar Smatanova for translating the Russian article into English.

1
Rice JM, Wilbourn JD. Tumors of the nervous system in carcinogenic hazard identification.
Toxicol Pathol
2000
;
28
:
202
–14.
2
Whysner J, Ross PM, Conaway CC, Verna LK, Williams GM. Evaluation of possible genotoxic mechanisms for acrylonitrile tumorigenicity.
Regul Toxicol Pharmacol
1998
;
27
:
217
–39.
3
Wrensch M, Minn Y, Chew T, Bondy M, Berger MS. Epidemiology of primary brain tumors: current concepts and review of the literature.
Neurooncol
2002
;
4
:
278
–99.
4
Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions.
J Epidemiol Community Health
1998
;
52
:
377
–84.
5
Strange RC, Spiteri MA, Ramachandran S, Fryer AA. Glutathione-S-transferase family of enzymes.
Mutat Res
2001
;
482
:
21
–6.
6
Landi S. Mammalian class theta GST and differential susceptibility to carcinogens: a review.
Mutat Res
2000
;
463
:
247
–83.
7
Ali-Osman F, Akande O, Antoun G, Mao JX, Buolamwini J. Molecular cloning, characterization, and expression in Escherichia coli of full-length cDNAs of three human glutathione S-transferase Pi gene variants. Evidence for differential catalytic activity of the encoded proteins.
J Biol Chem
1997
;
272
:
10004
–12.
8
Zimniak P, Nanduri B, Pikula S, et al. Naturally occurring human glutathione S-transferase GSTP1–1 isoforms with isoleucine and valine in position 104 differ in enzymic properties.
Eur J Biochem
1994
;
224
:
893
–9.
9
Engel LS, Taioli E, Pfeiffer R, et al. Pooled analysis and meta-analysis of glutathione S-transferase M1 and bladder cancer: a HuGE review.
Am J Epidemiol
2002
;
156
:
95
–109.
10
Geisler SA, Olshan AF. GSTM1, GSTT1, and the risk of squamous cell carcinoma of the head and neck: a mini-HuGE review.
Am J Epidemiol
2001
;
154
:
95
–105.
11
Hashibe M, Brennan P, Strange RC, et al. Meta- and pooled analyses of GSTM1, GSTT1, GSTP1, and CYP1A1 genotypes and risk of head and neck cancer.
Cancer Epidemiol Biomarkers Prev
2003
;
12
:
1509
–17.
12
Houlston RS. Glutathione S-transferase M1 status and lung cancer risk: a meta-analysis.
Cancer Epidemiol Biomarkers Prev
1999
;
8
:
675
–82.
13
Johns LE, Houlston RS. Glutathione S-transferase mu1 GSTM1 status and bladder cancer risk: a meta-analysis.
Mutagenesis
2000
;
15
:
399
–404.
14
McWilliams JE, Sanderson BJ, Harris EL, Richert-Boe KE, Henner WD. Glutathione S-transferase M1 GSTM1 deficiency and lung cancer risk.
Cancer Epidemiol Biomarkers Prev
1995
;
4
:
589
–94.
15
Habdous M, Siest G, Herbeth B, Vincent-Viry M, Visvikis S. Glutathione S-transferases genetic polymorphisms and human diseases: overview of epidemiological studies.
Ann Biol Clin (Paris)
2004
;
62
:
15
–24.
16
DerSimonian R, Laird N. Meta-analysis in clinical trials.
Control Clin Trials
1986
;
7
:
177
–88.
17
Grant R, Ironside JW. Glutathione S-transferases and cytochrome P450 detoxifying enzyme distribution in human cerebral glioma.
J Neurooncol
1995
;
25
:
1
–7.
18
Garcia-Closas M, Wacholder S, Caporaso N, Rothman N. Inference issues in cohort and case-control studies of genetic effects and gene-environment interactions. In: Khoury MJ, Little J, Burke W, editors. Human genome epidemiology. 1st ed. New York: Oxford University Press; 2004. p. 127–44.
19
Renehan AG, Zwahlen M, Minder C, et al. Insulin-like growth factor IGF-I, IGF binding protein-3, and cancer risk: systematic review and meta-regression analysis.
Lancet
2004
;
363
:
1346
–53.
20
Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias.
Biometrics
1994
;
50
:
1088
–101.
21
Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test.
BMJ
1997
;
315
:
629
–34.
22
De Roos AJ, Rothman N, Inskip PD, et al. Genetic polymorphisms in GSTM1, -P1, -T1, and CYP2E1 and the risk of adult brain tumors.
Cancer Epidemiol Biomarkers Prev
2003
;
12
:
14
–22.
23
Elexpuru-Camiruaga J, Buxton N, Kandula V, et al. Susceptibility to astrocytoma and meningioma: influence of allelism at glutathione S-transferase GSTT1 and GSTM1 and cytochrome P-450 CYP2D6 loci.
Cancer Res
1995
;
55
:
4237
–9.
24
Ezer R, Alonso M, Pereira E, et al. Identification of glutathione S-transferase GST polymorphisms in brain tumors and association with susceptibility to pediatric astrocytomas.
J Neurooncol
2002
;
59
:
123
–34.
25
Kelsey KT, Wrensch M, Zuo ZF, Miike R, Wiencke JK. A population-based case-control study of the CYP2D6 and GSTT1 polymorphisms and malignant brain tumors.
Pharmacogenetics
1997
;
7
:
463
–8.
26
Kondrat'eva TV, Imianitov EN, Togo AV, et al. L-myc and GSTM1 polymorphism in cerebral glioma.
Vopr Onkol
1999
;
45
:
523
–7.
27
Kondratieva TV, Imyanitov EN, Togo AV, et al. L-MYC and GSTM1 polymorphisms are associated with unfavourable clinical parameters of gliomas.
J Exp Clin Cancer Res
2000
;
19
:
197
–200.
28
Trizna Z, de AM, Kyritsis AP, et al. Genetic polymorphisms in glutathione S-transferase mu and theta, N-acetyltransferase, and CYP1A1 and risk of gliomas.
Cancer Epidemiol Biomarkers Prev
1998
;
7
:
553
–5.
29
Wiencke JK, Wrensch MR, Miike R, Zuo Z, Kelsey KT. Population-based study of glutathione S-transferase mu gene deletion in adult glioma cases and controls.
Carcinogenesis
1997
;
18
:
1431
–3.
30
Wrensch M, Kelsey KT, Liu M, et al. Glutathione-S-transferase variants and adult glioma.
Cancer Epidemiol Biomarkers Prev
2004
;
13
:
461
–7.
31
Pinarbasi H, Silig Y, Gurelik M. Genetic polymorphisms of GSTs and their association with primary brain tumor incidence.
Cancer Genet Cytogenet
2005
;
156
:
144
–9.
32
Butler MA, Ruder AM, Daly AK, et al. Polymorphisms in GSTM1, GSTT1, GSTP1 and NAT2 and susceptibility to primary intracranial brain gliomas.
Proc Am Assoc Cancer Res
2003
;
44
:
128
.
33
Ada AO, Suzen SH, Iscan M. Polymorphisms of cytochrome P450 1A1, glutathione S-transferases M1 and T1 in a Turkish population.
Toxicol Lett
2004
;
151
:
311
–5.
34
Coles BF, Kadlubar FF. Detoxification of electrophilic compounds by glutathione S-transferase catalysis: determinants of individual response to chemical carcinogens and chemotherapeutic drugs?
Biofactors
2003
;
17
:
115
–30.
35
Corrigall AV, Kirsch RE. Glutathione S-transferase distribution and concentration in human organs.
Biochem Int
1988
;
16
:
443
–8.
36
Hayes JD, Pulford DJ. The glutathione S-transferase supergene family: regulation of GST and the contribution of the isoenzymes to cancer chemoprotection and drug resistance.
Crit Rev Biochem Mol Biol
1995
;
30
:
445
–600.
37
Liu L, Zhou YH. Review and meta-analysis of glutathione-S-transferase polymorphisms and the risk of brain tumors.
J Neurooncol
2004
;
6
:
328
.
38
Okcu MF, Selvan M, Wang LE, et al. Glutathione S-transferase polymorphisms and survival in primary malignant glioma.
Clin Cancer Res
2004
;
10
:
2618
–25.
39
Cardon LR, Palmer LJ. Population stratification and spurious allelic association.
Lancet
2003
;
361
:
598
–604.
40
Aynacioglu AS, Nacak M, Filiz A, Ekinci E, Roots I. Protective role of glutathione S-transferase P1 GSTP1 Val105Val genotype in patients with bronchial asthma.
Br J Clin Pharmacol
2004
;
57
:
213
–7.
41
Olshan AF, Li R, Pankow JS, et al. Risk of atherosclerosis: interaction of smoking and glutathione S-transferase genes.
Epidemiology
2003
;
14
:
321
–7.
42
Rahman SH, Ibrahim K, Larvin M, Kingsnorth A, McMahon MJ. Association of antioxidant enzyme gene polymorphisms and glutathione status with severe acute pancreatitis.
Gastroenterology
2004
;
126
:
1312
–22.
43
Martinez-Lara E, Siles E, Hernandez R, et al. Glutathione S-transferase isoenzymatic response to aging in rat cerebral cortex and cerebellum.
Neurobiol Aging
2003
;
24
:
501
–9.
44
Weiserbs KF, Jacobson JS, Begg MD, et al. A cross-sectional study of polycyclic aromatic hydrocarbon-DNA adducts and polymorphism of glutathione S-transferases among heavy smokers by race/ethnicity.
Biomarkers
2003
;
8
:
142
–55.
45
Davis F, Ali-Osman F, Friedman H, Vick N. Genetic/Neruocarcinogen Risks and Outcomes for Brain Tumors. National Cancer Institute [2004 [cited 2005 Jan. 24]; Available from: URL http://spores.nci.nih.gov.